Sun Grid Utility Goes Live for Employees

museumpeace writes "CNET is reporting that Sun Microsystems turned on its Grid computing utility, hosting large ERP applications for its employees to test out the server infrastructure and user acceptance of the Computing-as-metered-utility model. General availability is scheduled for October. The rates? "Sun is offering processing and storage in a pay-as-you-go arrangement of $1 per CPU per hour, delivered via an Internet connection". Sun is still retooling its Thin Client interfaces and support SW. Experts quoted in the article wonder if Sun can make any money this way." Slashdot also covered the original announcement back in February.

Comments

So I could use the internet

[the_skywise]

to run multiple copies of Everquest simultaneously with bots and harvest tons of Gold!

Re:So I could use the internet

[infonography]

you mean mug people with the power of SUN! Where is my +8 Sword of w00tness?

Competition.

[Seumas]

I wonder how they plan to compete with the distributed/remote computing power provided by all of the unpatched and unprotected Windows based systems in the world that are freely available to anyone with a couple scripts and an internet connection?

$1/CPU/hour?

[Anonymous Coward]

Isn't that many times what it's worth? 365*24=over $7K for a year's worth of computing. Hmmmmm...

Re:$1/CPU/hour?

[Seumas]

While I haven't read the article, I believe that's $1 per CPU-hour.

If you're running a program that consumes 25% CPU, you'd be paying $1 for every four real-time hours.

Likewise, if you're running some massive number-cruncher, you may be using a dozen CPUs at 100% and paying $12/hr.

On the positive side, you're not paying for hardware or maintenance or facilities or manpower (directly) either.

Re:$1/CPU/hour?

[Neil Blender]

If your using 100% of the processor 24/7 for a full year, then no, it's not worth it. However if you need a little here and there, it can be worth it.

Re:$1/CPU/hour?

[Tango42]

Considering you could buy such a CPU (and the rest of the computer to put it in) for far less than 7k, yes, however you probably actually get 10 CPUs for 6 minutes, rather than 1 for an hour, so the processing is faster than you'd otherwise get. I can't see this being very popular though. The kind of people wanting to run CPU intensive tasks enough to be willing to pay someone for it are likely to be universities and companies that could get their own systems.

Re:$1/CPU/hour?

[pclminion]

Isn't that many times what it's worth? 365*24=over $7K for a year's worth of computing. Hmmmmm...

But say you wanted to run the job ten times faster. You'd split it across ten CPUs. Each CPU would perform 1/10th the work, but in parallel, so the job gets done in 1/10th the time. But the total number of CPU-hours you've used remains the same. So you pay the same price but get the job done ten times faster.

If you wanted to do that yourself, you'd have to buy 10 CPUs and once the job was done you'd have a bunch of CPUs you didn't need.

Re:$1/CPU/hour?

[dsginter]

Let's say that you have some number crunching that will take about 7000 CPU hours. Are you going to be happier waiting a year for your desktop to solve the problem or would you pay $7000 to get the answer in one hour?

Sun is betting that there are many people/businesses that fall into the latter category.

Re:$1/CPU/hour?

[TheRaven64]

Our 128 Xeon cluster cost around £100,000 (around $200,000). It has an expected life span of around 3 years, but will not be used 24x7 at maximum capacity. Assuming that it was (and assuming Sun's CPUs are the same speed), the equivalent from Sun would cost $3,363,840. Now even considering that our technician's time keeping the cluster running and up to date probably doubles the initial investment cost over the lifespan, and adding on air conditioning costs, it's still a good five times cheaper to buy our and operate our own system. Now, at 10 an hour, Sun would be the clear winner...

Re:$1/CPU/hour?

[Anonymous Coward]

With state universities cutting budgets right and left even with double-digit tuition increases year to year, the Sun Grid for on-demand stuff might be pretty attractive to the bean counters.

Re:$1/CPU/hour?

[Enrico Pulatzo]

Well, the article says $1 per CPU per Hour so that would be ($1 * time)/(# cpus) so maybe using more CPUs gets you cheaper computing :)

Re:$1/CPU/hour?

[Madd Scientist]

that is why sun is trying to build a system that they can't afford. a super system for the people who need a super system to share. and sun is exactly the people i would trust to build it.

Re:$1/CPU/hour?

[LurkerXXX]

We've gone over this before. The price isn't bad. You don't buy time on this system when you have one CPU's worth of stuff to compute and don't need it for a years time.

You buy time on it when you need a LOT of CPUs worth of stuff done NOW.

Imagine you have some projection software package that you need to run once a quarter for your company. You need the data within a week of the beginning of the quarter. You require 10,000 CPU hours to get the numbers all crunched. It's the only "big-computing" job you have.

On one computer the task would take you a little over a years time (8544 hours in a year). That won't quite be up to the task, remember you need the job done in a week. That's 10,000 CPU hours to fit into 168 hours of real time. You'd neeed 60 processors chugging away for those 168 hours to get it done.

How much is a 60 CPU cluster going to cost you to build? It's not insanely expensive, but it's not cheap. It looks a lot better to you to build that cluster than to spend $40,000 a year though! Right?

Wait. Clusters take up space. A 70 CPU cluster (better add in a few for redundency since this job has to be done in time) is not going to fit in the broom closet. That floor space is going to cost you.

Hmm, those 60 CPUs throw off a lot of heat when they run. Better add some more cooling to the building. Another decent expense.

Damn, look at that electric bill from the extra 70 CPUs and cooling for them. This nickel and dime stuff is starting to add up.

And now for the killer. You've got a new 70-CPU cluster. Your going to need someone to manage it. Cluster work is a bit different from what's what your used to, and your IT staff is already busy with their current workloads. It's time to hire a guy to manage the cluster. BZZZZZZZZZT. That hire alone makes the $40,000 a year for grid CPU time a deal.

Work the numbers yourself. It's not really a bad deal if you only occassionally need massive computing.

Re:$1/CPU/hour?

[hypnagogue]

Then follow that logic through: if Sun successfully establishes a market for leased CPU, then the Grid @ Home screensaver will come out offering to pay you some fraction of the proceeds to use your computer's idle cycles -- and presents that capacity to market at a lower cost than Sun's $1/hr.

Without a barrier to entry, this market will race to the bottom, and the price will settle out at around the amortized cost of the CPU + power and HVAC.

I think Sun has priced themselves out of the market before it even exists.

Re:$1/CPU/hour?

[pclminion]

Then follow that logic through: if Sun successfully establishes a market for leased CPU, then the Grid @ Home screensaver will come out offering to pay you some fraction of the proceeds to use your computer's idle cycles -- and presents that capacity to market at a lower cost than Sun's $1/hr.

You assume that all parallel jobs have a negligible communication overhead. Many parallel algorithms are communication bound, and these simply won't work when distributed across the Internet. Are you seriously comparing a rag tag assembly of 1000 random PCs tied together across the Internet, to a 1000 node hypercube with dedicated gigabit interconnects?

Re:$1/CPU/hour?

[TheSHAD0W]

You forgot the convenience factor. You'd have to wait a week, twiddling your thumbs, while that cluster ground away at the data. If there's a problem with the data, you may not find out until the end of the week, at which time your bosses will be pissed because you'll be telling them the projections will be delayed.

With Sun's service, you'll probably get the result within a few hours, not a week. If there's a problem the tests can be re-run with plenty of time before the presentation.

Of course, your bosses may be even more displeased about the extra $10,000 cost of the run than they would've been about another week's delay. Hope you talk fast!

Re:$1/CPU/hour?

[hypnagogue]

Are you seriously comparing a rag tag assembly of 1000 random PCs tied together across the Internet, to a 1000 node hypercube with dedicated gigabit interconnects?

Doesn't have to be ragtag at all -- if I was faced with a supercomputing task and had the option of buying a 1024 node system or leasing the time from Sun, I might consider buying the system if I knew I could recoup investment by leasing the time. Every prospective customer is therefore a prospective competitor.

When that happens the market clearing price will drop to amortized hardware cost plus utilities.

Re:$1/CPU/hour?

[pixel-fodder]

You are ignoring many of the costs associated with running your cluster (who pays your electricity bill?) but that's academic. You're probably right, for your particular use case it may be cheaper to buy and feed your own 'grid' but for many cases it doesn't make sense - if you want a large number of CPUs for a short period (ie. less than 100% utilization) then renting CPU time may make more sense. Also, you should seriously consider competing with Sun if you think you can charge 10c for the same service - competition will create a true commodity market. Rich (Sun employee)

Re:$1/CPU/hour?

[Asgard]

Don't forget to add in the time to get the software package working every quarter on the cluster. It doesn't make a lot of sense to pay for year-round storage for an app that runs only once a quarter, but changes in the cluster enviroment will probably require some sort of maintenance to the app before each run.

Re:$1/CPU/hour?

[Anonymous Coward]

"I think Sun has priced themselves out of the market before it even exists."

Er, uh, you do understand Sun can set their prices to meet the market, right?

Being the first to drive very large scale managed and virtualized compute services...Sun is initiating the market price at $1. That will certainly change in time.

Re:$1/CPU/hour?

[Saanvik]

Actually I think this might be very appealing to research groups at universities.

As part of your grant proposal you include a flat cost for computer time rather than costing out hardware purchases. Not only that, but you can also start your project as soon as the money is approved, you don't have to go through all the hoops to buy, ship, house, and administer the hardware.

Re:$1/CPU/hour?

[bradleycarpenter]

Ok, problem here. Do you really think a large research institution or someone doing very sensitive important work is going to trust a grid of millions of desktops to process his work. How can he know that these systems are secure? With the Sun Grid you know the data is on the grid. Not on Barrys computer that has 500 viruses and hasn't been updated in 5 years.

Re:$1/CPU/hour?

[drsquare]

If Sun can make a profit on $1, then it obviously costs them less than $1, in which case you can do it for less than $1, and so there's no point in using Sun's service. The 'hire' wouldn't be needed all year, as the cluster would often be sitting dormant.

Re:$1/CPU/hour?

[LurkerXXX]

Sun can make a profit on $1 because they can accumulate enough customers to keep their cpu busy most of the time. For a company that only keeps the CPU busy 5% of the time, that 95% idle time makes the CPU not worth buying, but worth renting occasionally.

If you don't pay your 'hire' all year, you are going to need to find a contractor willing to deal with large gaps of time in between his use.

Now imagine you are the contractor. If one company that is only going to use you for 1 week every three months, and a gig comes up that is offered to you for 6 months of solid work, with no breaks, which customer are you going to drop for the other? Probably the one that will pay for 6 solid months of work NOW rather than the one that will pay you the same amount of money, but only sporadically and spread over the next 6 years. A company contracting out that 'hire' will likely find themselves going through a string of different contractors, one every or every few jobs run. Each one having to be brought up to speed on the new-to-him cluster.

Sending a check to Sun is going to be much more reliable and more cost-efficient.

Re:$1/CPU/hour?

[Anonymous Coward]

If my phone company can make a profit on $0.05/min, then it obviously costs them less than $0.05, in which case you can do it for less than $0.05, and so there's no point in using the phone company's service. The 'hire' wouldn't be needed all year, as the phone lines would often be sitting dormant.

If my ISP can make a profit on $10/month, then it obviously costs them less than $10, in which case you can do it for less than $10, and so there's no point in using the ISP's service. The 'hire' wouldn't be needed all year, as the Internet connection would often be sitting dormant.

If my gas station can make a profit on $2.50/gal, then it obviously costs them less than $2.50, in which case you can do it for less than $2.50, and so there's no point in using the gas station's service. The 'hire' wouldn't be needed all year, as the fuel pump would often be sitting dormant.

And so forth.

Sun ain't stupid, you know.

Re:$1/CPU/hour?

[Jimy]

hmmm...i think you are right. The way i like to put it. Its like you are renting out a really expensive car for some special occasion. You can't go on for ever with a rented car. It is only to be used for some specific purpose. The big questions is actually about the market scenario. For sun to make profit on this strategy it needs to know how much is the market out there who want to rent a limo like this one.

Re:$1/CPU/hour?

[Madd Scientist]

i've done a couple of computer/math research grants, and i can't think of any CPU based research where you would know how many CPU cycles you'd need before your research had even begun. if you could calculate that, then you could have just as easily calculated your original problem.

Re:$1/CPU/hour?

[dexomn]

They have to pay for all those punch cards somehow.

Re:$1/CPU/hour?

[Madd Scientist]

well... maybe not... like if you knew algorithm x could take in dictionary of size y and solve it using a function of y which would give you CPU cycles needed, then maybe... but thats a fairly specific example. if most projects are running the same code on different data sets, then maybe they could estimate based on their dataset if it was applicable.

Re:$1/CPU/hour?

[Anonymous Coward]

That's like renting a movie for a year instead of just buying it.

Re:$1/CPU/hour?

[W12x40]

This is great if you only do modelling now and then. If I do a really big STAAD model, say over 100K nodes, it shuts down my desktop for 60 to 90 seconds - nothing else can be done concurrently. Multiply that by fifteen iterations per revision, and you really can't get anything else done in the background. If the system does what I tink it does, I upload the first version of the model, run it in some sort of browser window, minimize and go back to CAD/Excel/Mathcad, whatever. This assumes that when I send the data, I get instant CPU time.

We don't need a big box in-house because those sort of big projects only pop up every month or so. I'd save four billable hours and charge $4 to the purchasing card.

The target audience might be overflow from big research projects and universities, but also small jobs from outside the research realms.

Re:$1/CPU/hour?

[ghukov]

also consider the price of maintenance and repairs, and having to pay someone to take care of the cluster. Not to mention the cost associated with powering, cooling, and securing a cluster.

Re:$1/CPU/hour?

[Donny Smith]

I see no advantage in either case (over existing solutions, that is).

If you look at a basic mail/Web/mySQL server with some 200 emails and 2-3,000 hits/day, it probably consumes about 2-3% of CPU resources (average figure).
That'd be $1*2%*24hrs = 48 cents a day or some 150 bucks a year. I'm sorry but that's already available at any hosting company at approximately the same price.

And let's not be naive - I'm sure somewhere in the fine print it probably says you must cough up X bucks for setup and buy a minimum of Y time units, whereas many hosting companies allow monthly extensions and charge nothing for account setup.

I'm lazy to calculate now, but at higher percentage of CPU utilization Sun's deal is probably even less attractive.

Re:$1/CPU/hour?

[Anonymous Coward]

computing power free @ planetlab

Re:$1/CPU/hour?

[Neil Blender]

And let's not be naive

Yes, let's not. The world is made up of more than simple web hosting and mysql and blogs. Some businesses use more computing power in 1 minute than you will in your entire life. And sometimes it make senses to rent a huge amount of processor time for a short interval rather than setting up the infrastructure to accomplish the same task. It's the same as you renting space in hosting company rather than colocating your own box. Only on a much larger scale.

Re:$1/CPU/hour?

[jetmarc]

> ... or would you pay $7000 to get the answer in one hour?

What if I spend the $7000 hour fixing bugs of the software that doesnt scale as expected?

Honestly, which real-world problems are solved in this manner anyway? Except for cryptography I fail to come up with any.

Re:$1/CPU/hour?

[Donny Smith]

>on a much larger scale.

I am aware of that, I was commenting on the small-scale example from grandparent post.
As far as large scale computing is concerned, I believe that vertical solution providers, specializing in particular industries like oil & gas, biotech, etc. will be able to provide a better solution for large scale customers as they have domain know-ho.
And xSPs and hosting providers can sell unused compute power to grid service providers at a very low cost. I do not believe Sun can run data centers more efficiently than RackSpace, GoDaddy and other guys with many years of expeience in huge scale datacenter management.

Re:$1/CPU/hour?

[pclminion]

Doesn't have to be ragtag at all -- if I was faced with a supercomputing task and had the option of buying a 1024 node system or leasing the time from Sun, I might consider buying the system if I knew I could recoup investment by leasing the time. Every prospective customer is therefore a prospective competitor.

I think this view is too optimistic. The price can't drop to the "amortized hardware cost plus utilities" because there's more to it than that. A cluster computer is not like a piece of artwork which you can just rent out to anybody who wants it. You need to maintain it, fix parts as they fail, pay an employee to do this for you, build an accounting system to track the money flows, pay for marketing and sales support in order to develop contracts, support clients when their shitty code breaks, etc. You basically have to go into the business of being a cycle provider.

It doesn't follow that everybody who requires supercomputing services is necessarily interested in entering the supercomputing business.

Everything Old is New Again

[Anonymous Coward]

This is just the reincarnation of the mainframe era. Everyone (Sun, MicroSoft, et al), want to put us back in the days where the storage/cpu and most importantly the applications themselves are in their "capable" hands.

I'm not even going to enertain the idea of having MY data stored on another (microsoft/sun/etc)server, and paying for the rights to access/modify it.

There is a reason it's called the PC,and not a dumb terminal.

Re:Everything Old is New Again

[Kjuib]

What if we changed the rules a little... convert all the PCs everywhere into terminals... so we can connect and log into our PSs (Personal Servers). We can all have them at home with our own info on them, and can access them anywhere by anyones terminal.

Re:Everything Old is New Again

[Anonymous Coward]

Sure, it's a nice concept what you describe, but omit the peronal in Personal Server. You don't own it. Not the hardware, nor the software.

Backwards compatibility afer vendor installs new common application, maybe? maybe they will charge you for "upgrading" your data to the new standard. (not open of course).

Want to install some new software on your "personal server", better get permission.

I'll stick to my desktop, and being able to ssh to my box from anywhere.

Re:Everything Old is New Again

[guaigean]

I can already do that via ftp, http, and ssh in via putty. Why do I need a corporation to charge me for it? (Yes, I realize that Joe Sixpack can't do this YET, but it's not far off with the number of gui tools out there)

Re:Everything Old is New Again

[Anonymous Coward]

Then why don't you store your savings under your matress? Wait...you mean you pay a bank to keep your money and manage it in your accounts? And you have to go to them to get access to it? Oh the humanity!

Also, once ZFS comes around for Solaris 10, the Sun Grid will be orders of magnitude more reliable than your PC. At least for your data, as the weak link will always be the DSL/Cable Modem/T1/whatever people use for Internet access.

Eventually people will not have any mechanical devices in their PCs...everyone will pretty much have stateless thin clients to access their data services.

Re:Everything Old is New Again

[oldmanmtn]

There is a reason it's called the PC,and not a dumb terminal.

There are no dumb terminals - only dumb users.

This isn't targeted at PC users. This is for (for example) the hedge fund that needs 50 machines for 8 hours, once a week, to run a complex model. This gives them the power they need for a fraction of the price of the raw hardware, and they don't have to pay anybody to maintain it.

I've had projects where I really wanted 1000 CPUs for a week, just so I could do scalability testing. There's no way we could afford $1,000,000 to buy 1000 machines just for that one test, but we could probably have swung $50,000 to get them for five 10 hour days or ten 5 hour days.

Re:Everything Old is New Again

[Alwin Henseler]

This is just the reincarnation of the mainframe era.

And what is wrong with mainframes? Putting all the computing power in one place, is about sharing that equipment between all users, making optimal use of the hardware. As opposed to everyone having their own box that comes with everything and the kitchen sink, but sits around doing nothing 98% of the time. Talk about waste...

I'm not even going to enertain the idea of having MY data stored on another (microsoft/sun/etc)server, and paying for the rights to access/modify it.

Agreed... but I can think of many compute-intensive tasks, where the privacy implications aren't that important: scientific research, running simulations, design/engineering, etc. Besides: if that 3rd party (on purpose) leaks or abuses confidential data that you trust with them, then you know who to sue for damages. And think of what running any software/OS with an automatic update mechanism enabled, means here. Kind of the same: to a certain degree, you simply trust that 3rd party.

There is a reason it's called the PC,and not a dumb terminal.

In my view PC's haven't gotten all that much smarter in the past decades, seeing how ordinary users struggle with them every day. And not faster either, only very efficient at wasting CPU cycles on eyecandy and countless software layers. Don't get me wrong here: you can do lots of wonderful things with PC's today that weren't possible 10 or 20 years ago, and I wouldn't want to swap mine with a vintage machine for doing everyday work. But just FYI: for example a meager 3.5 MHz. Sinclair ZX Spectrum boots up in what, 1 second? Doesn't need updates, doesn't have virusses, and even if it's the first time you've ever used a computer, you can learn how to program and run a "Hello World!" thingie on it in under 5 minutes. Back in the 80's, kids (myself included) used to do this with machines that were on display in shops. Now try those things with your brandnew 3+ Ghz. machine.

Back to that mainframe issue: I can see lots of people who would love an appliance-like PC, with software on it remotely managed. Think: by your local ISP, by a user group of your choosing (like a Linux distro), you name it. With your private data stored in any location of your choosing, locally or on some other remote system.

Just putting computing power remotely as well, doesn't make much sense to me for most scenario's. The speed at which computing power becomes cheap for the masses, makes that a niche market for selected applications. As always: pick the right tool for the job.

Re:Everything Old is New Again

[JPriest]

People don't like to go backwards. Would you trade your PC for one? No? then stop arguing for the idea, if you don't want to do it then don't expect anyone else will want to either.

Re:Everything Old is New Again

[That's Unpossible!]

Hardly "insightful." I give the parent a -1 Shortsighted.

This is just the reincarnation of the mainframe era. Everyone (Sun, MicroSoft, et al), want to put us back in the days where the storage/cpu and most importantly the applications themselves are in their "capable" hands.

The application is in the hands of the person creating it. You are purchasing scaleable processing power.

I'm not even going to enertain the idea of having MY data stored on another (microsoft/sun/etc)server, and paying for the rights to access/modify it.

Good to know. Too bad it's not what grid computing is concerned with. At all.

There is a reason it's called the PC,and not a dumb terminal.

Honestly... +5 Insightful? "The mods are crazy."

Re:Everything Old is New Again

[fracex]

Where's the (-1 paranoid) option?

Re:Everything Old is New Again

[pdo400]

I don't know if you're a simple troll or a simple-ton, but either way your logic is utterly flawed.

I work producing a software product that I personally wouldn't use. Does that mean nobody would use it? Apparently not since my company is profitable and I get paid.

Also, 'backwards' is a matter of perspective. For ordinary home users who have trouble keeping their desktops clean and their start menus in order, don't keep backups of their data, and throw away their PCs every couple years because of viruses and spyware, a thin client solution would certainly be considered a step 'forward'.

Re:Everything Old is New Again

[DAldredge]

Sun sells those too, they call them SunRays.

Re:Everything Old is New Again

[serialdogma]

Should that not be +1?

Re:Everything Old is New Again

[Mac Degger]

Yeah...but I'd sure love to use this instead of losing my home pc for a week when I want to render out one and a half minutes of 3D.

Re:Everything Old is New Again

[JPriest]

If they are remotely using another environment what makes you think that environment cannot also aquire a virus? You can restrict their permissions, but you can do that on a PC too.

You have to give them enough access to install their own programs and read email attachments or you are limiting yourself to 5% market share from the start. If you give them admin access on platform you are endangering everyones data, not just theirs.

The idea that everyone not posting to slashdot is some clueless noob that can't point and click is your miscalculation, not mine.

As PC's get cheaper people are only moving further away from thin clients. If people are so willing to use a thin client then why have all the attempts to push them failed so miserably?

Re:Everything Old is New Again

[Anonymous Coward]

I'm confused. Is that an Imperial simple-ton or a metric simple-ton? Or perchance a complex-ton after all?

OpenOffice

[gkozlyk]

Maybe it can be used to improve OpenOffice load times.

Re:OpenOffice

[Anonymous Coward]

I think the UNIX linker is a linear process, so, sorry, you still gotta wait for OO.org. However, you can customize the splash screen with your favorite nudie image to make the time more enjoyable.

Seti and Folding @ Home^h^h^h^h Datacenter

[paulius_g]

Yah! Now I can get my stats back up. /me cracks knucles

Ouch that hurt... Sun's spirit will always be amongst us.

Re:Seti and Folding @ Home^h^h^h^h Datacenter

[Anonymous Coward]

You wouldn't beleive how close to the truth that actually is ... 415,000 credits after less than 5 days computation ... ;)

Gambling?

[weilawei]

Now, I can finally run gigantic neural networks trained by genetic algorithms to predict the outcome of sports betting. :) I actually gave up the project (I was coding it in C using the HDF5 library for data storage, etc.) because I have no hardware capable of running it reasonably. This brings some nice computing resources well within reach of individual hackers. I don't need a 10,000 CPU block for an hour slice, just a hundred CPUs for an hour.

Re:Gambling?

[cloudmaster]

If you just need a hundred CPUs for an hour, couldn't you use one CPU for 100 hours? I mean, that's only a little over 4 days, or ten days of running while you sleep, or about two weeks of running while you're at work.

Re:Gambling?

[weilawei]

I actually had plans of using the system for something slightly better than theory if I could make it work. $100 investment, plus bet...

Decision still to be made on name

[Timesprout]

But CloudNet, AirNet and UpperAtmosphereNet are currently leading contenders. Analysts feel however it may be something in a similar vein which is finally chosen.

Re:Decision still to be made on name

[Anonymous Coward]

First of that joke to actually be funny. Nice.

Re:Decision still to be made on name

[Seumas]

Arnold jokes aside...

My vote is for HairNet

The economics of the situation

[Anonymous Coward]

This might be a good idea if you need a lot of computing power only occasionally. If your needs are fairly constant, you can buy a heck of a lot of computer for a buck an hour.

Am I missing something here? This just doesn't seem like it makes sense for most people/organizations.

Just to get it done away with...

[Eberlin]

With a Microsoft partnership, we now understand how we plan to have the oomph to run Windows Vista when it comes out.

$1 per CPU per hour...the true money-making scheme here is that if you run Linux, they'll charge you the $699 for each processor on behalf of SCO.

So for $50 bucks an hour, you can run a SWING application almost without a performance drop.

With the licensing model, you can run apps with it, but you can't alter any data that passes through without our permission. Want to see the results of your calculations? You'll have to sign an NDA.

Sun Microsystems? Hey weren't they that big dot-com company that wanted programmable toasters? What are they doing these days?

sun's shining

[dankelley]

Hey, look. Sun has a new idea. Hm, if they get 9 more, maybe their stock price will reach the price I paid for a nanosecond, so I can dump it :-)

Seriously, I like the sound of this. One can argue about costs, etc., but at least they have something other than inertia that might encourage a scientific user to choose Sun.

Re:sun's shining

[Anonymous Coward]

The idea is not new, it's just a rehash of a bunch of old technologies, packaged up in a new shiny ribbon.

Re:sun's shining

[Berner]

I think they're going to make loads of money on this... Say that each of these processors sit in an individual blade, each blade costing between $3-4000. If this blade gets hired more than 10 hours/day you will probably get the money you invested back inside a year. Even if the processor isn't used that much you will make money off of it since a Sun machine is good for years and years of continuous operation (we have a couple of dozen at school) you have the time to make that money.

Not for big problems, then

[Space cowboy]

Considering that a "CPU" can be had for $400 (2.8GHz Celeron D without even trying, just a search on google).

So 24 hours a day, $400 -> 16 days work. Let's add in 25% for "stuff" (electricity costs, etc., being generous...) and you're still saying that a problem that takes 20 days or more, you're better off buying a throw-away PC and running Linux on it.

So, it must be aimed at the smaller problems. Like what ?

Simon

Re:Not for big problems, then

[Keeper]

Actually, I would imagine this would be more useful for solving larger problems that are run infrequently, where you want do do a distributd task once a month that takes 1000 machine hours and get back a result in 1 hour.

Re:Not for big problems, then

[Paul Jakma]

For the same $400 you could get 100 CPUs for four hours. If your problem divides up reasonably well, then instead of spending your $400 CPU and waiting 16 days, you could instead get your answer in hours.

Maybe you could do it cheaper by buying your own CPUs, but you could be waiting two weeks for your Dells to arrive. How much is it worth to you to get an answer within hours or days versus a few weeks or months of waiting?

--paulj

Re:Not for big problems, then

[Wesley Felter]

Your costs are way off; try adding 250% for other stuff (especially management).

Also consider the case where your problem needs 20 CPU-days -- in 8 hours.

Re:Not for big problems, then

[Glonoinha]

If you have a problem that takes 400 CPU hours to run, your answer is either inanely worthless, or mind-bendingly valuable (I needed to throw one of those in there for the SETI group, but I won't say which.)
Well that or you need to optimize your code, or get a faster machine.

That said, it probably isn't worthwhile to the guy with a $400 problem - more likely they are looking to appeal to the kinds of guys that want to crack 128-bit encrypted data streams in real-time, or run two neural networks against each other in a zillion games of chess in order to teach (evolve) their neural network, or crunch two terabytes of data picked up by an Indy race team over three days at the track. Brute forcing 1024-bit encryption is totally possible, but the data isn't generally valuable a thousand years after you start decrypting it. Throw enough horsepower to decrypt 1024-bit RSA in real-time and you will find yourself rich (or dead.)

Knowing the winning numbers to the lottery thirty minutes after they are announced is pretty worthless.
Knowing the winning numbers to the lottery thirty minutes before they are picked is worth a hundred million dollars.
Amazing difference having the answers an hour earlier makes - I'm not saying that these computers will give you that much of an advantage, but I'm still saying ... I currently work on problems where an hour difference in processing time can make a single data-crunching run cost about an additional $100,000.

Re:Not for big problems, then

[Space cowboy]

You know, I've got several linux boxes. I don't believe any of them cost $100 for every 20 days runtime! As for 250%, Oh boy! I have a bridge to sell you!

And, if it's my computer, what management are we talking about ? It's a program running on a computer. I start it. I wait. I analyse the results. What's to manage ?

If I buy 100 of these things, you use a simple batch script (I wrote one at college in about 2 days). Typing 'batch ' was all that was required to start something. Typing 'batch list' gave you a list of all running programs on the cluster, and typing 'batch progress []' gave you progress reports (written by the app into the file 'progress' in outputdir) for one, many, or all programs on the cluster. Similar commands for 'kill' 'stop' 'cont' etc. Easy. Our nodes were DECstations but I'm sure Linux is equally amenable...

My PhD was in Neural nets, we used to have nets that would take 20 hours or so to train, you would do that 10 times to make sure your results weren't anomalous, and then the input variables needed to be changed, and the whole thing re-run. This could quite happily take several weeks.

Now I can see that we could have rented time on the grid, but my point is that our use would categorise as continuously-required, and I don't think the sums add up.

So, assuming you huge problem is *really* parallelisable (lots aren't) you can split it over several hundred nodes and get an answer back relatively quickly. But most of us who need answers to such problems need to do this over and over again, so again I ask - where does the break-even point happen ? I suspect it's lower than Sun think it is...

Simon.

Re:Not for big problems, then

[Paul Jakma]

I don't believe any of them cost $100 for every 20 days runtime! As for 250%, Oh boy! I have a bridge to sell you!

You're not paying for and hence you do not have:

- data centre floor space with:
    - heavy duty UPS
    - generator backup
    - climate control
    - security
    - redundant networking
    - multiply redundant storage
    - tape backup silos / HSM
- The 24x7 staff to:
    - monitor security
    - test the generator weekly
    - monitor the backup processes
    - monitor and maintain the network
    - monitor and maintain the hardware

etc. etc. If you think your costs as "Joe Bloggs the guy who runs a few Linux PCs at home" are comparable to a corporate affair then you're simply kidding yourself, particularly when you're not billing yourself for your own time ;), and your SLA with yourself is pretty flexible and forgiving ;).

A lot of corporates have thought what you thought "Ah sure, it can't be expensive to run a few servers in our own 'data centre'", and they typically either under-estimate the costs, or they end-up with very shoddy server facilities. Then they'll have reliability problems due to:

- servers overheating cause they're stuffed into a cupboard (seen this)
- lack of staff expertise (all too common)
- utilities failures (they couldn't afford the large UPS + diesel generators + cut-over switches + electricians expenses)
- the gradual increasing burden of maintaing installed plant, which if not planned professionally slowly but surely turns into a huge sprawl of unmarked cables, till it gets to point even simple rewiring tasks are a massive (and error-prone) undertaking.

Eventually, to a lot of these types of small corporations, locating in a managed data-centre and letting someone else take care of the details becomes very very attractive. (particularly for coporates whose primary business is *not* computing).

You are almost certainly underestimating the costs.

--paulj

Re:Not for big problems, then

[Space cowboy]

Look, what part of this is unclear ? I am not trying to compete with Sun!

In the university environment I mentioned above, they already have all that. They were paying for it regardless of if I push 10 boxes on the shelves in the spare rack-space and wire them up to the switch. Which is exactly what I did. With their blessing. "It'll be lost in the noise, go right ahead"...

So. Zero extra cost apart from electricity.

Most corporates have an IT dept as well, and I would expect the same to apply.

In the personal space, I put 10 machines in the spare room and leave the window open (it has a grill). Zero extra cost apart from electricity. Sure we can all make up costs and "apply" them, but for the case of an employee/student/individual just wanting "to do it", I can't see the big deal. Buy the machines unless you're highly parallelisable.

Simon.

Re:Not for big problems, then

[museumpeace]

economics and scalability are not that simple to mix. Your $400 machine simply can't handle the dataset or the array sizes or the threads or much any metric you'd get when you size up any number of huge but very practical programs that businesses sometimes run. To name but one example: the monster linear systems that model supply chain and what-if a huge number of variables so the company can find the right product mix...the list of such programs is bigger than you may be aware.

Re:Not for big problems, then

[Paul Jakma]

Look, what part of this is unclear ? I am not trying to compete with Sun!

Didn't say you were, you were however trying to extrapolate costs from what you think your home costs are.

They were paying for it regardless of if I push 10 boxes on the shelves in the spare rack-space and wire them up to the switch.

And that slack capacity, in terms of floor and rackspace and network, has a definite cost. So does planning ahead in order to balance cost against sufficient slack (people's time has a cost).

So. Zero extra cost apart from electricity.

Uhuh, that's cause that data-centre is free. As are the admins who run it, no doubt. All free.

Most corporates have an IT dept as well, and I would expect the same to apply.

Yep. IT departments come free when you register a company, with special pixies who never need to eat , sleep or have any life beyond being IT pixies. All free.

I put 10 machines in the spare room and leave the window open (it has a grill). Zero extra cost apart from electricity.

There is a cost, that room is a cost for one. You could have bought an N bedroom house instead of an N+computer room house and saved yourself 20,000 or more - which would be sufficient to lease a rack from a co-lo for 5+ years. Alternatively, maybe you'll need to go buy a N+1 house now, cause you need another spare room (eg, you get married and have a kid, or whatever), now you need to pay for moving (costs your own time if nothing else) as well as the more expensive house.

And again, the computers in your house could not meet the kind of reliability guarantees many businesses want, or expect (even when self-hosting). Go spec redundant power for your house and then think again about $1/hour. And again, your time isn't free either.

Sure we can all make up costs and "apply" them, but for the case of an employee/student/individual just wanting "to do it", I can't see the big deal.

So while you feel qualified to comment on whether or not $1/cpu/hour is a fair price or not, you don't think it's actually necessary to include any actual real costs other than electricity in determining what such a fair price is?

It's obvious you work in acadamia ;).

--paulj

PS: I'll be dropping around a few computers to your house later on. If you could put them in your spare room, hook them up to a UPS and give me a key to your house for access and/or field calls at any time and do maintenance for me on them, that'd be great. Cause you sure come at a /great/ price..
I'll pay for your electricity of course. ;)

Re:Not for big problems, then

[Anonymous Coward]

The $4000 machine I am building as soon as I sell my house will be able to handle the data set , array sizes and threads this year, in a decade the same machine will be $400.

It is just mind boggling how powerful the machines are getting to be.

Re:Not for big problems, then

[Anonymous Coward]

I currently work on problems where an hour difference in processing time can make a single data-crunching run cost about an additional $100,000.

Uh huh, and you bought what your bosses were selling?

Re:Not for big problems, then

[Anonymous Coward]

Most corporates have an IT dept as well, and I would expect the same to apply.

Which are busy doing other things and whose personnel may lack knowledge in setting-up and managing clusters of computers. What if the person does exists and amazingly isn't already busy with other work, what happens when he/she leaves (say it's a small company)? Now you need to quickly go and hire someone else who can perform his cluster-management job, in addition to his other duties, which may cost extra. Not to mention that in the meantime you may have a problem, I believe third-party consultants/tech-support run $100/hour up.

Now, you may think that such a task would not take up much time, and it wouldn't if you don't give a rats ass about anything. A corporation may not have that option, if that clusters doesn't work properly for some reason then the costs of the delay may outweigh any savings you get from having your own ghetto-cluster. A corporation may also not be able to wait weeks while their program finishes running as that delay would cost money in essence. Potentially, if its not done within a day you may as well not do it as in which case you'd neither a very big ghetto-cluster.

In summary: to a corporation reliability may be more important than cost (have you seen the triple-everything redundant servers you can buy?) as the cost of failure may vastly outweigh that extra investment.

Re:Not for big problems, then

[afidel]

400 CPU hour problems can fit in to lots of other categories. Weather prediction, where the results have to be out in hours but could easily take in excess of 400 CPU hours are an example. Another good example is chip routing, at a previous job the chip routing could take days or weeks on multi-cpu boxes cost tens of thousands of dollars, the results were valuable, but not mind-bendingly so. Another good example might be special effects for a mid budget film, buying the hardware to get results in a short enough time to get a return on capital would take too much capital to be justified, so you simply lease CPU time and get the movie out in weeks or months instead of years or massivly overbudget.

Re:Not for big problems, then

[Space cowboy]

I try to avoid using sarcasm in posts, but for some reason, the "pixies" irritated me...

It takes tremendous resource-planning - this 20-day project. I mean, there's absolutely nowhere that any multi-national organisation, university, home, garage whatever could possibly put another computer. Christ no. Everywhere is *completely* budgeted for, for that 20 day period. For the space of a desktop PC. Of course it is.

And (to make sure those 20 days are properly accounted for) I'll have to employ at least a dozen busybodies to run around, make up even more fictional costs to justify why they're there. Hey look - even the costs have costs. For 20 days. There's no way I could plug it in myself (!) and run the program, then switch it off. Just no way at all!

Oh and by the way, [laugh, snort!] I *certainly* do *not* work in academia. [laugh, again!]. I *do* think you've been eating some of that pixie dust though - I've bought and sold companies (2 of them actually), and I've never come across IT pixies. I've never had a problem finding a home for an engineers project either, if there's a potential gain for the company. Or if it's neat. Especially if it only lasts 20 days.

I've never yet seen a business plan for a computer-room that didn't have huge amounts of spare capacity built right in. And making use of it (for 20 days) seems like a good thing to me, not a bad one.

Actually you're welcome to put stuff in my spare room (for 20 days) if you like (there are 2), but you get no access, no backup, no reliability guarantee, no redundant power & no maintenance. I charge $1000/day. That's because it's my house not my business, mind. If I was setting up in business, I wouldn't be using my house...

The reason I'm emphasising 20 days is that if this is the break-even point for 1 project, you can give away the computers afterwards to schools/colleges/community. No long-term issues there. Or (more likely in a corporate culture), you could re-use them on the next project - at which point the grid costs are even worse of course, but then there are some logistical costs. Storage is cheap when the boxes arent' being used though...

The problem is you're seeing it as a long-term plan to have this resource lying around, needing maintenance, running continuously, etc. etc. I'm seeing it as "get in the car, buy the box, run the program, ditch the box (sensibly, of course)". If the cost is (in the end) the same, why keep the boxes ?

Simon.

Re:Not for big problems, then

[Paul Jakma]

I try to avoid using sarcasm in posts, but for some reason, the "pixies" irritated me...

Ah, sorry, I should have put in a smiley. It was intended as good-humoured banter. My posts generally should be taken as such.

Everywhere is *completely* budgeted for, for that 20 day period. For the space of a desktop PC. Of course it is.

I did not say "budgeted for", I said it had a cost. Whether you track that cost and budget for it or not is a different thing. You may feel it well-worth it to simply provide for lots of slack in terms of resources rather than spend lots of time accounting and micro-budgeting for everything. And you could even be right. But those costs remain however, whether you count them or not.

However, if you're planning on setting up a "utility computing grid" then you need to determine exactly how much those costs are, and then you *do* have to go and meticulously examine and account for every little cost, as much as you can. Otherwise, you'll never know how much to charge. ;)

I've never come across IT pixies.

Nor have I, that's my point. If you find any, let me know ;). But round here, people don't do things for free - either because they get money or some other benefit. ;)

I've never yet seen a business plan for a computer-room that didn't have huge amounts of spare capacity built right in.

I've definitely seen computer-rooms that were stretched to capacity and in a pretty poor state, because they were badly planned or had out-grown the original plans. Much pain.

On the flip side, if you massively over-provide, then you're wasting money, as in two to four years your kit which you never really used will be obsolete anyway. Pure cost. It's *really* easy to waste massive amounts of money on a regular basis in a corporation if budgets are not carefully controlled and examined (but OTOH it's really hard to control budgets finely while still retaining flexibility and avoiding red-tape) - amounts large enough to start affecting your bottom line significantly.

Course, you never notice, because you don't track capacity and budgets carefully enough. ;) Until your competitors start eating you (better margins + lower prices -> they take your customers away from you, and you can't comprehend it's partly cause you're sinking significant quantities of cash into black-holes).

making use of it (for 20 days) seems like a good thing to me, not a bad one.

Yes, it is. If you have it, use it. But grid computing is aimed more at those who don't have it, who want it soon, and don't need it continiously. If you happen to keep a bunch of racks of clustered 1U dual Opterons + high-bandwidth interconnects + redundant SAN spare for when you think you might need them, then Suns' grid is not for you. ;)

If the cost is (in the end) the same, why keep the boxes ?

If you don't even take the time to work out what your own costs /really/ are, then you have no basis for determining whether or not the cost actually is the same, or less or more.

You simply don't know. And I still think you greatly underestimate the true costs of IT. ;)

--paulj

Re:Not for big problems, then

[Zemplar]

Not to mention that anyone that would use the Sun Grid would even THNIK about running their simulations on equipment that doesn't use ECC memory!

Re:Not for big problems, then

[nonlnear]

Uh huh, and you bought what your bosses were selling?

In any financial sector where real time commodity (or derivative) pricing is needed, then being an hour behind the market could easily cost millions - or more. The $100,000 he quoted is peanuts.

Re:Not for big problems, then

[redwyrm]

more likely they are looking to appeal to the kinds of guys that want to crack 128-bit encrypted data streams in real-time, ...

STFU.

Re:Not for big problems, then

[tfiedler]

>>How much is it worth to you to get an answer within hours or days versus a few weeks or months of waiting?

How many answers really need to be found in hours or days? I manage a couple of clusters and while they are nice for finding answers on problems that take weeks to compute, no one really looks at the data the second it finishes. Sometimes, it'll be a week after before anyone looks at it.

Having said that, I wouldn't mind using Sun's grid for the simple fact that doing so would let me force graduate students (they write a lot of the (sloppy) software for the different principal investigators) to write to tighter standards!

Re:Not for big problems, then

[Paul Jakma]

Zero extra cost apart from electricity.

Btw, see this interesting blog about the cost of just electricity in a big data centre:

http://blogs.sun.com/roller/page/marchamilton/2005 0829

Why am I not impressed?

[RomanySaad]

OK... that comes out to $1600/yr per computer for a business... how much is the storage?
What about security?
And also why not "follow the power analogy entirely" and charge based on computer power/hr?
It's too expensive... rather than pay $1000-$2000 every 2 years, will businesses pay $3200... why?

Re:Why am I not impressed?

[jasongetsdown]

why?

Simple. You'll never upgrade your hardware again. Or at least you'll only be upgrading cheap thin clients. But then again you'll be paying the annual equivalent of a new compy per user. If you subtract the associated costs of maintaining all that complex equipment vs. thin clients it might be well worth it.

My question is how do you deal with the latency? Is display over IP really that sprightly? And what happens when I want to put my 300meg photoshop document on a CD?

Who is going to use it?

[GGardner]

Seems like the main categories of potential users fall into two camps:

The high-energy physics folks, who generally get government and university subsidies for their high-performance computing needs, and so certainly get computation much cheaper than $1/cpu-hour.

Commercial folks, maybe in the financial services sector who are (rightfully) paranoid about security, and just aren't going to send their sensitive data from Wall Street to California, so matter how much SSL-this and triple-DES that happens on the way there.

Re:Who is going to use it?

[Neil Blender]

Several years ago at the company I worked, we tossed around a few ideas that would take massive computing power (probe design, for one) We realized our end product might take 100+ CPUs 6 weeks to do one set of work and could only do one set at a time. We had a joke that this might be viable business - if we had a time machine.

In the end, it was decided (quickly by some, slowly by others) that it would be foolish to pursue it given the infrastructure we would need start off with. Givin this sort of service, it might have worked. Pay-as-you-go, so to speak.

Re:Who is going to use it?

[K8Fan]

What about 3D rendering? There are lots of people renting time on render farms right now to make deadlines. If I can run PRrenderman on these CPUs, and the price includes storage for my rendered frames, it might be price competitive to buying a lot of processing speed that I'm not going to be using 24/7.

Re:Who is going to use it?

[Anonymous Coward]

Commercial folks, maybe in the financial services sector who are (rightfully) paranoid about security, and just aren't going to send their sensitive data from Wall Street to California, so matter how much SSL-this and triple-DES that happens on the way there.

What about Pixar rendering their next movie?

It probably costs them well over $1/CPU-hour to host their own cluster that gets fully used a tiny fraction of the time it potentially could be.

my OWN CPU cycles, thankyouverymuch

[Anonymous Coward]

I don't need to pay for some huge buisness' CPU's to get my computations done. I just hammer out the code myself, broadcast it to my 300 CPU server-farm in my basement, and voila! Free CPU cycles! (Except for the annoying monthly electricity bill)
--
Do you get those pesky Nigerian 419 emails? check out this site!: http://urgentmessage.org/ [urgentmessage.org]

Corporate Espionage issues?

[HowIsMyDriving?]

What if you lease time from Sun, and the computers that your data are being crunched on your compiditors computer? What about someone wanting to see what you are doing, and what is going on? This is not breaking encryption or finding aliens, this could be used by insurance, health, or governmental agencies for crunching large numbers, or searching databases. What have they done to address this?

Re:Corporate Espionage issues?

[DAldredge]

Google on Solaris Zones and Solaris Containers

Re:Corporate Espionage issues?

[gedhrel]

It is, indeed a big problem. Solaris zones, solaris containers, obviously have nothing to offer here. The issue with colaborative grid computing is that your computation (or some of it) may well be run on your colaborators' computers.

This is a major issue with many HTC (high-throughput computing, which is what this is, really) solutions. In fact, many sites look to create their own dedicated "pools" of compute nodes rather than push out computation to less sensitive areas.

So, generally, you are stuck with the "don't compute it somewhere you don't trust" issue. This is a problem in universities, too: we (in the UK) have data-proection act issues with many of the HTC jobs we run, since they may well be image processing for the medical faculty here.

An alternative approach, which requires work and doesn't always succeed, is to break down your computation into segments: compromising the security of a computation or result relies on compromising all the segments. Eg, for many image processing tasks (X-ray processing), it's possible to dish out tiny pieces of the picture to be processed. Only the originator of the computation actually knows which pieces go together, and how, to recreate the solution. Sometimes this is a viable solution; at other times, not. It often requires retooling the algorithms used, which is a highly expensive proposition requiring a great deal of skill.

Re:Corporate Espionage issues?

[Wesley Felter]

All the Sun Grid machines are owned by Sun, so that issue does not exist.

Re:Corporate Espionage issues?

[Anonymous Coward]

maybe i don't want Sun to see what I'm running?

Re:Corporate Espionage issues?

[Anonymous Coward]

maybe i don't want Sun to see what I'm running?

True, but then why would you want to hire Sun for the job?

Re:Corporate Espionage issues?

[Anonymous Coward]

Sun can't see it, Sun supply the computing resource which is an isolated environment (some parts logically and some parts physically).

The resource is provisioned and monitored by Sun but what goes in inside there is none of Sun's business (we don't need it to be and we don't want it to be).
We don't even have a login to the grid unless the customer requires consultancy/administration as the environment (OS and Apps) can be either entirely installed by the customer or supplied by the customer and installed by Sun.

Cost analysis

[Anonymous Coward]

Hmm, how will they make money?

If their grid has 1000 cpus running at an average of 10% utilization over time.

they will then make in 1 day (1000*24) = 24,000 * 10% = 2400 from their clients this is in 1 day, if they maintain that 10% over a year span
that will be 2400 * 365 = $ 876,000, with a possible $8.76 million each year. With a 3 year lifecycle this stands a chance to make money from people who need short term intensive processing.

Now who out there could use this, i know i could, but i wouldn't pay for it.

$1/CPU/hour is damn expensive...

[nweaver]

A good compute cluster can be had for $2500 a dual-CPU node. Assuming another $500/node/year for operating costs/upkeep, thats still

$1250 for a CPU-year. Compared with $8000/cpu/year for Sun's solution. So you better need BURSTS of CPU but not sustained CPU. And you better not be able to smooth out the burst demands with a batch-job system.

Re:$1/CPU/hour is damn expensive...

[Anonymous Coward]

You forgot the cost of:
1. Facilities: To run a large "cluster" you need the space to house the machines, the electricity to power them and the AC to cool them. $$$$

2. Maint. Contracts: You either buy one up front or you have to buy new compenents when those in your cluster fail (which they will). $$$$

3. Admins: Is the cluster going to run itself? Who installs software? Who maintains the queues? Who kills out of control jobs so that others can run? $$$$

For some organizations it is much cheaper to just build a cluster, but for many leasing time is an attractive option.

Re:$1/CPU/hour is damn expensive...

[TheRaven64]

I think your prices are a bit low for a good compute cluster - don't forget to factor in things like high-speed, low latency interconnect etc. Also, don't forget to account for technician time to keep it running (ours has roughly 40% of a technician's time allocated to it). Also, don't forget things like electricity and air conditioning (and clusters need a lot of aitcon). Even then, it works out around a factor of 5 cheaper to buy than to rent (offsetting the costs over three years).

Of course, that's assuming that you actually use your cluster. If you have one job that runs every month, and takes a day of cluster time, then it is cheaper to buy from Sun.

CPU-hour is a very wooly term. An hour on an ageing SPARC or Pentium is worth a lot less than an hour on a POWER5 with 36MB of cache.

Re:$1/CPU/hour is damn expensive...

[Glonoinha]

You forgot the most expensive (and often overlooked) part of infrastructure : the infrastructure staff.

Add a few $65,000 / year staffers in there to install / support those $2,500 machines and you are looking at $13,500 per year (every year) per machine. I know, that's what my company bills my department for each server I have on the network.

Re:$1/CPU/hour is damn expensive...

[Anonymous Coward]

$13,500 per year (every year) per machine

Wow your cost structure for IT sucks. Imagine if Google's costs ran as high - with 100000+ machines it'd cost them well over a billion dollars a year.

Re:$1/CPU/hour is damn expensive...

[Glonoinha]

Actually I thought about that - if infrastructure let everybody double the number of machines they had on the network without adding staffers, in theory the IT cost per box would be half and we would have twice the horsepower.

Wrong scale

[jd]

CPU charging has been done within Universities for mainframes and super-computers for a long time. Usually, though, it'll be in terms of clock cycles used and priority. The former can be metered exactly, as the OS will have that information whenever it swaps the process in or out. The latter can be fixed at the start.

(For real-time processes, you can even fix the clock cycles in advance.)

The advantages of doing things on this scale are that most heavy tasks will take in the order of seconds - at worst, minutes - and it is unreasonable to charge people for absolute time when an unknown fraction is spent in process swapping, paging and I/O blocking.

A flat per-cpu rate won't work well, as people will simply write programs to hog the CPUs as much as possible, thus taking the least absolute time, thus resulting in the least cost. It will also reduce the timeslaces available to competitors, driving UP their costs.

In other words, it is a recipe for creating a gang-land in electronic form, where the roughest and most brutal coders will "win".

There is also no incentive for Sun to improve such an OS, either, as they profit from latency. An increase in latency increases the fees, so reducing efficiency is a source of income.

It is, however, entirely in line with their Networked Computer idea, which I can only guess they are trying to resurrect. (If you can rent CPU space, you don't need to buy a local CPU.)

Re:Wrong scale

[TheRaven64]

Presumably they mean an hour of CPU-time rather than an hour of wall-time on a single CPU.

Re:Photo from inside the facility:

[Timesprout]

Guano Jim is about to become Guantanamo Jim because we all know only terrorists break NDA's

Flip Flop

[SilentReallySilentUs]

Sun's main problem is they are not able to stick to one strategy for rescuing themselves from the mess. Last year Java Desktop System was all over the press. 100$ per developer..I don't hear of it anymore - now it is 1$ per cpu. They are going to have a hard time getting the trust of enterprises to use a CPU in their servers. Good luck Sun, more importantly Good luck Java

Re:Flip Flop

[farble1670]

i assume you are not CEO anywhere, and that's good. successful long term companies don't make it by betting it all on one strategy. they dabble in a bunch of areas and see what pans out.

moreover, server vs. client based computing isn't black and white. one works better in some situations, one in others.

Re:Flip Flop

[SilentReallySilentUs]

You probably haven't heard of "Our goal is to organize the world's information" Compare that to - Network is the computer - We are the . in .com - $100 per developer - Java Desktop System - $1 per hour of CPU - You guys are screwing the revenue model - Scott McNealy to JBoss - Open source Solaris If a company doesn't show consistency in any commitment, nobody is going to trust it. And it shows in Sun's performance.

Re:Flip Flop

[farble1670]

the things you mentioned are not even peers. JDS is a product, where things like ". in .com" are marketing blurbs. to clear things up for you, the ". in .com" blurb meant that some high percentage of all network traffic (i forget the details) went through sun hardware. so? are you confused? does that somehow contradict sun's current direction?

as for $100 / developer vs. $1 / CPU, the former is a workstation billing model, the latter server based. they do need both since they offer solutions on both ends of the spectrum (e.g., JDS linux vs. sunray). so, are you confused by a company offering both workstation and server solutions?

yes, sun took the wrong approach to open source early on, but they did a 180, some time ago. are you still crying about it? are going to forever damn a company because they made a mistake (and corrected it)?

as for "our goal ..." ... google has no where near the breadth of products, and longevity that sun does. i would expect you to be able to reference fewer marketing blurbs from google. and btw, what's google mission again? to duplicate all of the services that yahoo already provides? maps? IM? web mail? a portal? truly revolutionary.

Re:Flip Flop

[SilentReallySilentUs]

Well, I have nothing against Sun but I am worried. It pains me to see a relentless innovator like Sun resorting to gimmicks like $1 per CPU to boost its sales. I am particularly worried about Java, which I like. While competiting frameworks like RAIL, PHP, .NET, are all catching up, the incredibly slow and bureacratic JCP is churning out more volumes of useless and incredibly gloated frameworks. I am not a CEO but simply a stock holder and user of Sun software. BTW, I have helped create http://www.collaze.com/ . Will appreciate it if you can take a look and give your feedback.

Re:Flip Flop

[farble1670]

It pains me to see a relentless innovator like Sun resorting to gimmicks like $1 per CPU to boost its sales

perhaps, but that's the reality of staying in business. during the .com boom, cost was not so much an issue. it's different now. sun's customers are very concerned about cost. the $1 / cpu "gimmicks" are a response to this.

sun has a rep for producing decent quality, overpriced solutions. they are trying to change that perception (the overpriced part). they understand that they will not get customers to buy their servers and linux based desktops unless they can show that the TCO is less.

No market there

[Anonymous Coward]

I'm going to wait until Microsoft copies that idea and rolls in a talking paperclip before I subscribe. Then I *know* I'm getting value for my hard earned dollars.

Re:No market there

[Cromac]

You joke, but MS Research has been working on building distributed computing into Windows for a while now. It probably won't be all that long until they either have a client, or something after Longhorn that will automatically distribute CPU load across a LAN.

http://www.betanews.com/article/Microsoft_Heats_Gr id_Iron_with_Bigtop/1104374194
http://www.microsoft-watch.com/article2/0,2180,174 6291,00.asp

It may be on the back burner at MS for now, but as we've seen many times if they perceive a market they're missing out on they can throw enormous resources at a project to get it to market.

Re:No market there

[Anonymous Coward]

That's why I hate Microsoft. The maintaining of the IT economy, any economy, really, is, above all else, an intellectual and creative endevour. People speaking with their minds as it were. Microsoft, like any entity that speaks only with its wallet, is little better than entities that speak only with their fists.

Anonymous Coward (the same one, actually.)

Re:No market there

[Anonymous Coward]

I guess that would make findfast.exe a nice source of revenue...

my cpu cost me over $290

[Anonymous Coward]

Energy costs not included.

So to profit, I gotta get paid $290 a year on top of energy costs. But since my cpu is idling most of the time I may as well sell it for below cost (so I can reduce my losses).

Any takers?

-Johan

Missing the point??

[SatanMat]

-- It seems that most people here are missing the point


"Just buy a cheep computer to do some of the work for less $$$"

I would think that this would be aimed for people who have a specific problem that would overwelm your current setup and is more of a one off, or infrequent large problem.

Think of the times when you would need a supercomputer for just a few hours...
like an old mainframe setup without most of the headaches.

In the Future®

[ch-chuck]

Why did my database stop working? D'oh! I forgot to pay my computer utility bill!! (smacks forhead).

What CPUs?

[Saiyine]

Should I rebuild the bunch of pentiums and 386sx I store in my garage?

--
Dreamhost superb hosting.
Kunowalls!!! Random sexy wallpapers.

Wrong model

[Wesley Felter]

Most batch nodes are space-shared, not time-shared. That means only one job is running on a particular node at a particular time. This does not create any perverse incentives; developers are still incented to make their code as efficient as possible. If the job is blocking on I/O, that's the user's fault, not Sun's.

bang for the buck

[cybergrunt69]

OK, so for one year of CPU time, maybe it is initially cheaper to buy a whitebox and install linux than it is to use this Sun solution.

However, if you got a linux whitebox to run this, not only would you have to worry about power costs, but also every other detail that comes with making sure your machine is running. What about patches, upgrades, network, bad hardware, runaway processes, general administration, backups, storage, etc? Most of the people here would be able to do the standard stuff that's needed, but I'm sure a business that needs "xyz" computed would gladly pay the 2x price. Not only would it do away with all the minor details, but they'd also have their results back in a significantly shorter amount of time! I'm too lazy to do math right now, but I'd say a year of cpu time could easily be done in less than month. That alone could be _the_ deciding factor and the justification for the expense.

Re:bang for the buck

[jonbryce]

I think power costs are so small you can safely ignore them. I don't know how much electricity costs in the US, but based on UK prices, you would be looking at about $0.003 per hour to power a computer.

Re:bang for the buck

[jrockway]

Wrong. In Chicago, electricity is $0.08275 per kilowatt hour. Let's just say 8 cents to make the math easier. My Linux box uses about 200W of power (it's a 420W PSU, but I'll assume I'm not maxing it out). This is 0.2KWh / h or 144KWh per month. Multiplied by that rate, it's $11.52 an hour. Over a year, that's $138.24. For a college student like me, that's not pocket change!

Needless to say, I've decommissioned that machine and only turn it on when I need to use it. My fileserver / e-mail / web server is now running on an NSLU2 -- which uses 8 watts. That costs only about $5 a year to run.

Re:bang for the buck

[jrockway]

$11.52 per month, I mean.

Note to self: Engage brain before posting, please :)

Re:bang for the buck

[Anonymous Coward]

Just so happens I was running some financial information this morning. To power/cool a Sun V240 costs about $.048/hour if you use the US national average of $.10/KWH. To power/cool an older Sun E450 (which my company has over 650 of still in production world wide) is $.204/hr.

That's just power/cooling. That does include maintenance (which two years ago, the last ready data I have) was $.395/hr for platinum 24x7x2 response support for the E450. Space was $.140/hr. Helpdesk and L1 24x7 support was $1.575/hr. L3 support was $.452/hr. L2 support is the responsibility of the business unit themselves. I haven't included networking costs, storage costs, or any of a dozen or so other costs. And remember, these are wall hours, whether the systems are used or not.

'display over IP' infrastructure

[GillBates0]

1. gridhost% setenv DISPLAY crappy_host.mydomain

2. gridhost % while 1
   ?  /usr/openwin/bin/xinit /usr/dt/bin/Xsession &
   ?  ???
   ?  echo PROFIT!!!
   ?  end

Free for employees?

[jericho4.0]

Some pretty interesting things might come of just letting it be used by employees for now. I'm sure a few of them have had ideas that would need oodles of power to flesh out, that nobody would pay for the big iron to run. I, of course, would recompile Quake 3 and get myself some geek cred with a stupidly high FPS.

How long for a portage plug-in?

[LaminatorX]

emerge update -distcc -sungrid

for example

[taniwha]

think of chip design - you want about a gazillion machines for QA simulation in the 2 months prior to tapeout and they're going to be idle for the rest if the year

Mind you the cost of chip design software is the limiting factor here, not the cost of hardware to run it on

Re:for example

[Anonymous Coward]

According to the OP's figures, it'd still be cheaper to buy the gazillion machines, and throw them away afterward, if you're not going to use them again for the rest of the year.

big margins

[rnd()]

Sun has some fairly substantial profit margins built in at the $1/CPU/hour price.

Consider that if you have a moderately large data set that you need to crunch it's not at all uncommon for it to take 3 hours on a 300 node cluster. That's $1800 if each machine is a dual proc machine.

So suppose Sun has a 300 node cluster, for example, where each machine cost $1800. Ever 3 hours one of the machines is paid for. In other words, a 300 node cluster is paid for in 38 days. Well, the hardware is, anyway.

I really don't know who the main clients would be of this kind of service, however. I'm guessing that if your company can't afford a 10-20 node cluster (fairly cheap) and still needs to do large scale computing, renting CPU cycles from Sun would make sense, though it would very quickly cost more than the 10-20 node cluster would have. So it's really going to benefit customers who need large scale number crunching results more quickly than they can obtain them simply by building a smaller cluster and waiting for the results, or customers whose problems involve data sets that are large enough that they need to be distributed over 100+ machines in order to be solved.

Who has large data sets like that and no cluster access? Not university researchers, not government agencies, and probably not most firms doing significant number crunching.

So I see the niche as firms with large data sets and someone who can write the MPI code, but who lack the willingness or finances to invest in a cluster of their own.

In a year or two when the same service is selling for $0.25 per cpu hour it will be a much more compelling proposition.

Re:big margins

[Com2Kid]

In response, if you read one of the articles linked to from the main article:

Sun Chief Operating Officer Jonathan Schwartz submitted a project to the Sun Grid--graphically rendering data from a protein folding experiment. It took only a few seconds, but cost $12. The 12 hours of CPU time for which Schwartz was billed was consumed by hundreds of machines simultaneously clicking away at the rendering problem for a few seconds each.

Results n seconds.

Doing a quick read through I did not see exactly how large the cluster was, but there is no reason that Sun could not scale it to a few thousand nodes, providing results in a time frame that would be cost prohibitive for most companies to setup clusters to compute.

Come on, if you are running a huge simulation a few tmes a year, maintaining a 1000+ node cluster year round is just not cost efficient.

Re:big margins

[rnd()]

Sun's pricing is per CPU hour so the more machines you use the more it costs. There would be no price improvement to be had by adding more machines to the cluster. If anything, the value would be slightly worse due to the additional communications overhead introduced by MPI having to talk to the additional machines.

You are correct that a company who needs a 1000+ node cluster a few times a year is a perfect client for the service. I don't know how many such companies there are.

Re:big margins

[Wesley Felter]

Sun's pricing is per CPU hour so the more machines you use the more it costs. There would be no price improvement to be had by adding more machines to the cluster.

But there is a value improvement to the customer of getting the job done faster.

Re:big margins

[Anonymous Coward]

The $1/ CPU hour is the base rate, according to a previous article ( can't find the article right now ); Customers contracting for extended / consistent service will see discounts.
    #1

Re:big margins

[alelade]

I totally agree, the 1$ per cpu hour is an outrage. When you think about forming large computer clusters it may make sense but that is gradually changing. Check this out : http://www.clearspeed.com/ These guys sell a 64 bit 50 gflop co-processor for 10000$ added to the system through pci-e port, and this is a comperatively new practice of good old co-processor technology. As it develops more, we can expect better prices, higher power. For the same processing power, an array made with these babies currently will need to be about half the hardware price. Also it will mean about 1/10 cluster size, further reducing maintenance costs significantly.

Just wait til the spammers and marketers see this

[msmercenary]

Why bother trying to win over customers, when a $150,000 marketing budget can be so much more effective leasing 1000 CPUs to DDoS the competition for a week... ;)

Economics of Sun Grid Computing

[WillAffleckUW]

1. cost for a CPU
2. cost for the box for the CPU
3. cost for data storage
4. cost for monitors
5. cost for cooling for above 1-4
6. cost for power for above 1-5

Now, ask yourself, will the price of power go down if oil will cost $100 (current median bet in the Oil Futures stock simulation using real dollars as per WSJ)?

yes, there are ways to reduce those costs. but not everyone can.

Re:Economics of Sun Grid Computing

[Alsee]

I'd just like to clarify a few of your items.

4. cost for monitors

I guess monitorS plural is technically correct. You probably want a minimum of two interface terminals. :)

However you certainly do not need or want hundreds or thousands of monitors for the hundreds or thousands of CPUs.

2. cost for the box for the CPU

Sort of. Just taking a wild guess here lets say motherboards with 4 dual processors. So figure a one eigth of the cost of a motherboard per "CPU". Figure 16 motherboards in a rack? Ok a rack is going to be rather more expensive than your typical box, but we're only looking at 0.008 of the cost of a rack per CPU.

3. cost for data storage

Sort of. I'd expect they do a network boot and store data on a central RAID. Not cheap, but you you are not going to need a separate big drive for each processor. Considering that it is shared storage and shared data the per CPU drive space can be fairly low, maybe a couple of Gig per processor. Hypothesising a massive SUN center with 25 racks, they could quite easily handle even a storage devouring tasks like rendering feature length movies at full theater resolution at a mere 1 Gig per CPU, with ample room to spare.

Storage is almost a fixed factor depending on the type of tasks, it doesn't need to go up much at all per-CPU. And the good thing is that you can just slap more storage into the RAID to scale to whatever size needed.

5. cost for cooling for above 1-4
6. cost for power for above 1-5
if oil will cost $100

Well lets do some back of the envelope calculations.

Lets assume you're brain damaged and you set up this grid on Long Island with obscene 17 cents per kilowatt-hour electricity costs. Lets assume that 17 cents is based on an assumption of $50 per barrel oil. Lets assume that oil goes up to $300 per barrel and that the price of electricity is solely and directly related to the cost of oil... lets assume it multiplies by 6 to an obscene $1.02 per killowatt-hour for electricity. Lets assume each dual core CPU spits out and obscene 200 watts of heat, or 100 watts per core. Lets assume you blow an obscene 100 watts per CPU in support costs between the harddrives and the fans and office lights and you install a big fat 10,000 watt neon sign over the building, chuckle.

That's 200 watts per CPU at $1.02 per kilowatt-hour, or 20 cents per CPU-hour.

So even with some of the most expensive electricity in the world and even at $300 dollars per barrel for oil and even with inflated electricity usage and whatever other silly assumptions, the 20 cents per hour per CPU has *just begun* to make a dent in the $1 per CPU hour they are charging. Under more reasonable assumptions... particularly if you locate somewhere with cheap hydro power... you can run the system for a few years before the electicity costs add up to anywhere near the initial hardware costs.

A large grid with hundreds or thousands of processors is certainly an expensive endeavor, but most of the items you listed aren't very signifigant. The dominant factor in that list is the cost of the CPUs and motherboards. If you are capable of paying six or seven figure sums for the processors and boards, well the rest is just an "adjustment factor" to the price.

-

Computing@home

[Saiyine]

What about releasing a grid client so every one could earn some bucks by letting their cpus work for others?

You know, like the spam bots in windows, but getting money!

--
Dreamhost superb hosting.
Kunowalls!!! Random sexy wallpapers.

Re:Computing@home

[Anonymous Coward]

Hello /.errs,

I just want to second the "interesting" score, since FYI I'm working on something like that in my spare time for about one year.

But besides allowing every one to earn some bucks as you suggested, it offers several advantages to the buyers and it will be able to recycle much more energy than what Sun approach can do.

http://www.cpushare.com/

See my post in comment to Sun approach with some comparison:

http://www.cpushare.com/hypermail/cpushare-discuss /05/08/0031.html

It requires some security kernel support (seccomp or trusted computing) so the client will only work on linux in the short term (seccomp is in kernel >= 2.6.12).

Re:Computing@home

[timeOday]

One problem is most of us don't have much bandwidth at home, so this grid would only be useful for extremely compute-intensive jobs which do a lot of processing on a tiny amount of data without much communication among nodes.

Another problem is you'd have to do every computation on at least 2 different machines to have any confidence in the result.

Re:Computing@home

[Anonymous Coward]

Long term I doubt bandwidth will be a big issue. More interesting will be latency but there's a lot that I can do with CPUShare to improve latency because the computing farm is distributed across the globe and not centralized in labs like for Sun.

The RTT is already measured and you will be allowed to screen for low RTT (according to supply/demand law, price should increase if you ask for lower RTT or higher MIPS/MFLOPS, but this is up to the market to decide, CPUShare can't control it).

About computing in different nodes, the form to create a buy order already includes the first 8 bits to screen A classes of the IPv4 address space (like it already includes the MIPS/MFLOPS/RTT screeing). So you can compute on different A-class. Over time trusted computing will make it a lot more reliable than it can be today, but you will never be allowed to rule out bitflips in the RAM/CPU (so computing twice sounds better idea even if you use Sun farm, just in case something goes wrong in the hardware).

Re:Computing@home

[flutkatastrophe]

You should really mark your kunowalls link as not safe for work (nsfw).

Cool but what about licensing?

[adawg]

Sun wants us to run an app on their powerful grid, but what do software vendors think about us running their single lincense across, say, 100 CPUs?

So now, on top of the $1/CPU/hr, we have to buy a license for each of those CPUs.

Or else this will be very good for open source.

Re:Cool but what about licensing?

[SagSaw]

So now, on top of the $1/CPU/hr, we have to buy a license for each of those CPUs.

It might make sense for Sun to work out agreements with software vendors to license their software to these machines on a CPU time basis.

$1/CPU/hr for the CPU time, $X/CPU/hr for the use of (insert high-end software package here).

Re:Cool but what about licensing?

[bradleycarpenter]

Sun partners with most of the big whigs in computing. I'm sure they get stuff for free or a very good price on licensing.

Re:Cool but what about licensing?

[adawg]

Sun could have the best friends in the whole world and they wouldn't get software licenses for free. Maya, Renderman, and Shake licenses (very often used on grids) cost more than the machine they are run on. If a potential Sun Grid user has to negotiate and buy licenses for each of the CPUs they are renting from Sun, this will be a big flop. And that's one of the biggest issues for grid deployment: the predominent licensing structure - aka software company revenues - isn't prepared for distributed computing. So until someone breaks ranks and offers a sane model for licensing, only the most influential customers will be able to wring temporary licensing deals. It is certainly in Sun's interest to facilitate this effort.

Re:Cool but what about licensing?

[Anonymous Coward]

Depends, the software is licenced for a single computer (or site or whichever terms)

I've never seen where you have to have a second licence for a dual processor computer. Most likely this is exactly the same, doesn't matter how many CPUs the task is spread over.

How about 1$ per java download?

[SilentReallySilentUs]

Sun can use paypal for worlwide collection or money.

Rather like virtual servers

[commanderfoxtrot]

The concept here is very similar to virtual servers. Instead of charging per disk and RAM, they're formalising the charging process for CPU.

You could apportion a similar costing structure to a normal e.g. unixshell/linode virtual server to charge by CPU hours.

Aggregation of Demand

[IEEEmember]

The whole idea of resource sharing via networks is based on the fundamental concept of aggregation of demand. The expectation is, that given an operational network, relatively small demands from around a large geographic area can be aggregated and thus be satisfied with one or a few specialized centers. By aggregating specialized service production into fewer centers, the required services can be provided at lower costs than would result from provision of the equivalent services by means of a large number of smaller machines. This principle of demand aggregation, to obtain the advantages of scale, applies to hardware, to software, and to operational costs.

From IEEE Computer Society magazine Computer, August 1973
reprinted in August 2005 in the 32 & 16 Years Ago column.

Re:Aggregation of Demand

[Alsee]

Chuckle, that reminds me of this quote:

"I think there may be a world demand for about 5 computers" - Thomas Watson, founder of IBM

-

Yes, yes....

[Anonymous Coward]

The execs will get to that, right after they finish squandering the money they got from taking it in the ass off Microsoft.

College

[Spazmania]

Yeah, they metered us this way back at college. It was supposed to simulate a "real" computing environment. We got a quota each semester. An hour logged in to a Sun was the same "price" as 1000 pages off the laser printer. I printed a lot of manuals. A LOT of manuals.

Re:College

[Fjornir]

Damn. They billed you for login time?? My quota was on normalized CPU hours, and at a certain nice level spare processor time was free.

Re:College

[Spazmania]

Depended on the machine. The Sequent was billed on CPU time but the Sparcs were billed on connect time. We had 300 "allocation units" for a 12-week quarter and connect time on a Sparc cost 5 AU's per hour. Needless to say, no CS student stayed within their quota; everyone sought and received an increase later in the quarter. We also got dinked for megabyte-days of disk storage. Everyone started with a 2 meg quota, but most CS students got theirs bumped up to 20 megs.

Back in the days when we walked 6 miles in the snow, uphill, both ways.

Re:College

[afidel]

Man your post reminded me of why I got into Linux. I went to school in Rochester, NY (RIT) and walking the mile or so from the dorms to the CS lab across campus was difficult when there had been a snowstorm with a couple feet dropped in 12 hours (a regular occurance in Rochester) so I got into Linux in order to be able to export an X display back to me in the dorms without spending loads of cash (there were no free Windows X servers in those days AFAIK). There's nothing quite like laziness to motivate learning =)

Re:College

[Anonymous Coward]

What are these quotas of which you speak?

Who is going to use it?-Victims of "/."

[Anonymous Coward]

"Seems like the main categories of potential users fall into two camps:"

Three. All those people expecting a Slashdotting. Now they can laugh in Taco's face.

Old is new again

[Nom du Keyboard]

Sun is offering processing and storage in a pay-as-you-go arrangement of $1 per CPU per hour

Sounds exactly like the timesharing model I used on a Burroughs B5500 system circa 1970.

To Paraphrase "The Right Stuff"

[wsanders]

"I've tried A! I've tried B! I've tried C! All the while the aircraft is hurtling flaming ball plummeting out of the sky like a - I've tried D! I've tried E!"

Cost of breaking SHA1?

[Anonymous Coward]

Given last week's story about the new break for SHA1, one has to wonder how much it would cost to find a collision! Never fear! Back of the envelope guy is here! :)

Assume the CPU can do 1-4 billion ops/sec. 2^63/2^30 = k * 2^33 seconds, where k is the number of ops/step and includes a 1/4 ... 1 fudge factor mentioned above. At $1/3600 seconds, we're looking at k * 2^33 / 3600, or about k * $2.4 million.

I don't know how many many ops are required for each step, but I'll make a "safe" assumption and pick 10-20. So I'm going to make a ballpark guess centered at about $15 million, with a 10x fudge factor in both directions (i.e. plotted on log scale with confidence interval bars going from $1.5 to $150 million).

As soon as someone knocks SHA1 down to 2^53, we'll be looking at a guess of $15,000. And if they knock it all the way down to 2^43, I'm sure a lot of people would be willing to shell out $15 to make a pair of documents with the same hash.

p.s. It's fitting that my CAPTCHA (also in a story today) is babbles.

How much will it cost to calculate pi to 10^100

[Anonymous Coward]

place ;)...

Obligatory reference...

[Mars Ultor]

Imagine if you hooked up enough of these into a pay-as-you-go Beowulf cluster!

Lot of unanswered questions.

[mjuarez]

What CPU will it be? A 4Mhz 8080? Or a Pentium II? Or a 1.0Ghz UltraSparc V? Or is it only one kind of CPU? If so, which kind, what speed? And will the $1/hour/CPU increase as they upgrade to new processors, or will it stay constant?

Wouldn't it be better to offer, say, $1 for each million MIPS? Would be a lot more straightforward.

Re:Lot of unanswered questions.

[Wesley Felter]

IIRC it's an Opteron.

Given that processors get cheaper over time, you'd hope that the price never increases.

Pricing according to a Meaningless Indicator of Processor Speed doesn't sound like a great idea to me. Also, if the grid is space-shared, your job is consuming the same resources regardless of its IPC.

Re:Lot of unanswered questions.

[bradleycarpenter]

They are going to mostly be Suns new opterons. They are simply using their inventory on the grids. Good way to burn-in those systems. :)

Re:Lot of unanswered questions.

[timeOday]

Instead of that, do it market-based: start off for some negligible fee, then sell the hours to whomever pays the most.

Where do I sign up...

[sjs132]

to sell my spare CPU Cycles? At a $1/Hour, I'll let them have the 8 Hours that I'm at work using my office computer! That will give me some extra munch money... :)

Re:Where do I sign up...

[Anonymous Coward]

Dunno about Sun, but you can sign up at http://www.cpushare.com/ to start earning CPUCoins right now (real money via paypal is supposed to follow in a matter of months, as my spare time permits)

Don't forget the cost of configuring the "Grid"

[Anonymous Coward]

We're rolling out Grid computing at the Huge Neighborhood Financial Site where I work and Grid is basically ssh / rsync on steroids. If you can imagine a beowulf cluster of .... well, you get the idea. It takes quite a bit of overhead and training to install and configure, and so far the functionality I've seen is no more wonderful than "rsync my-script && ssh ./my-script". You are going to drink lots of Solaris/Java koolaid before it's working up to its potential. I'm a tad skeptical but for Huge Neighborhood Financial Site, who refuses to allow Open Source into their datacenters, it's probably all for the best.

Can't really see this being successful.

[GeorgeMcBay]

To really take advantage of a platform like this, you need to write code tailored for it. Who is going to risk spending large sums of money and lots of time developing for such a platform when it is controlled by a single-vendor and could disappear at any time?

Re:Can't really see this being successful.

[Anonymous Coward]

Uh, there is already tons and tons of code written to take advantage of grid style computing. All Sun is doing is offering to host the Hardware for you as well. You get to use, say 80hrs, of time for that once per quarter you need it and never have to buy the infrastructure that would otherwise sit around the rest of the time.

Re:Can't really see this being successful.

[Anonymous Coward]

Dunno about Sun but with CPUShare (http://www.cpushare.com/) I have the problem you outlined.

People will have to port their application to the seccomp/trusted-computing programming model. That takes time for already working applications.

So with CPUShare I intend to release a GPL'd version of the server to use in the intranet. However this will happen only after the "internet" version is finished and working.

I hope this will make people more confortable about porting their app to the seccomp or trusted computing programming model without risking to lose the whole investment.

The seccomp feature is extremely good for legacy grid computing too, so it's not like there will not be an advantage. Trusting what comes from the server is an extremely bad idea, even on the intranet, especially for large organizations.

With the seccomp programming model I guarantee at the kernel level that the client cannot be hurted permanently by the grid computing (no matter if the software comes from the internet or the intranet).

Re:Can't really see this being successful.

[Anonymous Coward]

Exactly, Sun supply hardware and an infrastructure that can support MPI, ROCKS, Java/Jini, N1 Grid Engine, whatever the customer whats.

This is purely a platform on which customers can run their apps, it is NOT a proprietary top-to-bottom stack which customers need to adapt to.

Re:fr1st p0st

[Anonymous Coward]

So Sun did it first? I would have always expected that Apple employees would be the first to receive GRIDS.

Applications that make sense

[acaspis]

As expected most comments about "who's gonna pay for this" and "it's cheaper to run your own server".

But think about business models where the grid provider sells not only CPU cycles, but also trust.

Scalable web hosting: Your PHP code is replicated on-demand to as many grid server as needed to handle your peak loads. The grid provider guarantees that server-side code and data remains confidential. End of the slashdot effect as we know it.

MMORPG: Small startups can deploy worldwide networks of game servers in no time and compete with the big boys.

The next Google: Anyone with a smarter search algorithm can go online without investing in huge datacenter first.

Rendering farm: Your CG movie is due to premiere next month, and your 10,000-node rendering farm can complete the job in time. Wouldn't you pay extra $$ to anyone who can save the day and guarantee that screenshots won't be leaked to the Net ?

Combine this with a micropayment infrastructure where the grid provider sends bills to end-users on behalf of the service provider. Huge potential.

AC

RIGHT scale

[csirac]

As other replies have aluded to, there's a difference between CPU time and wall time.


$ man 3 clock
          clock - determine processor time used

LIBRARY
          Standard C Library (libc, -lc)

SYNOPSIS
          #include

          clock_t
          clock(void);

DESCRIPTION
          The clock() function determines the amount of processor time used since the invocation of the calling process, measured in CLOCKS_PER_SECs of a second.


So as you can see, even if they used this standard POSIX function to measure time it'd still be counting clock cycles, just measured in hours.

top(1) and ps(1) can also tell you the amount of CPU time (measured by cycles spent on the process but reported in units of TIME) a certain process has taken up since starting.

For instance, I've been running Firefox for about 3 hours now, but it's only used about 11 minutes (10:59.39 minutes exactly) of CPU time (in the 2nd-last column of the ps list):


USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
csirac 7377 2.3 19.3 296996 151956 ?? S 5:51AM 10:59.39 /Applications/Net/Firefox.app/Contents/MacOS/firef ox-bin

yay for slowlaris

[Yonder Way]

Not all CPU hours are the same. An hour on a moderately fast SPARC processor is not as valuable as a moderately fast Intel Xeon or AMD Opteron or a PowerPC.

Re:yay for slowlaris

[Anonymous Coward]

For starters the Sun grid is running Sun's Opteron Servers (v20z).

Besides you will be swallowing your words about SPARC when they release their 8-core Niagara servers early next year.

Silly name calling aside Solaris is easily as fast as Linux.

Flexibility

[tez_h]

Others have noted that the cost of a bespoke system is lower. These slighly ignore the fact that this is aggregated over some time period, like a few months or a year, where the for most part of that time the computer may be sitting idle if there aren't jobs to fill up the time. But Sun's solution offers granular computing time for any number of CPUs at a linear rate, meaning you don't pay for the time you don't use, and you can pay for lots of CPUs without penalty. I would imagine this convenience and the additional conveniences of infrastructure, maintanence, security, updates, etc are what you're paying $1/CPU/hour for.

Since Sun is attepting to utilise economies of scale, as this gets more popular, presumably prices would become more and more accessible.

-Tez

The Power of The SUN...

[JrbM689]

...In the Palm of your hand. Too bad Palms don't multitask. (Paraphrased from Dr. Octavius of Spiderman 2)

I can shoot down one of these

[RelliK]

Rendering farm: Your CG movie is due to premiere next month, and your 10,000-node rendering farm can complete the job in time. Wouldn't you pay extra $$ to anyone who can save the day and guarantee that screenshots won't be leaked to the Net ?

And how the fuck are you going to transfer *hundreds of gigabytes* of data required to render a frame over the internet? How are you going to receive the data back? (2MB - 12MB per layer per frame).

Does that thing even have Renderman installed? (at $5k/CPU I highly doubt it). Does it have Shake? Does it have Houdini? Does it have Maya?

Besides that, how the fuck are you going to get approval to send _anything_ out of the studio? You obviously have never worked in the industry.

I'm also skeptical as to whether there is any use for this. What sort of environment do they run it on? Solaris/SPARC? Solaris/x86? Linux? Windows? What sort of software does it have installed? Would it ever be possible to replicate the in-house environment on this "grid"? (you know, with all the custom software, directory structure, environment variables, aliases, etc.) I know for a fact that there is no way we could outsource our rendering to Sun even if we tried.

The whole "CPU-hour" thing is a very nebulous concept. Environments differ wildly from one company to another, so you can never have a universal "CPU grid" in the same sense as you can have an electric grid.

Re:I can shoot down one of these

[acaspis]

How are you going to receive the data back?

We'll all have fiber-to-the-home (or at least to the office) by the time grid computing comes of age.

Renderman installed? (at $5k/CPU I highly doubt it).

I'm sure software vendors will be more than happy to adjust their licensing models to any standards-based national grid infrastructure. Especially if the infrastructure does DRM for them.

get approval to send _anything_ out of the studio? You obviously have never worked in the industry.

Indeed. Possible approaches:

• In theory, Trusted Computing (TCPA/TCG) could allow customers to ensure that their tasks are running on secure grid servers. "Secure" means secure against software attacks. You could use some kind of anonymizing front-end to make it hard for anyone to find out which servers they should target. Does your threat model include movie pirates plugging logic analyzers into millions of servers hosted in nuclear-plant-like data centers ?
• Have the servers process your data in encrypted/obscured/anonymized form. Might not work for rendering, but there are plenty of applications where it would help.

you can never have a universal "CPU grid"

I bet Sun's answer begins with the letter J.

If that's not fast enough, think virtualized operating systems. And for any compute-intensive project, the overhead of cross-compiling for a few architectures would be negligible. I once had to throw away several cpu-days of Povray output though, because the pseudo-random textures didn't come out the same across all platforms.

Re:I can shoot down one of these

[Spectra72]

"...Besides that, how the fuck are you going to get approval to send _anything_ out of the studio? You obviously have never worked in the industry..."

That's a very, very shortsighted POV. Right now, your Tax Returns have a good chance of being done in India. Companies are hiring temporary consultants *daily* to do highly confidential coding and other development. These things are handled quite easily with appropriate controls and disclosure agreements (are there problems and do things fall through the cracks? Sure.) We are moving into a Corporate landscape were everything that *can* be outsourced, will be outsourced. Maybe studios have been reluctant to send out their rendering up until now because there have been no viable alternatives. Not enough bandwidth for the transfers, not enough security controls (DRM anyone?). But maybe tomorrow there will be enough bandwidth and there will be proper controls. At that point..BAM! That work is off to someone else.

Your "We're too complex!" points are the same cries heard when HR was being outsourced, or general IT work was being outsourced. As soon as your company realizes it can save considerable dollars using an outsourced grid farm, it will. Maybe not from Sun, maybe not today...but soon. If someone showed your CEO that he could save 50% on development costs, with guaranteed security and SLAs and all he had to do was mandate a few changes to your software configurations (to make it less custom and more easily ported), why wouldn't he do it?

Re:I can shoot down one of these

[Mac Degger]

Uh...why not use the distributed rendering client that Maya, and Houdini have standard? Shit, Renderman loves to work that way...I live in a flat and have lots of budies help me out when I'm rendering something....all over TCP/IP (over a LAN, true, but ANY IP could take/use/send the data).

"Does that thing even have Renderman installed? (at $5k/CPU I highly doubt it). Does it have Shake? Does it have Houdini? Does it have Maya?"

It doesn't need to. It just needs the distributed rendering client. And your studio would need a phat pipe...but nowadays that's a trivial (if one-time expensive) thing to get.

Save money: procrastinate

[dazedNconfuzed]

There's a wonderful technical paper analyzing the value of procrastination. Don't recall the details, but the upshot was that for certain computing problems long enough (on the order of years compuatation time), at some point it's just better to do nothing at all until the price of the hardware drops and the CPU speeds rise enough so that you can actually get the job done sooner than if you started running the job on what you had ASAP.

If you have computing job that will take 4 years on today's processors, and processors will be 4 times faster 2 years from now, better to do nothing for 2 years and then run the job on computers that will complete the job in 1 year, getting you done a year early including a 2 year vacation.

Wish I still had that paper...

Re:Save money: procrastinate

[Mac Degger]

Heh...I know this one from science fiction stories from the 50's-70's...only there it was space travel (launch something now which takes 100 years to get there or wait 50 years for a technology which will get you there in 20 years).

Re:Save money: procrastinate

[Anonymous Coward]

So........

After 2 years. Computers will be twice as quick in 1 years time, getting the job done in 1/2 year leaving another years vacation.

After another year. Computers will be twice as quick in 1/2 years time, getting the job done in 1/4 year leaving another 6 months vacation and three months spare.

Sum to infinity and you get 4 years vacation with zero computation time - and your answer.

I think the maths are flawed... ;)

Way To Go!!!

[Anonymous Coward]

Congrats my Sun(son). But let the cost come down to 1cent/CPU/Hour.

Hey Look! Star Trek's on!

[ebooher]

Evening all, it's once again time for an EU Digital Rights Patent Pending Bullshit Theory of the Day! I ask for patience as I weave my tale tonight, for the future is never more than a dim bulb as if about to burn out and the visions gleaned will be open to wide reinterpretation.

I want to go on record as saying that I approve of the Sun Grid, and the ideal of computing utilities. While I love the fact that my Mac thinks by itself, I've been watching the wall and the graffiti has begun to get very interesting. The Sun Grid concept as it stands today is best for major academic and scientific researchers. People splitting atoms virtually who are making $300,000 US in their work efforts. Is that me? No. Is it you? Probably not. The current intended market are eggheads of a different sort. A minute of their time wasted waiting for a computer is $.50 US. Two minutes of wasted time pays for a CPU Hour from Sun Grid.

Why is this important? It isn't, not yet. The fact that Sun had a press release, issued a statement, and as geeks we saw it zoom by and have been waiting for months to see what they were talking about has nothing to do with geeks that frequent /. Doesn't even have anything to do with most geeks that don't frequent /. Frankly, there doesn't seem to be anyone who is trying to sit down and think about it like a pointy haired boss and not a geek.

But I digress. Star Trek, yes?

The Kirk vs. Picard debate aside, The Next Generation had great technology. Kind of had to if it was going to be used to save the ship every week. The main computer of the NCC-1701-D was the most massive mainframe anyone could ever think of and was pretty much the entire ship. Sure, the Galaxy class had a computer Core, but really when you think about it all that meant was massive storage arrays. Believe what you will, Star Trek has pushed (especially the American) the dreams of tech builders ever since it hit the air waves. Just look at every flip phone and tell me that isn't a Kirk age communicator. For that matter, look at any PDA and tell me that isn't a PADD. (Personal Access Data Device or something)

The big thing is tomorrow, the 4th dimensional tomorrow. There will be a computing "Grid." Whether from Sun, or Microsoft, or Apple, or Bob's House of Discount Computers. The ideal and need for one is there. Seti @ Home and Bit Torrent prove that much. Why do I say this?

Wireless Communication Technology.

When the away team beamed to a planet's surface, they were wearing sophisticated VoIP communicators on their chest. Tap the dialer, and speak the name of the person you needed to speak to, the main computer system found that buddy online, routed the call to their VoIP communicator, chirped them and allowed two way communication. Don't tell me these aren't on the way, Nextel has been using a feature that is one step removed for years now.

Pull out your PADD and let's get some work done. Integrated WiFi ensures a connection, but why does tomorrows PDA need to think all by itself? Heck, I'm not even a crazy, knee deep, senior geek anymore. Too much time spent doting on my 2 year old daughter. Yet, I still have no less than 7 UNIX systems around the house that are all accessed via a single system, terminal, ssh, X Windows server/client protocol, and for the times I need to help Windows friends VNC. Why do we think that with the move to more mobile systems, global wireless technologies and scaled down embedded processors that we all won't still do all that tomorrow? The PDA / laptop / digital paper / Bob's your uncle that will be along very soon now will be very similar to Sun's Sun Ray systems. Insert your Java Card, enter your password, and you are on your desktop. Today the Sun Ray is a thin client desktop. The protocol, in theory, will allow you to connect to your home office from any Sun Ray with your card. Provided the firewalls allow it, routing systems allow it, etc. But let's view the future as ide

Imagine....

[paulius_g]

Imagine a Beowulf cluster of these!

Oh wait...

Missing the point

[tsotha]

I don't see why so many people think this is aimed at people doing massively parallel processing. It's not.

I'm thinking of my own employer, who's got hardware up the wazoo that mostly just sits around heating the building. Take payroll, for instance. They probably run some payroll batch job a couple times a month, then the rest of the time the computer that does payroll is just sitting around. Sure, you could run other stuff on it, but then when it came time to do the payroll everything would run slow and people would be upset.

The way I see it, this is perfect for all kinds of periodic batch-type business applications where you really want to have a dedicated machine but it won't be utilized all the time.

Also, machines that mostly sit around have about the same maintennence requirements as heavily-used hardware. They still need OS patches, security patches, backups, etc. And they need to get modified when the new head of IT decides logs should go in /var/tmp/messages instead of /var/adm/messages or whatever. I know my employer could probably get rid of 2/3 of its data center and a corresponding fraction of the high-priced administrators currently making everything run smoothly.

You could argue companies can do the same thing by balancing existing hardware, i.e. have one box host multiple "bursty" applications so the CPU doesn't have much idle time. But that takes lots of effort to manage, and it doesn't leave you any spare capacity when you need it. This way you don't need any spare capacity, and when your business grows you never have to worry about running out of rack space or lead times for new hardware.

The downside, of course, is you're trusting Sun to have the capacity available when you need it. In a way, it really is like a utility - Sun will be hoping its customers "bursts" will all average out to some managable load. I wonder if it's true.

...and

[thepawn]

At 6:18 pm on August 24th, skynet became self-aware.

Business Models

[hiltonday]

What if this is just a pilot to generate some noise?

Surely as a value proposition, Sun could make money on the grid technology (software, architecture etc) by licensing it to owners of large amounts of computer power.

How about if Lawrence Livermore labs and Los Alamos National Labs got together to share their computing power using this grid technology?

Sun get revenues from licensing the grid-management software, and potentially hardware sales....

and the government (or universities, or whoever) gets a larger pool of computing power than any of their single supercomputers, that they can devote to different research groups.

The selling computer-time is a red herring I'd say, based on my experience of Sun.

Re:Business Models

[Anonymous Coward]

I agree that this probibly is purely selling cpu power.

Part of google's strength is their large computing grid, which allows google to launch services quickly and keep the services running under major loads. I think this is to place sun as the grid computing supplier.

The amount of experience sun is going to get with this technology puts them in the lead place to sell to everyone (baring google who have their solution & Microsoft who suffer from Not Invented Here Syndrome).

If it works their will be a major inititive from IBM it will be a threat to their services business, and from Microsoft because the software isn't part of windows.

The interesting possibility is for more software to be designed to work on a cluster, which will create a greater market for clusters, which may solve a chicken and egg situation.

Re:Business Models

[Anonymous Coward]

Well you're part right. The word is that one of the reasons Sun is doing this is to show the pointy heads on customer side that it can be done. The thing is that everyone more or else agrees that utility computing is the way to go, but given the standard psychological state (scared shitless of any chance, understandably) of a large scale IT buyer they are unlikely to actually build one huge grid themselves.

How do you upload the datafiles?

[WoTG]

I understand the economics of renting CPUs for big infrequent calculations, but how do you transfer the GB's of data required? For most business and research apps, aren't massive amounts of data typically attached to massive processing jobs?

Won't the average internet connection take more time uploading the datafiles than it takes to process the data? Is it really practical to ship hard drives full of data to Sun just to run a few calculations?

Re:How do you upload the datafiles?

[museumpeace]

yeah, good question. big deal they have the huge computer...its poor you who has the huge database, possibly with company confidential data. YOu not only have bandwidth issues, you probably have security issues. Like the article said, Sun is still ironing out the wrinkles in the Thin Client software. Once you get the data uploaded, you store it there but that may not help much if the data all change a bit in the 6 monthes that elapse between runs.

granted, there's more windows boxes than solaris..

[discogravy]

Don't sell Unix's (and Solaris') security holes short -- an unpatched Solaris machine is just as vulnerable as an unpatched Windows machine.

SunOS: the reason firewalls were invented.

CPUShare.com

[lbolognini]

Hey Andrea Arcangeli (a Linux kernel developer) has been running a similar project for a while at http://www.cpushare.com/ where you can actually make money off your spare cpu cycles. Lorenzo

Nice

[Anonymous Coward]

Just what we need, more ideas turned into weapons!

There may be peak periods..

[N Monkey]

The kind of people wanting to run CPU intensive tasks enough to be willing to pay someone for it are likely to be universities and companies that could get their own systems.

Chip companies need to run extensive simulations of the functionality of their chips before committing to actual silicon (which is a very expensive step). Although they tend to have large server farms to do this, these might already be busy.

When a (possibly even very small) change is made to the design of a chip you need to run a lot of CPU-time-hungry regression tests. If this is being done towards the end of the project, being able to do the tests in a shorter amount of time (by renting the extra CPU time) could be very beneficial.

I wonder how much I have to pay...

[ngyahloon]

... to get the answer 42.

Sun Is Dead, Not Broke

[gsgiles]

Anybody interested in the fate of Sun Microsystems should watch the great Danny DeVito movie "Other People's Money". Sun is dead, not broke, but they are headed there. Let's not be fooled as to why Sun gives their software away (java, solaris). They cannot sell it. They have consistently offered up the wrong product at the wrong time and at the wrong price. Bill Gates won becuase he is smarter than McNeely, makes better products, and is a vastly superior businessman. Get over it, all software has bugs and flaws, Microsoft is winning becuase it is better, and is getting better all the time. For the final nail in the coffin look at insider trading on yahoo about Sun. McNeely is dumping his stock bought at $.007/share at the market price, $3.70 last time I checked. Draining his stockholders equity in order to line his pockets kind of Enronish. Bill Gates net worth fluctuates more per day than McNeely is worth and for good reason. McNeely has the arrogance to keep his failing compnay in both the hardware and software business during the last twenty years when this has been demonstrated to be the path to failure (Prime,DataGeneral,Integraph,DEC, and SIlicon Graphics come to mind). What does he care he's partying on your (stockholders) money. Express elevator to hell, going down.

Re:Sun Is Dead, Not Broke

[Anonymous Coward]

This is the poorest troll I've read in a while. Insider selling is next to useless as a leading indicator (execs use it as their salary).

Re:Sun Is Dead, Not Broke

[gsgiles]

Agreed but when it is conflated with declining market share, declinign revenue and sales, it is a death knell.

Re:Sun Is Dead, Not Broke

[Anonymous Coward]

However, Sun appears to be turning out of the red this year, their R&D spending is turning out huge new technology, and they are going to refresh their entire SPARC lineup in the coming months. They bought into the huge compliance storage market, they are leading commercial-scale grid computing, they are lining up huge contracts with companies who could have chosen differently, etc. etc. etc.

They seem to be staying afloat for the time being.

Ownership is already a thing of the past.

[SavvyPlayer]

The majority of this discussion seems focused on the potential costs involved using Sun's service vs purchasing and managing your very own n-node cluster. Out of curiosity, who exactly purchases their own equipment these days? Organizations with any cash prefer leasing everything ranging from cellphones to PDAs to desktops to servers to SAN devices. With production equipment often hosted at an offsite facility, managed at least in part by a 3rd party.

The four primary reasons for this are:

1. Comprehensive support.
2. Homogeneity (1000 identical systems can easily be ordered and loaded with the same software image), with additional systems easily ordered on an ongoing basis.
3. Economy of scale -- leasing 1000 identical machines costs less than purchasing, supporting and regularly upgrading 1000 individually customised machines.
4. Automatic periodic equipment refresh, ensuring an org is able to leverage improvements in technology (CPU, memory, storage, bus speed, etc) on an ongoing basis.

Sun's offering is merely the next logical step in an already ubiquitous model.

The *PC* may not be dumb....

[Hasai]

....But after you receive your 56th "I pushed the PF26 thingie and it didn't go" call while chained to the luser helpdesk, you begin to reminisce quite fondly of those halcyon days when you needed a CS degree just to be allowed *near* a computer.

SUN page describing real uses of service computing

[museumpeace]

http://www.sun.com/2005-0816/feature/index.html