Slashdot Mirror

Except that by the time Sun "invented" the Blackbox, Google had already secretly built an entire data center filled with data center containers. And applied for the patent on the concept, which it got.

The concept HP described in an academic paper is not new, and is already being built into a UK data center that has won awards for its approach. The Infinity ONE data center in East Anglia, England is working with a local farming co-op to use cow manure to power a combined heat and power (CHP) plant that will provide electricity for the data halls. The data center is already operational (using utility power) and expects to integrate the poop power sometime this year.

One trend that makes the idea of turning organic waste into usable power for data centers is the moves by several firms to build facilities in rural locations, where high-speed networks allow them to take advantage of the cost advantages of such areas.

An example of this trend is the world's largest, the Lakeside Technology Center (sounds rural enough) located ... in downtown Chicago.
http://www.datacenterknowledge.com/special-report-the-worlds-largest-data-centers/worlds-largest-data-center-350-e-cermak/

What on earth leads you to suggest they don't have working disaster recovery?

The fact that their service was partially cut due to a power failure. We know accidents DO happen and power failures DO happen, like the explosion in The Planet's power control room [1]. The guy's at Amazon cloud should be prepared for predictable problems like a power outage, specially when one of their selling arguments is service continuity.

[1]

In its response, ASHRAE says there are already alternatives to economizers. Google and the other signers remain concerned about how the amendment to standard 90.1 might be interpreted by local building officials. They say tthe emergence of fresh air cooling is itself an indication of the speed with which best practices can change. “Not so long ago, economizers were a taboo subject for data center operators,” said Google's Chris Malone.It wasn't really until this 2008 Intel study that economizers gained momentum in data center design.

Believe it or not, trading by computer has now become so quick that companies vie for server space closer and closer to the place where trades are taking place. This article" even goes as far as suggesting that a 1 millisecond advantage in trading applications can be worth $100 million a year to a major brokerage firm with such low-latency trading becoming more common.

I think the ping to Asia may be a bit more than that!

Agreed. If you have the second most popular website in the world serving over 400 billion dynamically generated page views per month from a backend of over 30,000 servers, you absolutely do not want to rely on PHP to serve those pages for you.

This is the same old argument over and over again. Facebook and a couple of other large sites building on PHP don't work the way they do because of PHP, but despite being build on PHP. There's a lot of engineering effort put into those sites, with a not unimportant share of that devoted to things having nothing to do with what PHP handles. With such engineering effort being poured in, Facebook could have been developed in assembly or COBOL, neither of which would have been prove that they are inherently suited for such environments.

Also, don't forget that Facebook is migrating to their C++ based hiphop toolchain and have build their own internal framework, a framework that resembles the Java EE stack in a couple of ways and is NOT available to the general public. The kind of PHP that Facebook does is NOT your typical 'phpfreakz' kind of PHP that is most commonly practiced.

I've also personally talked to the lead developer of Hyves -- the Dutch version of Facebook with some 12 million users -- and he said 'PHP is a little shit language' (PHP is een kut taaltje). I understand that out of context this may not mean much, but he did said that. Additionally, Hyves, Tweakers (a Dutch news site) and a couple of other companies I know that use PHP professionally, some of which I know people working there, almost all seem to agree on that it's very difficult to find really skilled PHP programmers. There is an abundance of programmers that are of the Joe's Burgers variety and have hacked on some PHP script, but that's a completely different category really and not at all what a professional organization is looking for.

So, even though a select few companies are using PHP very professionally, the norm in the PHP market is that it's used by people who barely know the difference between a while loop and an if statement, let alone that they know anything at all about data structures, algorithms, or efficiently structuring their code.

Does PHP offers anything like that? Does the standard PHP library already comes with an MVC framework? Is unicode already natively supported? Is there any name spacing support? Is there ..

Those aren't real problems unless they are actively preventing you from developing your website, which is probably only the case about 2% of the time. The other 98% of the time you never need consider them.

And that's precisely where it all goes wrong. We don't need no education anymore, we don't need no knowledge about CS theory, we don't need no transactions, we don't need no unicode, we don't need no MVC... PHP is so easy... a baby can now program any web site. .... Until it all comes falling down.

It's the typical Visual Basic kind of mindset. A mindset about which you can read everyday on thedailywtf. Sure, idiots can produce crap in any language, but for some reason both PHP and VB seem to produce a disproportionate amount of that.

You sound like you did very well in school, and maybe even graduated at the top of you CS class. You also sound like you may have little real world experience.

I did okay in school and CS went well too, thank you. I'm currently the lead developer of a 12 person strong development team I've worked with for about 7 years. We're building an enterprise application consisting of several 100k LOC and every day have to deal with an infuriating amo

I'm not a particular fan or foe of PHP (it's a tool that can get a certain class of jobs done), but I should point a few things out...

PHP is basically JSP with plain Java SE on it, and a rather poor version of that approach that this. It's perfectly suited for Joe's Burgers 4 page website, but when something more 'advanced' is needed, PHP just doesn't cut it.

Agreed. If you have the second most popular website in the world serving over 400 billion dynamically generated page views per month from a backend of over 30,000 servers, you absolutely do not want to rely on PHP to serve those pages for you.

Does PHP offers anything like that? Does the standard PHP library already comes with an MVC framework? Is unicode already natively supported? Is there any name spacing support? Is there ..

Those aren't real problems unless they are actively preventing you from developing your website, which is probably only the case about 2% of the time. The other 98% of the time you never need consider them.

And I can produce results a hundred times as fast by using Java EE/Java/JBoss AS than with Django/Python and surely PHP. I'm not sure what this proves though... Also, don't forget that hacking together stuff and throwing it out is nice when you're 16 and in high school, but in the real world maintainability, correctness, stability, scalability, etc etc all count and in the end are way more important than just the ability to get some prototype like code out of the door, that 'usually-works-but-breaks-down-every-tuesday-when-it-rains'.

You sound like you did very well in school, and maybe even graduated at the top of you CS class. You also sound like you may have little real world experience. If you're in the bowels of some large corporate IT department where even your boss forgot you existed, you may have the time to plan everything, consider all of the potential problems, and do everything perfectly right the first time through, but it's more likely that either your boss or someone above him will see the cost of you spending six months planning something that will be perfect as way more expensive than just spending three weeks delivering something that is good enough. And this isn't even considering that you may have competition in the marketplace, competition that will put you under if you spend all your time planning while they are delivering actual products.

The YouTube staff now says the downtime was due to a technical error. That really narrows it down.

This makes sense when you consider Google and their relentless pursuit of reducing their energy bill.

A lot of people have no idea how much electricity they are consuming, except at the end of the month. Increasing awareness will encourage people to turn off unused lights in their house (and get the instant gratification of seeing the electricity consumption graph go down on their homepage). This serves a dual purpose. Cutting down on consumption will mean a surplus of electricity, which lowers the price. Google gets cheaper electricity, and it also helps the environment.

I don't think Google is particularly interested in selling your electric power consumption data, although they might want to look at large-scale statistical data for their own research.

That may be true. Although, this whole thing has me wondering if they're re-thinking their "Green Technology" push. According to this article, they've recently partnered with a European firm that specializes in Green Data Center Technology, most specifically, using "air side economization" cooling techniques (cool the data center with outside air as opposed to mechanical cooling). Now, while I think this is a viable and worthwhile technology and strategy (and before anyone flames me into oblivion for being a naysayer when it comes to becoming more energy efficient), I do have to raise the question as to whether they had fully commissioned this facility. Had they determined at what point(s) supplemental cooling was brought online? Had they tested the mechanical cooling systems to ensure they would come online successfully and in enough time? Did anyone ask any questions relating to these issues?

Just makes me wonder. I hope they share the information so that we can all learn from it.

2007: Youtube accounts for 10% of all traffic
2008: Youtube accounts for 35% of all streaming video; Streaming media accounts for 50% of all internet traffic. Doing the math, that puts Youtube around 12-15% of *all* traffic.

There's no good alternative because there's no good alternative.

Unless your site catches fire and the fire department insist you cut the power before they'll enter as happened to ThePlanet.

http://www.datacenterknowledge.com/archives/2008/06/01/explosion-at-the-planet-causes-major-outage/

"The fire department is not allowing the company to run backup generators, so the facility has been without power since the incident occurred."

Adequate fire protection doesn't help that much if the electrical room explodes with enough force to remove three walls.

The source article misses some of the coolest design features of this facility. It has the equivalent of a 12-foot high raised floor, using the entire lower level of the facility as a cooling plenum. The fans bring the cool North Sea air into the lower chamber, and they manage the pressure to direct the air up into the server area. There's also a Computerworld story with more details but an erroneous headline that suggests that it's the "first-ever" wind cooled data center. The story makes it clear that the facility has chillers as backup for when the wind dies down or air temperature doesn't support free cooling. Both Microsoft and Google are already running data centers with no on-site chillers.

The source article misses some of the coolest design features of this facility. It has the equivalent of a 12-foot high raised floor, using the entire lower level of the facility as a cooling plenum. The fans bring the cool North Sea air into the lower chamber, and they manage the pressure to direct the air up into the server area. There's also a Computerworld story with more details but an erroneous headline that suggests that it's the "first-ever" wind cooled data center. The story makes it clear that the facility has chillers as backup for when the wind dies down or air temperature doesn't support free cooling. Both Microsoft and Google are already running data centers with no on-site chillers.

Your tax dollars at work http://www.datacenterknowledge.com/archives/2009/07/01/nsa-plans-16-billion-utah-data-center/, someone has yet to update their wiki entry http://en.wikipedia.org/wiki/National_Security_Agency.

Read all the little bits, people...

2005: Google starts buying up dark fiber.

2007: Google buying land in middle of nowhere, near power stations, building data centers.

Now throw in GoogleTalk, Wave, Phone, Chrome, Android...and now the Nexus-1?

My hypothesis for the next 5 years of Google:

1. Become an ISP, like ComCast, TimeWarner, etc...
2. Become a telco like Qwest, Verizon, AT&T, etc...
3. Become a broadcaster like ABC, NBC, FOX, etc...
4. Become a wireless provider like T-Mobile, Verizon Wireless, etc...

Pay close attention, folks. This is going to happen, and it's going to happen fast.

Jeff Rothschild, the vice president of technology at Facebook, said in a recent presentation that Facebook has 230 engineers supporting data for more than 300 million users. He says Facebook seeks to maintain a ratio of one engineer for 1 million or more users. Facebook is vague about exactly how many servers it has, saying it's "more than 30,000." But at 30,000 servers and 230 engineers, that's a ratio of about 130 servers per admin.

Microsoft says it has automated its data center operations to the point where its admins can each manage between 1,000 and 2,000 servers. That matters, as the company may pack more than 300,000 servers into its new container data center in Chicago. It expects to support that facility with about 30 employees, including admins and facility maintenance staff.

"Number of servers per admin" is to limited to be properly measured, even in your examples. Can one admin manage 5000 *nix boxes that are all identical, running a clustered web server? Sure. Given the proper engineering and tools, it's very possible. Now throw in a bunch of other applications that users request or a business requires and now you will need more staff. Also, supporting a bunch of pizza boxes or supporting one piece of heavy iron can make a difference is trying to measure this number. I know of a couple of folks who admin one box. Granted, they are either a large SUN and IBM hardware, running databases, web servers, and numerous applications, but look at their job requirement in a slanted way and it would look as though they are responsible for only one system. While another admin can have a couple of thousand desktops all identical and simple. And their value would still be relative to whomever the users are.

Jeff Rothschild, the vice president of technology at Facebook, said in a recent presentation that Facebook has 230 engineers supporting data for more than 300 million users. He says Facebook seeks to maintain a ratio of one engineer for 1 million or more users. Facebook is vague about exactly how many servers it has, saying it's "more than 30,000." But at 30,000 servers and 230 engineers, that's a ratio of about 130 servers per admin.

Microsoft says it has automated its data center operations to the point where its admins can each manage between 1,000 and 2,000 servers. That matters, as the company may pack more than 300,000 servers into its new container data center in Chicago. It expects to support that facility with about 30 employees, including admins and facility maintenance staff.

"Number of servers per admin" is to limited to be properly measured, even in your examples. Can one admin manage 5000 *nix boxes that are all identical, running a clustered web server? Sure. Given the proper engineering and tools, it's very possible. Now throw in a bunch of other applications that users request or a business requires and now you will need more staff. Also, supporting a bunch of pizza boxes or supporting one piece of heavy iron can make a difference is trying to measure this number. I know of a couple of folks who admin one box. Granted, they are either a large SUN and IBM hardware, running databases, web servers, and numerous applications, but look at their job requirement in a slanted way and it would look as though they are responsible for only one system. While another admin can have a couple of thousand desktops all identical and simple. And their value would still be relative to whomever the users are.

Jeff Rothschild, the vice president of technology at Facebook, said in a recent presentation that Facebook has 230 engineers supporting data for more than 300 million users. He says Facebook seeks to maintain a ratio of one engineer for 1 million or more users. Facebook is vague about exactly how many servers it has, saying it's "more than 30,000." But at 30,000 servers and 230 engineers, that's a ratio of about 130 servers per admin.

Microsoft says it has automated its data center operations to the point where its admins can each manage between 1,000 and 2,000 servers. That matters, as the company may pack more than 300,000 servers into its new container data center in Chicago. It expects to support that facility with about 30 employees, including admins and facility maintenance staff.

Jeff Rothschild, the vice president of technology at Facebook, said in a recent presentation that Facebook has 230 engineers supporting data for more than 300 million users. He says Facebook seeks to maintain a ratio of one engineer for 1 million or more users. Facebook is vague about exactly how many servers it has, saying it's "more than 30,000." But at 30,000 servers and 230 engineers, that's a ratio of about 130 servers per admin.

Microsoft says it has automated its data center operations to the point where its admins can each manage between 1,000 and 2,000 servers. That matters, as the company may pack more than 300,000 servers into its new container data center in Chicago. It expects to support that facility with about 30 employees, including admins and facility maintenance staff.

Expansion is difficult, and the setup is very expensive to begin with.

I wonder how this compares cost-wise with Google's offshore data centers.

http://www.datacenterknowledge.com/archives/2008/09/06/google-planning-offshore-data-barges/

only 7 comments in and http://www.datacenterknowledge.com/ has been slashdotted :(

I think you're confusing cause and effect here.

The technique of shifting the CPU workload off the database server onto multiple web servers has happened because of people using slow languages not the other way round. Nearly lock free techniques have to be used if you're going to have hundreds of simultaneous users whether their code is running on the DB server or not.

Saying that it's possible to add more CPUs even to the extreme levels that facebook uses to combat the slowness of PHP doesn't stop something being CPU bound.

So I think I'll stick with the usual implication of "cpu bound" ... "on ordinary hardware" because most people need to pay for their webhosting so "DB Bound" ... "on infinite hardware" doesn't really cut it.

Re vendors: Iceotope makes the cooling system. The demo at SC09 is using servers from Boston Limited, a UK server firm.

Source for excerpt below

"Intel set up a proof-of-concept using 900 production servers in a 1,000 square foot trailer in New Mexico, which it divided into two equal sections using low-cost direct-expansion (DX) air conditioning equipment. Recirculated air was used to cool servers in one half of the facility, while the other used air-side economization, expelling all hot waste air outside the data center, and drawing in exterior air to cool the servers. It ran the experiment over a 10-month period, from October 2007 to August 2008.

The temperature of the outside air ranged between 64 and 92 degrees, and Intel made no attempt to control humidity, and applied only minimal filtering for particulates, using "a standard household air filter that removed only large particles from the incoming air but permitted fine dust to pass through." As a result, humidity in the data center ranged from 4 percent to more than 90 percent, and the servers became covered with a fine layer of dust.

Despite the dust and variation in humidity and temperature, the failure rate in the test area using air-side economizers was 4.46 percent, not much different from the 3.83 percent failure rate in Intel's main data center at the site over the same period. Interestingly, the trailer compartment with recirculated DX cooling had the lowest failure rate at just 2.45 percent, even lower than Intel's main data center."


And although the failure rate was similar, the electricity bills were night and day. So I'm not buying into this unless your running a HUGE data warehousing op with more transactions than WalMart...

Google is too big to fail.

Wow really? That's the best examples of datacenters you can find? Not a single mention of PUE at all. Lots of fancy talk and equipment but no real numbers.

Cisco's datacenter consumes 10MW. If it has a PUE of 1.50 (good for that kind of design) that's only about 6.6MW for racks.

If this were a Google datacenter with a PUE of 1.22 that's 8.2MW for racks.

If cisco can maintain a PUE of 1.5 for that design, they're wasting 1.6MW of power. That's about 13670MWh per year or at $0.08/KWh is about 1.1 millon dollars a year.

But please, continue to refute my statement with clear, unsupported, single-word denials. They carry so much weight in an argument.

Yeah, you're probably right. I mean, no one is putting in raised floor environments anymore. I don't know what I was thinking.

Quote all you want. I run an enterprise data center, and I can tell you that raised floor is certainly NOT dead.

But please, continue to refute my statement with clear, unsupported, single-word denials. They carry so much weight in an argument.

Yeah, you're probably right. I mean, no one is putting in raised floor environments anymore. I don't know what I was thinking.

Quote all you want. I run an enterprise data center, and I can tell you that raised floor is certainly NOT dead.

But please, continue to refute my statement with clear, unsupported, single-word denials. They carry so much weight in an argument.

Yeah, you're probably right. I mean, no one is putting in raised floor environments anymore. I don't know what I was thinking.

Quote all you want. I run an enterprise data center, and I can tell you that raised floor is certainly NOT dead.

Just to get it out of the way, yes, IAADRS (I am a Disaster Recovery Specialist - the "speaks bit and byte" and "cosfi" datacenter visiting type...)

Concrete ? Well, yes. Under the raised floor. What did you want ? Marble ?

No cooling from the raised floor ? why not ?
Overhead network cables and a "newish" cooling solution like here :

http://www.datacenterknowledge.com/

front page. Think...giant heatsink...(first overlord joke gets the boot 8p)

As for concrete, let me introduce you to this wonderful answer : it depends. Mostly on the concrete.

"Ramazan Demirboa

Civil Engineering Department, Engineering Faculty, Atatürk University, 25240 Erzurum, Turkey

Abstract

In this study, the effect of silica fume (SF), class C fly ash (FA), blast furnace slag (BFS), SF+FA, SF+BFS, and FA+BFS on the thermal conductivity (TC) and compressive strength of concrete were investigated. Density decreased with the replacement of mineral admixtures at all levels of replacements. The maximum TC of 1.233 W/mK was observed with the samples containing plain cement. It decreased with the increase of SF, FA, BFS, SF+FA, SF+BFS, and FA+BFS. The maximum reduction was, 23%, observed at 30% FA. Compressive strength decreased with 3-day curing period for all mineral admixtures and at all levels of replacements. However, with increasing of curing period reductions decreased and for 7.5% SF, 15% SF, 15% BFS, 7.5% SF+7.5% FA, 7.5% SF+7.5% BFS replacement levels compressive strength increased at 28 days, 7- and 28-days, 120 days, 28- and 120 days, 28 days curing periods, respectively. Maximum compressive strength was observed at 15% BFS replacement at curing period of 120 days."

http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V23-4KPFKFY-1&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&_docanchor=&view=c&_searchStrId=1085368183&_rerunOrigin=google&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=b8e7b8ce7b5e23b07db5805bf9ad9740

http://www.datacenterknowledge.com/archives/2009/10/22/peering-disputes-migrate-to-ipv6/ Only article I could find with a pic of the cake.

The discussion was comparing blades to minis on the basis of cost, not density, but if you want to talk density...

Well, the cost argument works just as well for rackmount machines. An appropriately specced 1U rackmount is about $6.5k. 8 Mac Minis is about $8k.

I've just measured the rack in my garage - it's 30" deep by 19" across, and a mac mini fits handily into a 7" square by 4 holes in the rack with space to spare. 6U is 18 holes, so I can get 4 slide-out racks of minis into 6U, and 8 minis (including PSUs and wiring) per rack in a 2-across by 4 deep configuration. That's a total of 8x4x2 CPUs (because they're core-2 duos in the mini), or 64 CPUs.

Yes. I estimated you'd get about 8 Minis per 2U. You've measured it out at about 8 Minis per 1.5U - though I suspect you'd have trouble keeping the density you're talking about properly cooled, given the lack of high-volume forced cooling in the Mini. The ones at the upper rear of the rack, in particular, would suffer.

I'm getting less and less convinced about the need for keeping computers cool, and it seems I'm not alone. Certainly the linux boxes in my garage (which isn't at all cooled) have had no problems over the Summer here in the Bay Area. Those boxes suck a *lot* more power than the minis and run a *lot* hotter.

Actually, it works out extremely well. Those rackmount servers have Nehalem based CPUs, which are roughly twice as fast as Core 2 CPUs at the same clock speed. Even being conservative, the 48 cores in 6U of rackmount servers will be quite a bit faster than the 64 cores in the Minis. The same is true of using Xserves, though the price comparison probably won't be quite as favourable.

I include part of my other post below

Sometimes good things come in small packages. The Nehalem chips *are* better chips, but core-for-core they're nowhere near 2x as fast (let alone the ridiculous 10x as fast!).

I invite you to look at the mac-mini benchmarks (which are obviously the right ones :) and the MacPro2,1 benchmarks (which use the same processor, and run slightly faster than the blades above). Here's a summary of the relative ratios for different types of computing:

• Integer. 7127 : 3162. E5405 is 2.25x
• Floating point: 11849 : 4927. E4505 is 2.4x
• Memory: 2752 : 2650. E4505 is 1.03x
• Stream: 2062 : 1912. E4505 is 1.07x

Bear in mind that the E4505 above is a dual quad core, and the mini is just a dual-core when you compare these. Also bear in mind that they're weighted averages - in some of the individual tests, the E5405 is almost 4x the speed.

Given that the minis are matching (or beating) the number of cores you're talking about, I'd say the advantage is on the mini's side, not the 1U servers. Odd but true.

The thing is, a Mac Mini is not designed to be a 24/7 server. It also lacks numerous "standard" server features like power supply redundancy, remote management capabilities, and hot-swap hard disks. A Mini comes with a 1 year warranty standard, while even a low-end Dell server will come with a 3 year warranty. A Mini is difficult and time-consuming to service. A typical rackmount is trivial to break down and nearly any arbitrary component can be replaced in a matter of minutes.

It's only anecdotal, but I have had a mini running almost constantly 24/7 for a few years now, almost constantly under load since it does the image-processing for network cameras linked up inside the house and garage. Traditional servers are designed around performance and they run hot, therefore needing all this industrial design to keep them c

Microsoft has already started moving/building data centers out of Washington state, which is both perfectly legal and bound to hurt the shortsighted WA state govt that thought it could just start changing tax laws on companies without any repercussions.

+1, informative.

http://www.datacenterknowledge.com/archives/2009/08/05/microsoft-migrates-azure-citing-tax-laws/

Summary: Microsoft decided it made business sense to build a huge data center in Washington state. Washington state then re-interpreted its tax code to extract much more money out of the huge data center. Microsoft halted development of the data center and announced that it would migrate most of its customers to a data center in a different state.

Washington state will now miss out on the taxes they would have collected. I can damn well guarantee you that the result is a net loss of tax revenue for the state. Classic case of knifing the golden goose.

The issue is when the systems designed to create redundancy actually cause the failure (a transfer switch causing a short, etc.) Also with a couple seconds of searching I was able to find one extended downtime caused by safety procedures and not lack of redundancy:

http://www.datacenterknowledge.com/archives/2008/06/01/explosion-at-the-planet-causes-major-outage/

I have seen other cases where entire datacenters were shut down because some idiot hit the shutdown control (required by fire departments for safety reasons, you don't want thousands of amps flowing through a building you are spraying water into), etc.

This is all about Dell buying service revenues in health care and government, two areas that are getting s huge boost from the Obama stimulus package. In today's conference call, Perot Systems said it gets 48 percent of its revenue from the healthcare sector, and 25 percent from government. That's a strong footprint in two key growth areas, which is why Dell is paying a 65 percent premium to Friday's share price for Perot.

It's interesting how the government is portraying data centers as the problem. The video Kundra showed is like a bad political ad: when the data centers appear, the music turns ominous and the background grows dark. But when cloud computing is mentioned, the music turns happy and the landscape becomes green. I'm all for eliminating redundant technology spending, but where does Kundra believe these "clouds" actually live?

This is being widely discussed in the hosting industry. The full jury ruling is online, and there's additional analysis and discussion at the Web Host Industry Review, TechDirt and Data Center Knowledge.

Why is it that we haven't built datacenters in places with natural cooling. gives a new meaning to the phrase, sent to siberia.
I know I am not the only one with this obvious Idea.

Why is it that we haven't built datacenters in places with natural cooling. gives a new meaning to the phrase, sent to siberia.
I know I am not the only one with this obvious Idea.

My computer doesn't have to be hooked up to the interweb to work...

Since you are not a candidate for Internet-connected, virtualized, on-demand scalable computing resources (aka "cloud computing"), you are not attracted to cloud computing's value proposition.

For those of us who need these things, vendors such as Amazon, Microsoft, and Google are building services we definitely want to buy. Amazon's simple storage service, for example, had 40 billion objects in its repository as of February, 2009.

Automated trading has been a growth engine for data center colo providers, who market "proximity hosting" space within their facilities to hedge funds who believe that they can get an edge by being physically closer to the exchange's servers than their trading rivals. In other words, once you max out the wire speed, it's about physical distance. Savvis says its ultra-low latency offerings can reduce connection speed to microseconds, rather than milliseconds. The NYSE's data center expansion purportedly will enable it to offer colo space to low latency trading operations.

It does make sense. That's exactly what i/o Data Centers is doing in Phoenix. They're installing a thermal storage system at their huge Phoenix ONE data center. The building chillers will cool a solution of water and 28 percent glycol. The thermal storage tank contains Cryogel ice balls, which freeze when the system is charging at night, and then cool the glycol solution during the day. The glycol solution is then pumped through a heat exchanger, which chills water in a separate loop used in the data center.

One of the companies that Andreessen/Horowitz are funding is virtualization/cloud computing startup webappvm, which recently demonstrated its technology at a Sun Microsystems event.

This recalls Google's Toilet ISP April Fool's prank. But there was actually a broadband-via-sewer called CityNet, which raised $375 million back in 2001. It also used robots to run the cabling. This story recaps that initiative, and can serve as topical reading while TFA is Slashdotted.

This recalls Google's Toilet ISP April Fool's prank. But there was actually a broadband-via-sewer called CityNet, which raised $375 million back in 2001. It also used robots to run the cabling. This story recaps that initiative, and can serve as topical reading while TFA is Slashdotted.

This is becoming an annual event for Heiliger, who also complained about server vendors at GigaOm's Structure 08 conference last year. Facebook used to buy a lot of cloud-optimized gear from Rackable/SGI, but no longer appears on the list of their largest customers. Makes you wonder if they're not going to follow Google's lead and build their own servers.

Data Center Knowledge has a roundup that looks at some of the problems with this approach, including security issues related to running a server on a desktop app and bandwidth consumption. If your browser-hosted site gets busy, you think your ISP won't notice?

Erlang is used in Facebook Chat, which just hit 1 billion messages a day. Eugene Letuchy discussed Facebook's use of Erlang at the Erlang Factory event.