Sun Releases Starcat

SilentChris writes: "Sun has released the Starcat server, a beast with up to 106 processors running Unix. Anyone have an extra couple [million] bucks lying around?" They're not cheap.

Comments

Yeah

[SpanishInquisition]

At last a platform to get descent java performance...

Re:Yeah

[WilsonSD]

This is in fact the fastest Java server in the world. Check out:

http://www.sun.com/smi/Press/sunflash/2001-09/sunf lash.20010925.3.html

-Steve

Re:Yeah

[Anonymous Coward]

Fast? How's that? The number is impressive, until you read the fine print. 72 CPU's at 900MHz?? Hell, a 16-way HP box with 750MHz procs will do about 120,000. How is 325,000 fast considering the hardware they used? That's like saying a 2048 way Beowulf cluster is "faster" than a dual 800MHz Apple G4 cube. Of course it's "faster"! But notice the lack of any other benchmark numbers....why ain't they there? Because the US3 chip is a dog with fleas. And who is going to buy a million dollar Sun box to do only Java? At least gimme some NASTRAN or Oracle or SPECint...anything but Java.

And don't forget kiddies, it's cost is way outta line with it's performance. "The new UltraSPARC3 - Last year's technology at today's prices." Their marketing department deserves some credit for saving this one. By pushing cost of ownership and other, decidedly non-traditional Sun qualities (notice that storage plug on their release page?) they're saving Scott's a$$ once again. There's a reason Sun's main page doesn't mention performance anywhere on it....(main page as of this writing being at www.sun.com/2001-0925/feature/ ) - they are behind HP and IBM in the chip race. And, since Sun doesn't have IBM Global Services' income or HP's Inkjet money to fall back on, well, I don't know how they are going to be able to spend enough $$$ on R&D to catch up.

Move over SGI, the Sun is about to set....

anonymous coward, signing off

just wait

[smnolde]

until it shows up on e-bay from a disgruntled former dot bomb employee who five-fingered it from a linux shop which stole BSD code.

Re:just wait

[nebby]

ass.getStick().remove();

Re:just wait

[Strog]

It's references to a couple articles on the front page right now. I read the other articles and found it funny myself.

BSD code theft

FiveFingerDiscount.com

Re:just wait

[Kalani]

Shouldn't that be:

ass.remove(ass.getStick); // ?

What are you removing from the stick?

Re:just wait

[nebby]

Ok Mr. Smarty man, I am telling the stick to remove itself from within the ass.

And I believe your code should have been:

ass.remove(ass.getStick());

But then again, I don't pick nits :)

Re:just wait

[Kalani]

Actually I didn't mean ass.remove(ass.getStick());

Because the getStick member function is an access mediator, it makes no sense to allow an attempt at access if that resource is invalid. Passing the address of the getStick member function to the "remove" member function allows the ass object to place its address into a table of excluded functions. The original function header is retained in this table and the code after stack-related instructions is modified to raise an exception when an attempt is made to access the specified function. That way, the ass can remove any access methods in an abstract way.

It's a very versatile ass.

Not cheap?

[sys$manager]

Did you expect a system that scales to 106 900MHz UltraSPARC III Copper processors to be CHEAP?

Re:Not cheap?

[sacherjj]

For as expensive as it is, I hope they aren't running their current site on it:

The site isn't doing so well.

Just in time.

[Multiple Sanchez]

Now maybe we can set up a decent mirror for the LOTR trailer..

For those beowolf comments

[segfaultcoredump]

Lets remember, that this system is not intended to replace a beowolf cluster of cheap pc's. It is intended to do something that most beowolf clusters can never do: present a single OS image with half a terabyte of memory that any cpu can access at very high speed.

This is a system that is very good at things like fluid dynamics and massive database operations. It is not a good idea if all you want to do is get to the top of the list for the SETI@Home project

Re:For those beowolf comments

[nion]

It is not a good idea if all you want to do is get to the top of the list for the SETI@Home project.

Been there, done that. Tech here working with the StarFire used to run Seti@Home on idle systems. 64 400MHz UltraSparcs. Team Sun@Home rose rather quickly in the ranks those days, I hear. ;)

Re:For those beowolf comments

[Anonymous Coward]

this system is not intended to replace a beowolf cluster of cheap pc's.

Is it intended to replace a cluster of Beowulf jokes?

Re:For those beowolf comments

[Anonymous Coward]

Psshh...

I'd invite you over to check out my liquid hydrogen evaporation-cooled Athlon Thunderbird, but it's been having power supply trouble lately.

When it runs, though, it blows this Sun POS away.

Re:For those beowolf comments

[Anonymous Coward]

yeah right shithead

Re:For those beowolf comments

[sporty]

Well... that's a matter of how you write it, no? I mean to say what's the difference between 106 (what an odd number) processors on this machine, using different threads for each processor versus 106 seperate machines doing the work?

Perhaps in terms of redundancy, mixed cpu speeds,eetc, in the most generic sense, they are the same. Of course, their layout, implementation etc, are QUITE different.

Re:For those beowolf comments

[Omnifarious]

I'd be worried about liquid hydrogen being a conductor. I'd rather use liquid helium. Almost as cold, and definitely non-conductive. Of course, you may end up inducing superconductivity in some of the copper in the chip and on the board, and that might be bad.

Re:For those beowolf comments

[whopis]

Yeah yeah yeah... but just imagine a beowulf cluster of these things!

Re:For those beowolf comments

[Webmonger]

In terms of memory bandwidth and latency, they are very different.

The fastest networking technologies do not approach the speed and responsiveness of a memory bus. Yet a cluster design uses networking in place of a memory bus some of the time.

If there's not a lot of data, it doesn't matter much. If there's tons and tons of data, a cluster design is inefficient.

Re:For those beowolf comments

[sporty]

So in a way, it does replace clusters, in that it replaces that slower networking down. Of course, now its eggs all in one basket.

Re:For those beowolf comments

[Anonymous Coward]

It's too bad Fluorinert is the only commercially available non-conductive liquid. It gels way too easily.

Re:For those beowolf comments

[elmegil]

Maintenance is also a lot different for a single system image versus a beowulf cluster, with tradeoffs either way. To manage that big a beowulf cluster means managing a minimum of 27 different system images, which means 27 times some types of maintenance activity (I'm assuming 4 proc boxes max, so a more typical config would be 53 or 106 system images). On the other hand, if one of those fails, needs to be upgraded, or whatever, the impact to the cluster is probably pretty minimal. Unless starcat does something really different, a CPU failure will cause that single image to fail (which is of course why you'd split the box into two and use an HA cluster of some sort :-).

Re:For those beowolf comments

[Peter La Casse]

If there's not a lot of data, it doesn't matter much. If there's tons and tons of data, a cluster design is inefficient.

Sometimes a cluster is still the only option, however: when you have tons and tons of data, but high cpu requirements. I'm thinking of things like the 30 teraflop project.

Infiniband was supposed to help with this by blurring the lines between the network and the system bus, but it's going to be a long time before it exceeds the capabilities of current technology.

Re:For those beowolf comments

[tolldog]

Agreed.
This would make a killer render system, assuming the renderer can handle that many threads.

This is why beowulf rendering is bad. Network performance for shared memory sucks.

With renders hitting the 2GB + mark for memory useage, do you really want a network passing that data arround.

What could happen with systems like this is that the render time vs. load time would get extremely lopsided. 30 minute loads and under a minute a frame. It would force a rethink of how the render jobs get distributed and ran.Best case would be a few of these, for each different departments render needs. But then we are talking 20+ million for rendering. That buys a lot of intel boxes.
If I was given one, I would try to use it. But I don't think I could ever seriously suggest buying one. But that is me and my particular application.

Re:For those beowolf comments

[Anonymous Coward]

Well, this Sun, and all of SGI's line are not homogenous memory architecures. They're offshoots of NUMA. Most of these machines are 4 way boards with their own local memory.

They can access non-local memory via a high-speed crossbar network (SGI's is called Craylink, dunno about Sun's).

The architectures are pretty similar because when SGI bought Cray, they spun off the lower end stuff to Sun. Ironic that Sun got the reputation for high end servers with the low end of what SGI uses now. An Origin3800 has been available for some time, and blows the doors off this new Sun machine.

Guess it just goes to show that technology is completely irrelevant for market success.

Re:For those beowolf comments

[ivan256]

Funny, I can get water right out of my faucet, and you can get mineral oil at the grocery store... Where are you from that you can only get Fluorinert?

(Of course I'd deionize the water before I poured it on my PC...)

Re:For those beowolf comments

[Anonymous Coward]

The water freezes way too early to be useful around the liquid hydrogen, it's liable to break hardware during expansion.

Here in Antarctica you'll be lucky to find running water. No, we don't have the Fluorinert mail-delivered.

Re:For those beowolf comments

[mallie_mcg]

Unless starcat does something really different, a CPU failure will cause that single image to fail (which is of course why you'd split the box into two and use an HA cluster of some sort :-)

Unless I am very miskaten most Sun boxen if a CPU fails, stops using that CPU and moves (or does it have to be told??) all processing away from that CPU. That way the system can stay live, albeit slower.

Re:For those beowolf comments

[Valgar]

On the bigger boxen, you have the happy ability of dynamic reconfiguration. Plus the kernel_cage_enable=1 tag in /etc/system (There is more but I'm going to keep this simple).
I'm pulled entire memory/cpu boards out of running E4500's without a problem when cpu failure occurs.
I even have had to script dynamic online-offline of processors for various load testing purposes to determine how the system scales as CPU's are added/removed
If there is one thing about the big iron I like, it is the ability to manipulate the system on a level such as this.
Now Stratus systems are even more amazing...but as they said in Conan. "That is another story"

Re:For those beowolf comments

[Ryan Amos]

Unless they're using superconductive alloys (which are not at all conductive until they hit their temperature) then that shouldn't be a problem. There probably is some sort of superconductive activity if they're using liquid hydrogen, in which case they WANT it to be superconductive.

Re:For those beowolf comments

[Omnifarious]

At extremely low temperatures, ordinary bulk metals become superconductive.

Re:For those beowolf comments

[Anonymous Coward]

Good luck finding something that is non-conductive, cheap, AND doesn't freeze or gel at low temperatures. Mineral oil and deionized water tend to dislike extreme temps.

Re:For those beowolf comments

[jaoswald]

Listen up.

The "extremely low temperatures" for good room-temperature conductors to become superconductors is lower than you can achieve by using cryogenic liquids at atmospheric pressure, or even by achievable lower pressures. Liquid He boils at ~4K at 1 atm. Liquid H2 boils around 20K.

Aluminum becomes superconducting at about 1.19 K at zero magnetic field. Cu, if it superconducts at all (I vaguely recall someone reaching this point, but I'm not sure), has a critical temperature in the millikelvin range or lower.

To clarify: 1) liquid helium is *colder* than liquid hydrogen. 2) don't worry about good conductors becoming superconducting without a low-temperature physicist involved.

Now you can get back to discussing other topics you don't understand.

Re:For those beowolf comments

[colmore]

"Here in Antarctica"

Maybe this comment is eight years too late, but sometimes the internet still amazes me.

Re:For those beowolf comments

[Anonymous Coward]

Aluminum becomes superconducting at about 1.19 K at zero magnetic field.

That's funny - so does aluminium!

Re:For those beowolf comments

[Omnifarious]

I can't see how anything you said contradicts my post (well, aside from the liquid helium being colder. I actually kinda thought it was, but assumed the previous poster knew more then (s)he did about low temperature physics). It was kind of a ludicrous thread anyway. I did talk about using liquid helium, and I was well aware that the superconducting temperatures were in the range of 0-2k. Yes, it ridiculous to assume that you'd be able to get good enough thermal conduction away from the processor to maintain it at these temperatures, but it's silly to think about cooling it in liquid helium in the first place. :-)

Re:For those beowolf comments

[mad_dwarf]

The 106 CPUs is quite bizarre. The Starcat has 18 System Boards and 18 I/O Boards..

18 system boards with 4 CPUs = 72 CPUS

BUT! They have this cunning little thingy (can't remember the name of it) that allows you to replace a I/O board with a board populated with 2 CPUs (but no memory), giving

17 I/O boards with 2 CPUs (got to keep 1 I/O board in obviously!) = 34 CPUS

Giving us a grand total of 106 CPUs.

Don't know many people who will sacrifice most of their I/O to get the extra CPUs.

Re:For those beowolf comments

[Anonymous Coward]

As the original troll who started this (actually quite interesting) thread, I admit I have no super-cooled Athlon system at home. I am not in Antarctica. And I have no access to Fluorinert.

But thanks to all those who piled on and made this the interesting thread that it was!

Re:For those beowolf comments

[elmegil]

Unfortunately, you're very mistaken.

The vast majority of Sun's boxes, including this one, are not fault tolerant. The CPU fails, the box panics, diagnostics presumably notice it's failed and map it out, and on reboot you no longer use the failed CPU. That's not the same thing.

distributed.net, RC5

[Anonymous Coward]

if all slashdot users throw in a few bucks, we could buy one of these and really kick some ass over on distributed.net. cmdr taco can pick up the electricity bill.

106?

[adaking]

What's special about 106? How did they come up with that number? I could see making the upper bound a power of two, a perfect square perhaps, but 106? Perhaps they were shooting for 128, but couldn't get those last 22 to play nicely with the others?

Re:106?

[Cheetahfeathers]

It has to do with how many CPUs fit on a board, and how many boards fit into the box. Get a bigger box, and that architecture can theoretically handle 1024 processors. :)

Re:106?

[segfaultcoredump]

The system grows to 106 in the following way:

There are 18 "cpu/memory" boards that hold 4 cpu's each. This brings the system up to a total of 72 cpu's and 576GB of ram.

Now, if you want an server that just does number crunching and dont care about I/O, you can then add 'MaxCPU" modules. Each module holds two additional cpu's (no memory) and occupies the hPCI module slot (a hot swap PCI case that can hold what looks like two to four pci cards). You can use up to 17 of the hPCI module slots to hold MaxCPU modules. (there are 18 pci channels on the system, and at least one must be used for accessing the boot disk).

So there ya have it, 106 cpu's and half a terabyte of ram. I think that in most cases, folks will opt to not use the MaxCPU modules and just stick to the 72 cpu limit.

Re:106?

[BLAG-blast]

Maybe the performance with 107 CPUs is lower than 106 CPUs. Adding more CPU does not equal more power.

Re:106?

[antek9]

Maybe they're just being sponsored by some Peugeot nostalgics. ;)

Re:106?

[Cheetahfeathers]

You can have each CPU board running a seperate OS, attatched to a seperate set of devices such as hard drives, ethernet, etc. In this case, more CPUs are more power. But if you are running them as one OS instance, yes, it might get a bit silly, at that level. It depends on what you're doing with it.

Re:106?

[Anonymous Coward]

Hah! I know insider info when I see it. Its only a matter of time until they can fit 4 processors on the so ersatz maxcpu board

Re:106?

[BLAG-blast]

You can have each CPU board running a seperate OS, attatched to a seperate set of devices such as hard drives, ethernet, etc. In this case, more CPUs are more power.

I've found that to be a nice idea, but if you try and use all the CPU power at the same time you'll find many bottle necks that reduce the combined CPU power down to less than one CPU suprisingly fast.

I'm always impressed by anything with over 16 CPUs that really works. And with the price Sun are pushing these babies at, you'd expect them to work well.

I miss transputers, I miss Occam, hmmmm, anybody know of an open source Occam project?

Re:106?

[cholokoy]

I think the world will probably need about 5 of these machines and not more .... :)

HP and IBM already has offerings that can equal Sun's and its only trying to play catch-up. These are machines that are solutions looking for a problem because its not easy to implement projects of this magnitude.

Re:106?

[Anonymous Coward]

It's because of the latest release of
Kaffe.
That way, Sun can have a good Java platform.

Re:106?

[fgodfrey]

Let's not forget SGI with the Origin 3000 (a system which I worked on a lot when I worked for SGI). It scales to 512 processors in a single system image (perhaps 1024 by now?) with a terabyte of RAM. In addition, you get 128 I/O channels with that each of which can have up to 12 PCI slots (or 4 "legacy" XIO ports). If you want redundancy, you can partition the system into multiple OS images. A 128p Origin as a single system image regularly gets over a month of uptime without a failure. So basically, while Sun can make this look like the best thing since sliced bread, it *has* been done before.

Re:106?

[elmegil]

Not with SMP it hasn't.

Re:106?

[agallagh42]

"I think the world will probably need about 5 of these machines and not more .... :)"

Yeah, and nobody will ever need more than 640KB of RAM. My company alone has two E10Ks (the predecessor to this beast), so I think there's a decent market for these...

Re:106?

[Furry Ice]

This isn't SMP, either. It's a NUMA architecture.

Re:106?

[Furry Ice]

Oops! I retract this statement.

Re:106?

[Anonymous Coward]

IBM and SGI can match the StarCat, but HP's big "Superdome" only supports 64 CPU and 256GB RAM. So Sun more than doubled HP's max RAM, and almost doubled number of CPUs using faster CPUs.

Companies are scrared of buying something in 1-4 million range from a company that might not be in business that much longer. SGI is very close to dead.

Solaris has a lot more support than AIX.

The E10k were aways behind the competitors, but with these machines Sun is back in first place.

Re:106?

[Anonymous Coward]

"A 128p Origin as a single system image regularly gets over a month of uptime without a failure."

And you are proud of that?

Re:106?

[Anonymous Coward]

"HP and IBM already has offerings that can equal Sun's ... "

Okay, I'll give you that HP's SuperDome is a pretty impressive system, but IBM has nothing in this ballpark except the z900 mainframe, and the z900 is less powerful than a two-year old E10000.

Re:106?

[Anonymous Coward]

What's the point of SMP when your bus-speed isn't much faster than the speed to remote memory on a NUMA box?

Re:106?

[Anonymous Coward]

The SuperDome is marchitected for 64-cpus, but is actually architected for 256 cpus (and thus 1TB of RAM). HP sells mostly commercial boxes and their marketing division thinks that the commercial market for boxes above 64-cpus is not worth the effort of fully qualifying larger configurations. But, if you really, really wanted one and were willing to pay the premium, you could probably get a SuperDome that big.

Plus, the SuperDome doesn't need to trade off I/O for cpus and even with the maximum I/O (and only 72 cpus) the starcat is way behind in the number of available PCI slots. A 64-way superdome will support 192 PCI slots, the Starcat will only do 72 slots max and every cpu you add above 72 cpus takes away 2 PCI slots.

As for cpus, don't be a megahurts sucker. The 900MHz Ultrasparc 3 was just a tad faster than last year's PA-RISC PA8600 at 550MHz in almost all objective benchmarks (spec, etc). HP's now shipping 750MHz PA8700 chips in their superdomes, these monsters have 1.5MB of data cache and 750KB of instruction cache ON CHIP, none of this pansy 32KB on chip cache. You can figure that an 8700 at 750MHz is probably about 1.5x faster than a 900MHz UltraSparc3 on a workload like spec.

You are right, SGI is almost dead, but SGI never got it together enough to play in the commercial market and thus rarely competed with Sun. In the technical market, they regularly kick Sun's ass - and the users don't care so much about the company's viability either, as long as they get their hardware and it works.

AIX does not have as many applications as Solaris does, but all the Big Boys are there like SAP, Oracle, etc. And the IBM Power4-based 'regatta' is going to kick some serious ass next week. 32 cpus, but each cpu is dual-core and their bandwidth to memory is massive. The regatta will probably easily beat out the Superdome performance-wise, which means the starcat will be left so far back in the dust that it isn't even worth looking back at.

So, if you are a sun shop, and will always be a sun shop, then the starcat is great, but if you have the option of considering other vendors, HP and IBM have got Sun licked at this level of performance.

Re:106?

[fgodfrey]

(Yeah, I know, I'm feeding the trolls) Please provide an example of another system providing a total single system image which does better at the same system size. You may have problems since there aren't that many machines out there that grow to 128p (I suppose the 106p Sun box is "close enough").

Re:106?

[Anonymous Coward]

It has to do with the size of the box and whether it will fit in a standard size truck, or fit in a standard size shipping dock. It already weighs over a ton, and will barely fit through most datacenter doors. The only way to go beyond this would to use a slower interconnect to join multiple Starcats together

Scheduling airplanes?

[sllort]

"those conducting computationally intense tasks such as scheduling airplanes"

Huh? I understand that the nation's air traffic controllers may need updated equipment in light of the existing crisis, but how hard can scheduling be? I could see a use for a massively parallel monster like this in, say, flow-through or structural analysis or something, but scheduling?

Any aviation experts out there... what am I missing?

Re:Scheduling airplanes?

[scott1853]

Ha ha, we created an online aircraft scheduling system that runs on a P2 with 256MB RAM, with IIS and MSSQL. I think you're right about that being an overkill. With that much power I'd expect it to simultaneously guide all the aircraft that are in the air at any given time.

Re:Scheduling airplanes?

[Derkec]

"those conducting computationally intense tasks such as scheduling airplanes"

"Huh? I understand that the nation's air traffic controllers may need updated equipment in light of the existing crisis, but how hard can scheduling be? I could see a use for a massively parallel monster like this in, say, flow-through or structural analysis or something, but scheduling? "

What your missing is that this isn't a matter of airtraffic control. This is a matter of determining which planes and crews to fly to which locations at what times to maximize revenue. This is a classic, big, nasty travelling salesman problem. The bigger of a beast of a machine you get, the closer you get to an optimized solution. I.E. Most passengers willing to pay this most money with the least use of resources. It's a huge problem that needs massive computational power.

Re:Scheduling airplanes?

[heliocentric]

But a lot of the cost of doing business things, which can be done using a graph, don't have to be.

Look into operations research. An early on thing you will learn in OR is the simplex method, where in you boil down a lot of your information (cost to move part A from here to there and buying B if needed, etc...) into a matrix and then use something like the Simplex Method you can maximizie say your profit, or minimize the delays, or minimize the total number of planes in the air, etc..

Now, does the world's flight system have a lot of inputs - you bet, planes, fuel, flight crew, union regulations, holidays, tons of things that you and I could probably never think of...

Now... I just wanted to post that there are ways of solving these problems without a graph - yes a graph is a lovely way for highly dynamic systems, but if you want to answer the question "What is the best schedule given these 1,000 limitations?" then look into operations research. Yes, I mentioned the simplex method, and I fully exepct following posts arguing Parametric Linear Programming or maybe some Markov Chains and their impact, but the core is still the same - if you have a set of limitations and are looking to maximumize (or minimize) something, then operations research is a fun thing!

Re:Scheduling airplanes?

[Anonymous Coward]

bullshit. Im a sqlserver dba and i dont think 256 MB of RAM can handle that many connections. unless the database isnt really doing anything

Re:Scheduling airplanes?

[Anonymous Coward]

Um since traveling salesman is a classic "hard" computer problem and there are thousands of aircraft with hundreds of airports thats a BIG problem. Also not all aircraft are identical nor are the routes. There are different route lengths, different salable costs for different routes (demand based) and many other factors. This is the kind of thing that big pc's can be used to improve. Process management in the airlines industry which has billions in capital at stake is a place where throwing a couple million at a computer seems like chump change.

Re:Scheduling airplanes?

[SilentChris]

"The bigger of a beast of a machine you get, the closer you get to an optimized solution. I.E"

Sort of accidently read that as "this machine is optimized to run IE". :) Don't mod me down, I submitted the story. :)

Re:Scheduling airplanes?

[scott1853]

I didn't mention the number of connections or even what type of aircraft it is. We cater towards small airports. A 747 couldn't even land at our customers airports. At most, we might have 40-50 simultaneous connections, which are all handled through perl scripts.

Couldn't help but notice you hid behind an AC name while mentioning you're a sql server dba. Why is that?

Re:Scheduling airplanes?

[acacia]

Um, there's a lot to it, that's why. Airlines operations are basically a big exercise in logistics.

My current client has initals that might contain the letters L, A, and U, and they use an RS/6000SP for exactly this type of work. Nice machine. This Sun box is a direct competitor to the SP, but very different architecturally. Large scale SMP vs MPP.

Clarification please

[scott1853]

Not being a server dude, are these things parallel processing machines or just a server farm contained within a refridgerator?

Actually for that price I hope it functions as a refridgerator, as well as a dishwasher, robotic maid and would provide protection during nuclear fallout.

Re:Clarification please

[Cheetahfeathers]

These can be either. It depends on how you configure it. And the fun this is, you can reconfigure it on the fly. You want a cluster in a box? You got it. You want 2 seperate instances of Solaris running, each using 1/3rd the resources of this box, while you pull out the hardware on the rest of the box for maintenance? You got it. This thing is _configurable_. You can hot swap everything except the backplane, pretty much. It's _sweet_.

Re:Clarification please

[elmegil]

You want a cluster in a box? You got it.

Whoah, Nelly. You need to be clear on what kind of cluster you're talking about here....if you want an HA cluster, there are always drawbacks to putting all your nodes in one box. s/nodes/eggs/ && s/box/basket/.

Re:Clarification please

[Cheetahfeathers]

Yes, there are drawbacks. But you all but can have an HA cluster in that one box. You can have everything except the backplane be redundant. Seperate power, ethernet, disks, CPU, memory, etc. It is not _quite_ as HA as properly seperating it into two seperate boxes, but it is pretty darn close. Properly put together, this would be more HA than many clusters with seperate boxes on each node of the cluster.

Not that I recommend that way of doing it, but it can be done. Of course, a proper HA cluster of these with one seperate cabinet for each node would be the best. :)

Re:Clarification please

[theyman]

Center plane, please ;)

That's it...

[daeley]

From now on, the official phrase should be "Whoa! Imagine a Starcat Cluster of those!"

And is it just me or did the old "Thundercats" show just pop into your head?

Re:That's it...

[Magumbo]

Um. I believe Starcat Cluster is a trademark owned by the Little Debbie Food Group Inc., Lubbock Texas. It consists of two Star Crunch patties glued together with Mallow Kreeme filling. Then dipped in a Chocolastic sealant.

Just watch out. I hear Debbie has quite a few lawyers.

Re:That's it...

[jayhawk88]

4 overweight geeks just had heart attacks thinking about that hypothetical snack food.

store.sun.com

[Wiggins]

Maybe they need one of those things powering sun.store.com as when I hit the page I got the following:

Configuration Error

1) Error calling config servlet: sunir.webdesk.common.checker.ConfigInternalExcepti on: Couldn't get sql connection

Then I tried to reload the page and didn't even get a response....

Re:store.sun.com

[Wiggins]

sorry about that... store.sun.com in the comment not sun.store.com......

They need one!

[MSBob]

Sun themselves are in a need for boxes like that as their website seems severely slashdotted right now.

Re:They need one!

[Anonymous Coward]

You're a retard. Don't you think the bandwidth might be the limiting factor? It works great from here, you should check with your service provider.

Re:They need one!

[Anonymous Coward]

Yeah. Internally, Sun uses rather old hardware. There are still a lot of SS20's around, doing server duty! Of course, they are still running well, so replacing them would mainly be to make it easy to swap out bad parts quickly, should something fail. Easier to replace a CPU on an U10 than a SS20... you don't need to hunt the U10 parts down, and hope you get luck. They are in stock already.

Luckily, Sun built them to last, in the SS20 days.

Re:They need one!

[friscolr]

sun put the dot in dot com,
but slashdot took it out.

*grumble*

[Doctor_D]

I've been submitting this story since 10:30 (Eastern Time) this morning when information became available from sun.com about the starcat. Three submissions and all three rejected, and then after my third rejection, this finally gets posted.
</rant>

Re:Come on people!

[antek9]

What's so funny? I already ordered my beowulf cluster of these, 20 pieces with all the additional options checked. Glad that my visa card was up to the challenge.
Or why do you think ordering of these babies is already Constrained: Demand for this product is so strong that it exceeds availability and is subject to shipment delay. If you proceed with your order, Sun will send you an e-mail with the scheduled shipment date as soon as possible. ??

*yawn*

[eXtro]

This isn't even a really impressive box. I'd rather have an sgi O3K system if I'm going for the ultimate in servers you can actually purchase. The SGI Origin 3800 has anywhere from 16 to 512 processors, 716 GB/sec system bandwidth and up to a terabyte of memory. It's also a single system image machine. Oh yeah, and you can cluster them to scale way beyond 512 processors.

Re:*yawn*

[Temkin]

And will they be around in two years????

Re:*yawn*

[Derkec]

My understanding is that SGI has been losing market share fairly steadily. What's wrong here?

Re:*yawn*

[Erik Hensema]

They are available now. See this beast. Sara's been running the system for about a year now, IIRC.

Re:*yawn*

[Urgoll]

True, an O3K can scale farther than the Starcat, but it isn't as redundant, doesn't have domains, and cost a LOT more than the StarCat.

SGI's future prospects are also dim, it's not a safe long-term purchase.

Re:*yawn*

[Anonymous Coward]

marketing. duh. IT decisions are not made based on technical merits.

SGI can't market worth shit. Sun is second only to Microsoft in marketing.

Re:*yawn*

[Anonymous Coward]

Oh, Yes, SGI. One does not seem to have enough trouble already. Where will SGI be in a year or two? It is anybody's guess.

Forget about the useless company which invited a brain-dead HP paperpusher as the CEO, who subsequently broke the company into pieces -- and then got rewarded by Microjunk's Mafia Don. For sure, the issue about SGI is not a technical issue, but they lost their credibility when they trumped their NT box and direction to the sky, but only got hammered when the box fell back on their heads.

We should thank Sun Micro (and Scott McNealy) at least for holding up the Unix front during that dark period of time. Without Sun Micro, Unix font will be self-destroyed by HP/Dell/Compaq/SGI collusion with Microsoft.

Re:*yawn*

[Jon_E]

yeah - but look at your iterconnects - NUMAlink?? - last time I checked not necessarily the best in bus contention and only at 3.2Gbps?

these Starcats are doing 43GBps on the backplane .. add to it the UIII 900MHz procs and it's a pretty sweet deal for I/O intensive apps

Re:*yawn*

[mhandlon]

Your outa your fucking mind HP has done alot more for Unix then Sun EVER did.... Jesus, Sun makes shitty servers/workstations and give everything away to colleges for free so the "new crew" learns on it and everyone thinks it's the shit.

HP gives away nothing so they're the enemy?

Re:*yawn*

[Refried Beans]

I think you missed the partitioning link on the web page. Also, how do you upgrade a StarCat? With an O3k, you just keep adding to the system. You never have to get rid of your old hardware. No more fork lift upgrades.

Re:*yawn*

[Anonymous Coward]

Check again, that's 3.2GBps per connection vs 43GBps overall system bandwith (which presumably means cross-sectional bi-directional). So Sun's connection bandwith would only qappear to be 2.39 Gbps. Impressive, but it doesn't beat the Origin 3K.

Of course the Sunfire is configured as a star, vs the more complex bristled cube/hypercube connections in the Origin with multi hops, so the Sun should win in latency.

IO intensive is no contest -- the Sun only scales to 72 PCI slots, the O3K scales to hundreds of PCI slots.

As far as SPECmarks -- 900MHz Sun wins on int, but a 500 MHz SGI wins fp by a decent margin.

Of course in the financial stability contest it's Sun all the way...

Re:*yawn*

[elmegil]

You don't get it Erik. The question was will SGI be around in two years, or will they have joined DEC (and their wonderful Alpha chip) on the bonepile? What good is an O3K if you can't get it serviced?

Re:*yawn*

[Anonymous Coward]

You clearly don't understand the differences between SMP, CC-NUMA, and cluster systems. Read the book "In Search of Clusters" which does a great job of explaining them.

Re:*yawn*

[Jon_E]

sorry for the misinformation - if we're talking peak system bandwidth it's actually at 172.8GBps (see http://www.sun.com/servers/highend/sunfire15k/spec s.html) and 43.2GBps sustained (9.6GBps and 2.4GBps per system connection across 18 boards respectively) - and right bisection bandwidth ..

IO Bandwidth up to 21.6GBps sustained .. true - only 72 PCI slots, but if they can interconnect into the the other Starfire line - you've got a low-cost expansion on the I/O side that would be pretty impressive ..

also keep in mind that Sun changed how it handles bus contention - so the benchmarks on these should look pretty sweet

Re:*yawn*

[phajek]

In hope of educating you, consider this. Sun and it's founders: vi, Virtual File System, NFS, Java, etc...

Now people take issue with HP for a valid reason.... During the mid 90s, HP spat on their HP-UX division in favour of being a strong NT supporter. It is *true* during this period that Sun was the only major vendor that stuck by UNIX.

Re:*yawn*

[defender]

that's rather unlikely.
The SGI systems have a rather different target market. SUN is targeting commercial customers. SGI's Origin and Onyx lines are targeted at a bit more scientific uses.
For example the O3800 Erik is talking about is used for scientific research. Apart from stability, speed and performance are where the Origin and Onyx systems *excel*.
For example, the "CAVE" system is powered by a bunch of Onyx2 boxes: http://www.sara.nl/Customer/vr_nieuw/facilities/ca ve/index.html
Also, if I remember correctly a lot of US government agencies also employ SGI/Cray systems in favor of SUN/HP...

Re:*yawn*

[elmegil]

So if the excell in stability, speed, and performance (though the distinction between the latter two eludes me), why are they only used for scientific research? After all, a business looking for 5 9's uptime is going to want those things in spades. I haven't heard too many scientific researchers insisting on 24x7 uptime (with a few exceptions).

Power! Must have power!

[supabeast!]

You know, if the brain had one of these to run simulations on, he an pinky probably could take over the world!

Not a good sign

[bstrahm]

Tried to get to Suns Order page to spec out a system for fun...

No such luck. I wonder if we /.ed DELL we could bring it to its knees. What kind of servers are running on each site ???

Of course

[Anonymous Coward]

They are using servlets. What do you expect? Anything else would be better cgi to perl, python, c or any other language. Just forget Java.

Re:Of course

[codingOgre]

Would have made a good troll if your comment made sense. They need to allow more database connects jack.

I need one of these for home

[hillct]

Now I can model the water flow down the shower drain (fluid dynamics). Cool. But can it model the flow of $$$ down the drain that would occur if I purchased it?

I'm all for new and spiffy hardware, but some things are just too much. I woner what the margin on one of these things is... I'm having trouble seeing how they expect to sell enough of these to cover the R&D costs... Maybe they'll sell them to the NSA so that agency can be brought into the 21st century...

--CTH

Re:I need one of these for home

[nebby]

Yeah, building extremely fast computers to solve new types of problems is a stupid idea. Thanks for the reality check, you probably just saved all those gullible physicists lots of money.

Re:I need one of these for home

[Doctor_D]

You've got to take your return on investment into account. Which would you rather have, an entire row in the datacenter filled with NT servers, or entire row filled with other *NIX boxes. Or just have a few of these in your datacenter and the rest filled with storage? If you look at it, having one of these and a wad of storage makes sense. Especially if you're running a worldwide corporation and maintaining their SAP and Oracle on the box.

Before I started working for Sun I drooled over the E10k, but thought I'd never see more than one in a computer room. How wrong I've been. I've since walked into datacenters and seen rows of 10k's all humming along. And I'd imagine I'll see the same thing with the 15k's in a few years.

Re:I need one of these for home

[Anonymous Coward]

Yea, same thing said about starfire. Yet they've sold close to 5000 of them. Go figure.

You don't get to leading edge technology by penny pinching.

Re:I need one of these for home

[Anonymous Coward]

Ummm they probably won't see the money for the R&D from selling these servers. But the technology used for these will trickle down to the lower end equipment and they will make their money there.

All I want for Christmas...

[digital_freedom]

Dear Santa,
I've been a real good gEek this year. I wrote several white-hat worms to fix IIS holes. I defended IP rights in the Linux kernel. I also mirrored the LOTR trailer.
Could I please get just one little old Starcat Server from Sun? Please make sure it is the 106 processor version with 576 GB of RAM.
I will be real good and use my idle time for SETI.

Your pal,
digital_freedom

P.S. Chocolate chip cookies are your favorite right?

Re:All I want for Christmas...

[Pope]

You could also host a lot of Return to Wolfenstein games...

Re:All I want for Christmas...

[Anonymous Coward]

im almost positive the server is onyl available for x86

Re:All I want for Christmas...

[yesthatguy]

I'm sure with that much power, you could find a way to emulate x86 without too much noticeable performance loss in hosting games :)

Re:All I want for Christmas...

[stefanjo]

Like anyone would pay that kind of money and then emulate a x86 on it :P Thats like emulating a c64 on ... hmm wait.. people do that.. oh never mind :)

/. the dot in dot com??

[Triode]

Hmph, seems as though the dot in .com can not hang with a little slashdot on them. I can ping it, but no response from port 80 from here...

Perhaps they should use one of those bad boys for the webserver?

Re:/. the dot in dot com??

[mach-5]

Looks like they should hook the StarCat up to their website.

I think their database is not running on a Starcat

[bartjan]

Configuration Error

1) Error calling config servlet:
sunir.webdesk.common.checker.ConfigInternalExcep ti on:
Couldn't get sql connection

106 CPUs?

[algae]

According to the specifications, it has support for up to 18 quad-processor boards. That's a total of 72 processors, so where are they getting the 106 number from?

Wow, up to 72 *hot swappable* PCI devices!

Re:106 CPUs?

[Anonymous Coward]

In addition to the 72 procs on system board, you can replace most of the hot-swappable PCI daughter boards with procs. So 72 proc on board + 34 PCI board swapped for procs and you get 106 procs.

Re:106 CPUs?

[Anonymous Coward]

No, you swap the complete IO board with a board that has two additional processors on it.

They will

[Anonymous Coward]

Trust me. The generation just before these (StarFire 4880, AKA the Serengeti line) were 1.5 million a piece and they sold something like 100 to the Navy all at once. They sold asstons more to private sector, but some applications do require such processing power.

Yeah, brain, but

[Anonymous Coward]

could we play quake on it?????

:)

Cool

[Buzzwang]

I'm glad to see that some companies are at least trying to accomplish new things and come out with new products given the state of our economy and markets and such. Even if people think it is overpriced or under-powered and what not that still doesn't degrade the fact that it is a relatively new product in a squishy market. Personally, I own a few hundred shares of SGI stock, but I'm still happy to see any tech comapny suck in their gut, tighten their belt and release a new product in this market. Makes me want to believe the tech markets will turn around sooner than people believe. Kudos to Sun for still working on new products and trying to generally improve things. Now, if Cray would follow suit, I'd be a happy man...

Re:Cool

[Anonymous Coward]

Uh, you can bet your bippy that the starcat has been in development for at least four years now. Sun really couldn't have delayed much longer - HP's already on their 2nd generation SuperDome and IBM will be announcing their butt-kicking Power4 platform in about a week. All the vendors, even Cray, are constantly working on new products and the slow-down in the market since Y2K really hasn't gone on long enough to make a significant dent in their plans. (Now, the plans for the next generation of products has probably been impacted at all vendors, but you won't see the effect of that for another year or two.)

Re:Cool

[cweber]

It's not like they started to work on this thing AFTER the economic downturn started. We were briefed on it under NDA almost two years ago, and it didn't sound like they were just starting with it. Projects like that take a bit of time. Too much time in fact for us - we ended up bying SGI gear which was ready and proven at the time.

Re:Cool

[elmegil]

I'm thinking since Sun bought part of Cray from SGI when SGI ate Cray, and those engineers did the work on the StarFire (E10000), they very likely did this box too....

Re:Cool

[Anonymous Coward]

Pete, since you work at Sun, you ought to know...

Re:Cool

[Anonymous Coward]

Well, I ought to know. Yes, the same peeps who did starfire did starcat.

'nuff said.

Re:Cool

[Anonymous Coward]

Did I say I didn't know?

Re:Cool

[Anonymous Coward]

Yep, implication is reality.

Re:Cool

[Buzzwang]

I didn't loose any money, I bought SGI stock after it tanked in hopes that it would rebound in the long-term. As a side not, I also own stock in Cray. :)

I don't think SGI is gonna disappear anytime soon, and Cray is still turning out super-machines same as Sun and a couple other makers. After the market relaxes a little and gets going again, things will pick up and start cruising. I'm just waiting for it to happen.

I want 2 of them!!!

[CaptIronfist]

I wonder how many players can 2 dedicated Tribes 2 Starcats can take. ( 106 cpu each ... mmmm )

That's, of course, if Tribes 2 ever gets ported to S'low'aris. ;)

Re:I want 2 of them!!!

[DNS-and-BIND]

Solaris picked up the "slow-laris" nickname from its performance on tiny single and dual processor boxes. It's not an operating system optimized for small things...you need at least eight processors before you begin to really realize performance benefits.

You can forget about Tribes 2 being ported anywhere...didn't Sierra close the doors? If Tribes 2 were ported, it's far more likely the x86 arch would be chosen as a target instead of the SPARC arch, as game-playing weenies don't care for anything without a BIOS.

Re:I want 2 of them!!!

[Anonymous Coward]

as game-playing weenies don't care for anything without a BIOS.

*ahem*. You seem to have forgotten to include the word "afford" in there.

Re:I want 2 of them!!!

[Anonymous Coward]

Ultra 1s and Sparcstation 5s and such are surprisingly affordable. And, there's always the Blade 100. However, these all lose badly in the megahurtz department, and are likely to be scorned by any serious gaming weenies.

Re:I want 2 of them!!!

[CaptIronfist]

didn't Sierra close the doors?

Actually, no. They closed down Dynamix, however the game is still Sierra's. So they transferred a part of the team to Sierra and they are now making patches. The Tribes 2 community had one yesterday.

Solaris picked up the "slow-laris" nickname from its performance on tiny single and dual processor boxes

Really ? I thought it picked up the nickname because of its incredible slow interface...

Interresting. ;)

The marketrons are going to _freak out_.

[Anonymous Coward]

"Wow, look at all the hits we're getting on the Starcat shopping cart! We're going to make a mint on these suckers!"

Re:The marketrons are going to _freak out_.

[SunManAndy]

I love the "Buy Online" tab at the bottom of the "beast" link!

I think I have room on my MasterCard...

partitions

[cornflux]

CNET article: The system can simultaneously use 900MHz processors with faster models yet to come. However, each partition requires all processors to run at the same speed, so faster chips will have to run in a partition of their own.

As someone who does nothing with these types of systems, nor follows them, I think it's great that you can have different processor speeds using "partitions."

I wonder if memory is treated the same way... i.e., separated by "partitions," or if you also have a choice to use it as one, large unified memory resource... or, I wonder if memory can be dynamically partitioned... hmm.

Actually, now that I'm thinking about it... are all of the processor partitions considered peers? I mean, are the partitions all treated as if they were a single processor... then treated equally?

Re:partitions

[Cheetahfeathers]

Yes, memory can be partitioned off dynamically. As can hard drives, ethernet ports, SCSI controllers, etc. Pretty much everything in this puppy is dynamically partitioned.

Re:partitions

[Doctor_D]

According to the specs each processor board holds 4 processors and 32 gigs of memory.

Now, if the starcat treats domains (partitions) the same as the E10k (I haven't been to training yet on it), then each domain at minimum will consist of 4 processors and 32 gigs of ram, ie 1 processor board. Basicaly these doamins are treated as seperate boxes as far as Solaris is concerned. You configure a domain to say contain 2 system boards, and then when you load Solaris, it then sees 8 processors and 64 gigs of memory. This way you can allocate resources as the need fits. But this means it doesn't look like the virtual processor that mainframes present.

The starcat may deal with processors above 72 in a different way, but I honestly don't know at this time how it deals with them.

Hope this helps answer your question.

Re:partitions

[Cheetahfeathers]

All the Sun Fire models (15k, 6800, 4810, 4800, and 3800) are based on the E10k technology. They are basically the next generation of the E10k. Now just wait and see what Sun puts out for it's _really_ high end server. ;)

Re:partitions

[cornflux]

Not really a follow-up, per se... but...

The Register has an article about the launch of the StarCat including a quote where McNealy said, with tears in his eyes, "God I hate my job."

Apparently, McNealy had a hard time speaking during the event, which was held in New York City, due to the death of a long-time Sun employee in the terrorist attacks on the WTC.

Re:partitions

[DNS-and-BIND]

Sun Fire servers use the UltraSPARC-III processor, the E10000 and friends use the US-II.

Re:partitions

[theyman]

Not only can they be partitioned by system board (so long as a system board has a CPU, memory and access to disk it can be made a domain) but, assuming it follows e10k functionality, you can 'blacklist' any component if it starts mis-behaving and it won't be used again until you say it's OK. Unlike 'redlisting' (ask your friendly sun bod about that...)

I'll bet their hoping

[ch-chuck]

to get on the National US ID Card database bandwagon with Oracle... It'll only need to store about 300 million records with DNA, fingerprint, picture for facial recognition software, key escrow, etc...

Re: They're not cheap.

[doorbot.com]

They're not cheap.

Since when has new (let alone the latest and greatest) Sun hardware been "cheap?" Sure, there might be some good values but it's never cheap.

And then there's the product itself. It's huge, it's fast, it's not intended for the home user, or even a medium-sized business. Let's just look at the specs, pulled from Sun's info page:

• Up to 106 UltraSPARC[tm] III Cu 900-MHz processors.
• Big memory - more than 1/2 TB.
• 
  Up to 18 fifth-generation Dynamic System Domains, which are fully configurable while applications are running.
• Hot-swappable Uniboard design CPU/memory boards that are common across Sun Fire server family.
• Redundant, high-performance Sun[tm] Fireplane Interconnect with up to 172.8 GBps peak bandwidth.
• Full redundancy of power and cooling systems.

Are you trying to be funny by saying it's not cheap? Was that a +5 Insightful comment that we all would have missed out on had you not enlightened us? Are you implying that there are other manufacturers who do sell these kinds of systems for cheap?

In your defense, you do have an astonishing command of the perfectly obvious.

Starcat and Oracle ID card

[mcplusplus]

Starcat sounds like a perfect companion for Ellison's ID card. With them the government can adequately track the behavior of everyone who isn't a terrorist.

I can't build one...

[tshak]

Configuration Error

1) Error calling config servlet: sunir.webdesk.common.checker.ConfigInternalExcepti on: Couldn't get sql connection

Hmmm... looks like their DB server is down.

Re:I can't build one...

[Anonymous Coward]

Those are some great servers, huh... we slashdotted Sun...

Re:I can't build one...

[Derkec]

We may want to forgive Sun for being a bit slow in getting their DB server back up and pretty. A huge chuck of their support staff is helping bale out clients whose data centers got blown up out east. On the other hand, it's your brand new product, you gotta make sure it's available to be bought up. But then again, who in their right mind would just go online and buy one of these? I'd bet ALL of these are sold through meetings between sales reps and IT purchasers.

Re:I can't build one...

[jaga~]

imagine getting commission on one of these? *whistles*

Re:I can't build one...

[tshak]

Very interesting. I posted the error because I was doing a test. Would everyone complain about JDBC or go off on some other nonfactual zealotry, or will they defend Sun. Now let's contrast this to a similar situation with MS :-).

And yet Steve Jobs will demonstrate

[Anonymous Coward]

at the next MacExpo that the Macintosh LCII can operate TWICE as fast.

This is excellent!

[Anonymous Coward]

Solaris 8 + UltraSparcIII with 106 CPUs and 512 GB of RAM makes ANYTHING from HP and IBM look like expensive junk.

Time to unleash the underlying power of the Sun platform....64 CPUs just isn't enough =)

Re:This is excellent!

[Anonymous Coward]

makes ANYTHING from HP and IBM look like expensive junk

...until next week when IBM launches the POWER4 and the Regatta.

That's great, but...

[Wakko Warner]

... can it run Outlook Express?

- A.P.

Re:That's great, but...

[DNS-and-BIND]

microsoft.com/unix...it's there, but don't expect anything close to stability.

Re:That's great, but...

[DNS-and-BIND]

Hey, slashdot messed up my URL...it should be microsoft.com/unix...for some reason it put in a link to slashdot itself...more slashcode wierdness, go figure.

Wonder what this would've done for Final Fantasy?

[weslocke]

I wonder how this would compare with the sheer amount of hardware that was thrown at the rendering of that movie. (I can't find the original link, but over and above the SGI stations that were used there were hundreds of clustered 'normal' PCs)

Just thinking "could they've just slapped a few of these suckers into place instead?"

Outlook Express

[Anonymous Coward]

Kiddiez can't get into your computer without it.

Where does this end ?

[selderrr]

holy smoke ! I don't know about your armada of assimilated dudes, but this blows my socks off

that's one hell of an impressive system... And all in all not that expensive, if you consider that our local university bought an IBM SP2-32procs about 3 years ago for the nearly same price tag... the starcat runs laps around the SP while knitting a sweater for half my country...

Many geeks (like me), closely following the Athlon-P4-G4 epic battle, have the impression that
a) desktop processors are getting damd fast, close to enterprise server level
b) sun/ibm/sgi CPU's not really advancing substantially compared to desktop.

Well, guess we're wrong. How far can this stuff go ? What's the AI power of a machine like this ?

Re:Where does this end ?

[Anonymous Coward]

Take a look at the SGI O3800 if you want your socks blown off for real. And then consider that the O3800 has been available for a while now.

Sun is still playing catchup. It doesn't really matter though because even in the high-end unix mainframe market, marketing means far more than technical ability.

Re:Where does this end ?

[platypus]

Take a look at this baby, ibm's new power4. They are putting two 1GHz-cores on one proc. My uninformed opinion is that apple should have gone with IBM instead of motorola.
Btw. Microprocessor Report is a cool, professional magazine and if you browse around their site you'll get a good overview of what's out there in the world of cpu's besides P IVs and thunderbirds.

Re:Wonder what this would've done for Final Fantas

[Anonymous Coward]

It would never have been as price effective if they had.

Brought TCO WAAAY down

[catseye_95051]

The good thing about Intel based render farms is that theya re cheap. The bad thing is... theya re cheap.

Intel based hardware does NOT have great MTBF (Mean Time Between Faiulre) unfortunately but inevitably as Intel boxes are commodities built for least possible expense. As a result large Intel farms mean near constant maintainence. Soem machine is always on the verge of failure.

In contrast so much horse power in a highly reliable box means both fewer machiens to fai lAND much lower MTBF per machine.

The result is much lower Total Cost of Ownership.

Re:Brought TCO WAAAY down

[Anonymous Coward]

Actually, the cheap, unreliable Intel boxes will have far lower TCO if you set it up right.

When one of them fails, we throw it away and replace it with another $1000 server. We put a CD in the drive, and leave. When we come back, its rejoined the cluster.

Far cheaper than the time spent waiting for the Sun tech who comes, and then bends the fucking pins on a CPU trying to install it.

Re:Brought TCO WAAAY down

[Doctor_D]

Far cheaper than the time spent waiting for the Sun tech who comes, and then bends the fucking pins on a CPU trying to install it.

Humm, on all UltraSPARC's that I've seen it's either a slot or pressure fit contacts, hence no pins to bend.

Re:Brought TCO WAAAY down

[elmegil]

Far cheaper than the time spent waiting for the Sun tech who comes, and then bends the fucking pins on a CPU trying to install it.

Let's see...

Sparc 1 box, CPU pretty much part of the MB, unless you get one of those fancy Weitek things. CPU fails, tech replaces MB, no pins to bend.

Sparc 2, Sparc IPX, etc, same story.

Sparc 20, suddenly we have CPUs with pins on them. Coulda happened. Of course, that hasn't been current tech for several years.

UltraSparc line comes out...pressure fittings for the Enterprise servers, no pins to bend. Deskside UltraSparc (like E250, E450), no pins to bend, the CPU is on a card just like Intel does these days.

Ultrasparc III line comes out, big servers don't even use pressure fittings--if you lose a CPU, you get a new system board. Deskside US III (SunBlade 1000) uses a card similar to older deskside units, and has rails to line it up and a torque tool to seat it. Don't see too many bent pins there.

So, apparently you got burned once a very long time ago with a Sparc 20. Don't you think it's time to get over it?

Help me out here

[donglekey]

Imagine ...

Re:Help me out here

[donglekey]

a Beowulf ...

Re:Help me out here

[donglekey]

Cluster ...

Re:Help me out here

[aidoneus]

...of these things!

(this garbage included to bypass the lameness filter. blargh.)

Re:Help me out here

[jaga~]

someone not invoking this old and tired phrase!

Re:Wonder what this would've done for Final Fantas

[catseye_95051]

Actually the reverse is probably true. The cost of running a compouter site is in the upkeep not the initial hardware cost. See m yTCo explainatio nabove.

They need to be using one themselves..

[ldopa1]

It appears that sun has been successfully /.'ed... I went to the pricing page, and got:

Configuration Error
1)
Error calling config servlet:
sunir.webdesk.common.checker.ConfigInternalExcepti on: Couldn't get sql connection.

When I finally did get the price ($4+ Million), I realized that the page also says "the Sun Fire[tm] 15K server helps redefine total cost of ownership in data center environments."

Talk about an understatement!

Blurr the lines...

[the_2nd_coming]

this new server realy blurs the lines between server and mainframe. it can partition its hardware and has hotswap hardware with hot partition resizing.

so where is Sun going with this? into the mainframe market or are they making it enough like a mainframe that they hope to steal IBM market share. I think the latter.

Why do you call it StarCat?

[Snaller]

That name isn't mentioned on Sun's page - or does their search engine show such a product?

Re:Why do you call it StarCat?

[Doctor_D]

That name isn't mentioned on Sun's page - or does their search engine show such a product?

Pretty simple, StarCat was it's codename in development. Just like Serengti was for the E3800, E48x0 and E6800.

Re:Why do you call it StarCat?

[Anonymous Coward]

Of course, if it was an open source system, it would be called the "KzSeries 900".

Is it just me

[evanbd]

or did we just Slashdot Sun?

1) Error calling config servlet: sunir.webdesk.common.webconfig.WebConfigException: Server busy.. please try again.

wow...

Re:Is it just me

[Anonymous Coward]

wanna know who *really* needs this beast?
http://lw4fd.law4.hotmail.msn.com/toobusy.html

Can't even try to order online

[xrayspx]

Cause we broke the java tool that tells you how much they are. I hope they aren't running store.sun.com on a 15000, after claiming essentially that it's unslashdottable.

Cookies won't work

[Anonymous Coward]

He wants cheese

too obvious

[Pinball Wizard]

you couldn't do kernel-level programming with VB until version 5, which was released in 1997. Before that, it was solely an interpreted, not compiled language.

gcc is not up to version 3.1. yet.

nice attempt though. taunting the linux users while proclaming yourself to be a 1337 VB hacker is always a good way to go.

Re:too obvious

[El]

Isn't "elite VB hacker" an oxymoron?

106 procs, so what

[jshawley]

I'm not sure what is so exciting about a system with 106 processors. When the SGI Origin 3000 can scale to 1024 with a single image of the OS running on it, now that is impressive. Maybe everyone should check out http://www.sgi.com/origin/3000/3800.html Though the site states that it only goes to 512, there is now an official installed system running 1024 that you can see at www.sgi.com/streaming/products.html#sara Now that is IMPRESSIVE!

Re:106 procs, so what

[segfaultcoredump]

The SGI origin has a ccNUMA architecture, which makes it great for some tasks, ok for others, and awful for yet others. (the trick is to make sure that your particular app falls under the 'great' category)

The sun system is an smp based system, everything connects to a common backplane and each board has equal access to all of the other boards. With the sgi, the speed of accessing memory on the local board or boards in the same cabinet is much faster than hits to memory in remote cabinets.

From what I can tell, Sun is planing on producing a special system board that goes into one of those 18 slots. Thus, with 19 StarCats you can create one big system with 1836 cpu's and 9.7TB of ram. (think of a system in the middle that acts as the center of a star) it will most likely be based on a COMA architecture rather than a ccNUMA. Like the SGI, memory access will depend on the distance between the requesting cpu and the storage location. The difference is that under COMA, if a cpu requests a particular bit of memory a lot, that page is either migrated or copied to a memory bank on that cpu's memory board (so if 5 cpu's all need read only access to the same bit of memory, then they can each have their own copy in a local memory bank. write updates are what make the system a pain in the ass to manage ).

Re:106 procs, so what

[pmz]

The UltraSPARC III architecture is designed to scale to many hundreds of processors in an SMP configuration. The 106 processor limit in the Sun Fire machine is just the latest offering from Sun. Sun also provides software to support 1024 processors in a traditional cluster of machines, so it would be possible to cluster nine or ten of the Sun Fire 15K machines (as long as I don't get the bill).

I belive the SGI system has a NUMA architecture, which has some "cons" relative to an SMP architecture, such as potentially high memory latency. It's mainly a tradeoff for being able to get 1024 processors into a single system.

If Sun releases a 512+ CPU SMP machine, then the Origin system will have some real competition in this ultra-high-end arena.

Re:106 procs, so what

[swordgeek]

Just wait for six months. This is the first beast in a series of pseudo-clustered Sunfires. This is roughly a stack of 6800's, and there's going to be a MUCH larger machine released very soon.

Re:106 procs, so what

[peril]

As you said, the idea actually is to cluster the backplanes and replicate memory state across the boxes relative to an executing user process(Single system image). The seems like an N^2 instance, as you have N processes accessing N memory locations, across a shared backplane. (assuming a nice uniform process/kernel thread density)

I dunno enough about the hardware architecture of memory to understand the bandwidth/latency concerns of replication across a shared backplane, but the first thing to come down will be the single system image. (Which will require kernel caged stuff to be replicated, accessed in a sane manner)

Shouldn't be too bad of an extrapolation to get user space library/code sharing working after a kernel strategy is defined.

--Adrian

Re:106 procs, so what

[brer_rabbit]

A stack of 6800's? Do they have eight chips in a cluster to make a 64 bit machine? Perhaps they'll move up to the MC680x0 line soon -- that would show the Macintosh SE who's boss!

Re:106 procs, so what

[swordgeek]

Hah! That's the first thing that's made me smile today (with the possible exception of the image of a certain contractor's head getting run over by a truck).

Sorry, I wasn't clear. SunFire 6800's. Loads up to six system boards with 4 processors and 1GB/proc. each, at present. (plus four I/O boards)

Re:106 procs, so what

[mt-biker]

With the sgi, the speed of accessing memory on the local board or boards in the same cabinet is much faster than hits to memory in remote cabinets.
...
The difference is that under COMA, if a cpu requests a particular bit of memory a lot, that page is either migrated or copied to a memory bank on that cpu's memory board

No difference there. SGI also do page replication and page migration.

Also, "one big system with 1836 cpu's and 9.7TB of ram"? On a monolithic kernel that has only just for the first time run on more than 64 CPUs? Sorry, I just can't see this getting off the ground.

The Travelling Salesman Problem?

[swordboy]

The Travelling Salesman Problem is very computationally intensive. Just picture hundreds (if not thousands) of destinations. Getting everyone to their destination in the most efficient manner is VERY computationally intensive. Take a look at the algorithms in the link above.

Can someone tell me...

[mickeyreznor]

... if this is the reason that I can't seem to connect to any sun sites?

Fujitsu Primepower already at 128

[Anonymous Coward]

The Fujitsu Primepower has been at 128 processors for quite a few months now. It has 563 MHz Sparc64 processors on 32 system boards. Each sysboard can have 6 (3x33MHz, 3x66MHz) PCI cards.
This system blows away the sun box on I/O here.

http://primepower.fujitsu.com/en/m2000.html

SGI? Heh.

[Anonymous Coward]

You own SGI stock? Wow, man, I'm sorry.... How much did you lose?

It's really funny how well Cray has done and how poorly SGI has done since SGI dumped them. SGI will probably get bought out one of these days, they're really struggling.

Fantastic!

[Anonymous Coward]

This new consolidated Sun box will make my job of removing Sun boxes and installing IBM zSeries machines a LOT easier! Much less bulk iron to recycle!

Where would you plug one in?

[orionpi]

200-240 single phase VAC, 47-63 Hz, with six 30 Amp circuits redundant with another six on 2 separate power grids

Can you get it bundled with 12 30A 240V 100meter extention cords? (So you could plug it directly in to every panel on the block.

Re:Where would you plug one in?

[stu_coates]

I can assure you that these guys are really thirsty... I used the 15000's smaller brother, the SunFire 4800 at Sun's labs in the UK earlier this year (prior to the launch), and with 5 of them running at the same time we managed to trip the power breakers quite often.

The SunFire machines themselves are great performers when it comes to running large databases. I did find the disk I/O a little dissapointing, but then again we where only using a very basic disk setup.

Only 106..?!?

[Anonymous Coward]

I wouldn't even brag about 106 processors when other vendors are doing much better. Take a look at SGI's 1024 processor machines. Their next generation CPU will allow up to 64K processors! (65536 CPU!)

Now that's cool....

SGI Origin 3000

[Refried Beans]

If you want a real system that scales and doesn't lose IO capability, check out an SGI O3k. up to 512 CPUs, as much IO you want.

http://www.sgi.com/origin/3000/

Re:SGI Origin 3000

[Anonymous Coward]

As much IO as you want??? What are you smoking. The SGI is built with the exact same back plane a the sun was 2 gens back. So no the SGI pos is not repeat NOT faster than the SUN in IO.

Re:SGI Origin 3000

[fgodfrey]

What are *you* smoking? The Origin 3800 is certainly *not* using the "same back plane a the sun was 2 gens back". In fact, if you were anything other than a hopeless troll, you'd realize that the Origin 3800 doesn't *use* a backplane at all. You get, on a 512p system, 128 I/O channels, each of them supporting up to 12 PCI slots or 4 XIO slots. I can't remember off the top of my head what the bandwidth per channel is but it's on the order of a gigabyte/second (I wanna say it's 1.6 GB/sec, but I might be wrong).

Simplex Method

[heliocentric]

But you don't need to do every problem with a graph. Look into operations research. What you do there is boil down all the limits.

Example, to fly this plane means R fuel cost, S crew cost, T maintenance, U airport costs, V etc... Now put in this information for all possible flights, and all the passenger demand info (you'd want regular flight days, as well as holidays, etc..) and put all of this into a gigantic equations where you are trying to maximize or minimize something - maximize profits, or maybe minimize certain costs. Then you can relate this to a rather large matrix. Now so far we haven't asked for any info you wouldn't need for your graph. Then you apply one of the methods from operations research to your data set and you get either a number, or maybe a bound range of an answer - without the contruction of the graph, or all of the graph traversals...

Yes, if you want to do something on-the-fly, like have a computer fly the planes for you, and update as weather changes then you'd probably want to think about a graph. But to generate the day-to-day schedules for your planes, look into OR.

Here's the textbook I reccoment to learn Operations Research: here

There are regular business apps for this

[vinyl1]

At my giant financial institution, we currently use an E10000 partition to do our monthly departmental MIS. It has 28 CPUs and 28 gigs. Our job does trillions of calculations, creates about a billion records, and loads them into two Oracle tables (only one index, thankfully).

It currently runs about six hours, but our merger with another giant financial institution will result in the creation of even more records, and our users always have big ideas that always seem to lead to more records and more calculations. Already, were up to 60000 cost centers, with 20000 accounts over 48 months of data (plus year to date, quarterly, last year to date, quarter to date, last year quarter to date, etc, etc). There are many complaints already that our monthly run takes too long, and with all the additional processing that will be required we're looking at 10 hours without more hardware.

I should mention for all you geeks out there that our coding is about as efficient as you can get. It's custom C++, full of forking and threading, with all the data read into memory at the begining and processed there. When we run, the CPU goes close to 100% and stays there for hours. Because of hardware limitations, we have to limit it to 22 processes at a time using about two and half gigs of memory each. If we could get more parallelism, we might be able to squeeze more processing into our available window.

hmmm

[Anonymous Coward]

Id think it takes whole lot of disk space oppposed to ram or computing power. 106 cpus will be good for simulations i suppose

Re:hmmm

[Jon_E]

um - both actually - how do you anticipate accessing or searching 300M records?

I think this is why it's being sold along with the HDS OEMd SE9960 - racks and racks of T3s make for a management nightmare and EMC doesn't look so good these days .. i like sun - they're not that reactionary and seem to think ahead - granted they missed a few key things (ie the storage market, dotCom buzzword fallout), but overall they've got a pretty take on being innovation driven ..

Nooo!

[7-Vodka]

Damit! I thought the headline said "sun releases starcraft"!

Re:Nooo!

[Anonymous Coward]

Get yourself a SUN Blade 100 ($1,000). then get yourself a SUN PCI Cards ($5000) install StarCraft on the sunpci card. StarCraft works quite well the Sun PCI cards. ANd it is fun too.

"Not cheap", he says.

[sharkey]

Man, you'd have to be Bill Gates to afford one of these suckers!

�Yay!

[vought]

en Suncat es yo quiero performance.

Quando Microsoft?

Up to 848 megs of cache.

[jellomizer]

106 processors with 8 megs of external cache. Think of it. XWINDOWS running compleatly in cache. Although I dought the archecture will allow XWindows to use the cache in all 106 processors. But just the idea makes me drool. Or just the ability to say. Oh I use my RAM as a place to swap memory for large programs.

Re:Up to 848 megs of cache.

[T-Punkt]

WTF is XWINDOWS or XWindows?

Sorry, I'm a UNIX guy, I don't know all this windows stuff, all I have is the X Window System...

Re:Up to 848 megs of cache.

[darkonc]

But just the idea makes me drool. Or just the ability to say. Oh I use my RAM as a place to swap memory for large programs.

I actually ran into this as a problem about 10 years ago. In 1992, we bought a machine from IBM that we nicknamed 'brutus'. The idea behind brutus was pretty simple: we bought the machine, and with just about every dollar we had left for the project we bought memory.

Brutus ended up with 380 Meg of ram (this was back in 1992, when most people were really happy to have 8 meg of ram and a 200 meg hard disk). It also had a single 800meg hard disk. This is where the trouble began. After loading AIX and all of the upgrades, etc, we had about 100 meg left for swap space. Unfortunately, AIX needs swap space for backing store before it will allocate memory, so we had a mondo-expensive box with about 200meg of ram that we couldn't get to.

We cobbled together enough money for a second 800meg drive, and while we were waiting (months) for it to be delivered, our IBM rep assured us that the extra memory wasn't completely wasted.

"The system will use it as a disk cache."

I pulled the extra memory boards and swapped them into a bunch of other (smaller) boxes while we waited for the extra backing store.

You may wonder what we wanted 380 meg of RAM for back then... It was a graphics lab, and some of the grad students were doing research on volume visualization. 380 meg allowed you to play with a 512x512x512 cube in ram (8 bits per voxel, 2 copies).

Re:Up to 848 megs of cache.

[elmegil]

Lucky for newer customers, Solaris hasn't required disk for every byte of backing store in a long long time.

Insecure

[Syberghost]

I thought we weren't supposed to use strcat anymore, because it's subject to buffer overflows?

Re:Insecure

[JLouder]

I thought we weren't supposed to use strcat anymore, because it's subject to buffer overflows?

You're right, a strnfire is a much better system.

I thought it said Starcraft =).

[antdude]

My eyes playing tricks :).

strcat?

[J'raxis]

We all know strncat() is better.

Re:strcat?

[Anonymous Coward]

I'm sure you'll find strlcat() is better.

"Descent" Performance?

[matthewg]

Is that performance so good, it makes you dizzy?

Wow! The Zerg will...

[Bigger R]

..REALLY be fast on that puppy!

Damnit!

[HellKnite]

We just bought two of the 6800's at the company I work for. They're sweet. But now they go and release the 15k's on us! Damnit! Here I am stuck with only 24 processors and 4 domains... *pout* oh well, I guess I'll survive... the pair of the 6800's make a nice set of book-ends.

Re:Clarification please (*real* HA)

[darkonc]

If you really want HA -- no problem. For an extra $4mil, I'll send you a second box, and throw in a couple of engineers to make sure it all goes together smoothly....

Given how much you can shuck and trash without powering down the whole machine, a second box is going to be for the people who really need the 'five nines' availability. Other than an extreme power failure or WTC-style disaster, it's hard to think of a situation that would require shutting down all 6 power supplies at the same time.

Obligatory stupid comment #2

[El]

Yes, but does it run Quake?

Will it improve my game any?

Re:Obligatory stupid comment #2

[Anonymous Coward]

Yes !! - http://www.planetquake.com/eclipse/

Still small CPUs

[macemoneta]

While Sun has increased the number of CPUs, they still don't have a good solution to address the large monolithic type of workload. IBM and others have them there, with very large CPUs available on the S/390 series, for example. I've had difficulty finding comparison benchmarks between the UltraSparc III 900 and Intel's high-end Xeon CPUs. Does anyone know how they stack up against each other?

Re:Still small CPUs

[elmegil]

What do you mean by small? Small MHz? Considering that SPARC is a RISC architecture, you can't really compare them apples to apples. It depends a lot on your workload. Sure I can run some or even many workloads faster though a Xeon, but I can't put more then 4 or maybe 8 of them into a machine very effectively.

If you're still running monolitic apps, maybe you need to redesign the app to take advantage of the scalability of threading....

Re:Still small CPUs

[Anonymous Coward]

Monolithic apps come from monolithic programmers.

Actually, your statement makes very little sense. 72 processors, Dynamicaly reconfigurable into separate domains which will run another kernel combined with Inner domain networking at buss speeds doesn't sound very monolithic to me.

Re:Still small CPUs

[Anonymous Coward]

They were passed by the xeon about 1 year ago. Intel is not looking back - sun is toast. Intel is going to have a 16 way chipset mainstream soon, I can already buy a 32 way Unisys, and will be able to a 96 way machine in less than a year that will smoke this puppy.

Re:Still small CPUs

[macemoneta]

Threading isn't always scalable; if the inter-thread communication that would result from threading an application is disproportionately high, then the efficiency of the application is maximized by minimizing threading. There are many such commercial/scientific applications. The solution to these problems is larger CPUs, not more small ones. That's the reason mainframes still exist; there is a class of problem that they are very efficient at solving :-)

Re:Still small CPUs

[Anonymous Coward]

Yes, IBM mainframe CMOS CPUs are roughly twice as powerful as UltraSPARC III CPUs given similar clock rates.

But the reality is almost every commercial application has been ported to multithreaded C and C++ code.

Mainframes are good at processing single-threaded COBOL code, but not running Oracle or DB2. For that, get a good RISC/UNIX machine.

Hell, even Java now scales to 72 CPUs.

Buy online

[Pacorro]

"Click here to buy online"

Who the heck buys this stuff online ???!?!?!

"Mhmm I think I'll charge this one to my Discover card"

Re:Buy online

[Anonymous Coward]

Well... I do have my Sun Corporate Amex... I wonder if I could expense one of those for my personal workstation?

A drooling Sun employee...

Insecure?

[Denor]

Sure, the starcat is great and all, but don't we want to be using starncat instead?

Wonder if....

[11 platter hard driv]

Wonder if I could run my windows on that.... They have this new product called windows 2002... oops, I mean XP

Super-Size that please?

[Peridriga]

SCHZHCHC.. Thank you for choosing Sun... Can I take your order

Umm... Lemme see... It's not breakfast anymore is it?

No sir....

Hmmm... Can I have one of those Sun 15k servers?

Ok... 1 Sun 15k server... That will be $1,808,110.00.... Would you like to super-size that order?

Yeah sure... Why not..

Aight... That will come to $4,140,830.00

Would you like to add on the extra disk array for only $480,400.00

No thats Ok

Ok... With tax your total comes to $4,430,688.10

Please drive around to the next window

Scheduling can be NP

[xixax]

I worked with a guy who was using genetic algorithms to schedule harvesting, and it really quickly racks up an insane number of permutations.

Simply put, with 10 airports and 10 airplanes, our first step has 100 possible _initial_ (opening) moves, and they permutations mushroom really quickly after that. He ended up using a small Beowulf that always had the current real world state, and this way he could shuffle scheduling on a daily basis, adapting to changes.

In comparision, our local bus network is re-scheduling, it's going to take one guru a couple of months to "zen" a new schedule.

Xix.

Re:Clarification please (*real* HA)

[_damnit_]

Anyone really concerned about that many 9's on a service would have boxes in different locations. (ie. two/three 15k's running Oracle Parallel Server with dedicated t1/t3's connecting them from different states/countries, etc). That's damn reliable. Plus, with the Dynamic reconfiguration makes it easy to replace cpu or i/o boards without bringing the whole machine down.

Where 106 probably comes from

[MadDog Bob-2]

I mean to say what's the difference between 106 (what an odd number) ...

Given the way other Sun boxen like the E3500 work, I expect that's the 15K has 18 boards, each of which takes 3 modules, either 2xCPU of 8GB RAM.

That means that 72 CPU / 288 GB memory is 18 boards, each with 2 2xCPU modules and one 18 GB memory module, and the box is full.

Since you always need some memory, the most CPUs you can get is 17 boards w/ 6 each and one with 4. Of course, that leaves you with 8GB of memory for your 106 CPUs.

The other end is (17 x 3 + 1 x 2) 8 GB memory modules for 424 GB on a pair of CPUs

But that's just a guess...

TCO

[TheMightyZog]

From the site:

With 106 UltraSPARC III Cu 900-MHz processors, more than half a terabyte of memory, and fifth-generation Dynamic System Domains, the Sun Fire 15k server helps redefine total cost of ownership in data center environments

Yeah, no kidding...

SGI can do 1024 processors :-P

[halfelven]

You think 100 processors are a lot? Take a look at SGI3000 which can come with 1024 processors at any time. Now that's a lot! ;-)

They say it themselves!

[kimihia]

Have you checked availabilitY? It is listed as "Constrained" which means:

Demand for this product is so strong that it exceeds availability and is subject to shipment delay. If you proceed with your order, Sun will send you an e-mail with the scheduled shipment date as soon as possible.

MMMM it's beowulfalicous....

[Rhinobird]

mmmm... it's like walking around in germany and running across some gold coated chocolate, and realising that the MRE chocolate bars you're used to just aren't gonna cut it anymore...

WOW

[Anonymous Coward]

Imagine, if these CPU's are bug free and my kernel has no panic's and the link to SETI stays up and.....

beast?

[porky_pig_jr]

what'd ya mean 'beast'? It is 560 processors short of being one.

Slow but reliable

[linuxbert]

most ATC systems run on really slow equipment.

most AFTN Mesassages (fliight plan, met data etc) fly between airports as 2400baud. some of the faster ones use 28,8 modems, or X25

the computers are often pentium class machines (running slack 4.something) or stratus servers (hpux ones)

the point is speed is not nessasary, ensureing the message gets there reliably before the aircraft does is nessasary.

Well, it's fancy and all , but

[Anonymous Coward]

will it run pine?

Here is a quiz ! ! ! :)

[Anonymous Coward]

What is the second OS that these systems run in parrallel to Solaris.....?

( NO ITS NOT LINUX )

Re:Here is a quiz ! ! ! :)

[Anonymous Coward]

You would think it would be the javaos they worked on a couple years back ....it was suppose to revolutionize embedded hardware.

the use of this type of system

[Ciannait]

Yes, there's a buy online button. But that's used to get info so one of their sales droids can contact you. It's not like you can slap it on your Visa card. :)

(Disclaimer, I work a lot with E10Ks, so this post is written mostly from my experience with those.)
The 15K is basically just an improvement on the E10K architecture, from what I've seen and heard from Sun's SSEs. The E10K started out life as the Cray SuperServer, and was sold to Sun for a song. It's not architecturally perfect. The E10K is set up to allow individual system boards to be part of domains (aka partitions), which can make for some great scalability in the domains. I've seen tiny little one-system-board domains, and domains with 13 fully populated system boards in them.

One of the major advantages to this platform is the fact that you can hot-swap everything except the centerplane. (Of course, I've never seen a centerplane fail.) The E10K also has Dynamic Reconfiguration, where you can remove system boards from a running domain, but unless your platform is set up in a certain, specific way, this doesn't work as well as advertised. I've personally never used it. The best thing about the E10K is the use of the System Service Processor, which handles all the administrative tasks for the entire cabinet. I've heard that the SSP is now integrated into the 15K, thus eliminating the need for a separate system to perform these tasks and monitoring.

The only thing I've ever seen this class of system used for is data warehousing. No modeling, no graphics rendering, just Oracle databases. Just because it has a large number of processors, doesn't mean they're going to be suitable for every task imaginable. (I used to have a 180MHz Indy R5000, that got 68kkeys/sec in d.net. My 166MMX got something like 350kkeys/sec.) These are workhorse processors, not sports-car style processors.

Though I wonder if Sun's gotten around to fixing that nasty ecache parity error problem with their processors... Having a domain randomly crash because the parity bit on a processor got flipped is no fun when you're dealing with a large production database. I have a feeling that problem will continue to plague them in the 15K.

Re:the use of this type of system

[Anonymous Coward]

Working in the HPC market space, these servers are wonderful for our work. There is nothing like seeing an E10k and soon an 15k fully pegged running financial models.

Re:the use of this type of system

[Doctor_D]

The E10K also has Dynamic Reconfiguration, where you can remove system boards from a running domain, but unless your platform is set up in a certain, specific way, this doesn't work as well as advertised. I've personally never used it.

Well actually the Ex500's also have DR, and of course the newer Ex800's. DR is a great idea, works great in lab environment. Honestly it's cool to see the running kernel jump from one system board to the other. But it's a product that I have never seen in the wild.

Though I wonder if Sun's gotten around to fixing that nasty ecache parity error problem with their processors... Having a domain randomly crash because the parity bit on a processor got flipped is no fun when you're dealing with a large production database. I have a feeling that problem will continue to plague them in the 15K.

Remember the 15k uses UltraSPARC III's rather than the UltraSPARC II's. And the newer UltraSPARC II's have mirrored ecache to take care of that problem. I do believe the UltraSPARC III's also have that same set up. But when the chips are down, I'd rather have the box crash to maintain data intregrity rather than spew garbage (can we say the floating point bug in the Intel Pentium?). But I think you summed it up best by saying "These are workhorse processors, not sports-car style processors."

Re:the use of this type of system

[dunstan]

Yes, Sun did fix the ecache parity error in ultraSparc III, and have produced UltrasparcII CPU modules with mirrored ecache, drastically reducing the likelihood of them happening.

It certainly won't affect the E15K, and many E10K customers have had their CPU modules swapped for the mirrored cache ones (and it really *does* stop the panics). Similarly the 3000 - 6500.

Re:the use of this type of system

[sql*kitten]

It's not like you can slap it on your Visa card

Speak for yourself :0)

data warehousing

[KyleCordes]

[The only thing I've ever seen this class of system used for is data warehousing ... Oracle databases]

I wonder what the relative merits are of:

1) A massive box, with general purpose (but very, very good) database software

2) A less massive box, with some kind of OLAP-specific database software

... or is Oracle also OLAP'ed enough to be in category 2?

Re:data warehousing

[Ciannait]

Honestly, you got me. I don't deal with Oracle at all, just the systems themselves. (I suppose I should learn about Oracle, but that'd require that pesky training that there never seems to be a budget for. :)

cooking the numbers

[spiffy_guy]

Here is another case of sun cooking the numbers. When they say 106 processors they mean 72. Why? Because in order to get 106 you have to use up all the PCI slots, and running 2 processor cards through PCI slots is not fast. That's why all the benchmarks they give quote the 72 processor system.

Speaking of the benchmarks. They quote two benchmarks. They say their java is faster than IBMs best reported score. That's really easy to explain, IBM hasn't reported a score for their fastest machines on Java. IBM has published results for its antiquated big machines and little 8 way boxes. I'm glad to know sun beats an 8 way with a 72 way.

Then that fluent benchmark faster than an IBM 128 way 1 Ghz... The only 1Ghz 128 way IBM makes is an Intel cluster. So they found an application that likes SMP better than a cluster. Maybe they should compare to IBMs big SMP machines to their big SMP machines.

Then there are no TPC benchmarks of any kind. Do you really think they would not publish the results if they were faster than everybody elses? The reason there is a lack of TPC results is not that they think TPC isn't a valid benchmark or that they don't have the results. They simply didn't want to embarrass themselves with the pathetic real world performance of their machine.

I'm not trying to down Sun. This is a decent server. However Sun should give a stern talking to its marketing guys for being so misleading.

Need an admin?

[hubertf]

8+ years of Solaris admin experience available.
Contact hubert@feyrer.de.

Re:Clarification please (*real* HA)

[elmegil]

Is OPS really capable of dealing with the latencies of a huge distance between nodes? I really doubt it....

Re:Clarification please (*real* HA)

[_damnit_]

Yes, I have seen it set up and working.