SolidNode

The mobile revolution is coming. and the world’s infrastructure isn’t ready yet. That’s the clear message from Equinix CIO Brian Lillie, who says new devices like the iPad are making mobile video a reality.

“Think what happens when the iPad takes off,” Lillie said. “People are surfing the Internet and streaming video on their mobiles, and it’s just starting. As prices come down for these new devices, the growth of traffic will surpass expectations.”

Lille was the keynote speaker this morning at Data Center World Spring 2010 at the Gaylord Opryland in Nashville, Tenn. He warned the audience of 800 data center professionals that the coming tsunami of mobile traffic will be disruptive to their networks and facilities.

“The mobile Internet will have a profound impact,” said Lillie. “We’re just at the beginning, but we’re not ready. This is the message we’re bringing to the carriers. And they know it.” He said some carriers are preparing to increase their network capacity ten-fold over the next decade.

Lillie said the growth is being driven by the development of mobile apps for the iPhone, Blackberry, Android phones and other mobile devices. As these new apps bring a universe of everyday tasks into the palms of users’ hands, usage is accelerating – along with the data traffic streaming across global networks.

As these apps and devices proliferate, network latency will become an increasingly important challenge for corporate IT departments and data center managers.

“We believe the mobile wave will be bigger than the fixed Internet wave,” Lillie said. “It has the potential to change the way we do business.” An example: Lillie said Equinix is redesigning its systems for mobile access, including customer management portals and capacity planning tools for the iPhone and Blackberry.

A similar transition awaits many of the companies represented at Data Center World, he said. “What are you doing to handle this growth?” Lillie asked. “You have to think about your networks and how they may need to grow. The numbers (for traffic volume) are getting to be extraordinary. Most of it is video, and it’s just starting to kick off.”

One technology likely to see broader adoption as a result of mobile traffic is WAN optimization, which can help companies better manage their existing network infrastructure to (See our WAN Optimization Outlook for an overview of the major players in this niche).

From:
Lillie: Mobile Growth to Have ‘Profound Impact’

Popularity: unranked [?]

Microsoft CEO Steve Ballmer today emphasized that “when it comes to the cloud, we are all in.” He shared that message first in a speech at the University of Washington, later in an all-staff email, and also in a major ad campaign the company is launching today.

Most of Ballmer’s talk focused on the end-user experience of cloud computing services. But he brought a data center with him: one of the next-generation containers that Microsoft data center GM Kevin Timmons described yesterday in a presentation in New York. The prototype (seen above) is the latest in a series of evolving designs for Microsoft’s containers, also known as an IT-PAC (pre-assembled component). The design is likely to undergo additional refinements as Microsoft continues scouting locations for its next major data center.

“It includes the equivalent of about 10,000 servers,” said Ballmer .”It’s a cool, next-generation concept. We used to have to stick fire hoses into these things to cool them down. (With this) next generation technology, you can put a garden hose in to one of these things to cool down.”

From a data center perspective, one of Ballmer’s most interesting comments came during the question-and-answer session with students, in which he hinted that Microsoft may offer a container packed with Azure technology as a product for on-site installation.

“When you walk outside and see one of those containers, it would be OK with me if we have to dump one into every country or sell some to some people who want to implement them,” said Ballmer.

Sell a container? These kind of statements are sometimes parsed out of context by media. So here’s the full transcript of the exchange:

QUESTION: “So, I’m curious that we shouldn’t care where information is because it should be completely abstracted away, but it seems the laws and regulations do care where information is. I’m just curious how we should manage and take care of that.”

STEVE BALLMER: “That’s why we talk about a partner cloud, a customer cloud and a public cloud. I mean, I think for a lot of reasons it will be many years before many government organizations will grow comfortable with the notion of their data or citizen data living outside of the jurisdiction.

As technology people we can talk about whether that makes sense or doesn’t make sense, and why the protections can be the same, but it turns out the regulatory environment, as you highlight, is imperfect. I mean, the truth of the matter is – our guys were trying to explain this to me a week or two ago – the same data held in the same place but under different operating circumstances has different regulatory blah, blah, blah, blah, blah.

And we can’t assume all of the world’s important countries are going to even standardize the regulatory framework. That’s why when you walk outside and see one of those containers, it would be OK with me if we have to dump one into every country or sell some to some people who want to implement them.

I love Slovenia, it’s a great country, but there’s only a million and a half Slovenes. This company is not likely to build part of our public cloud in Slovenian anytime soon. So, somebody should be able to implement a Windows Azure cloud in that country. They should be able to buy a device that looks like that or a set of devices and go do that and have that be affiliated for the rapid advance of technology with other things going on in the world.

So, I hear you and I agree that there’s a set of issues, but they don’t have to be constraints.

Here’s just one simple way to think about it. Will all of the world’s centralized compute, storage and networking infrastructure all be built out by four or five companies, Microsoft, Amazon, Google, the cloud guys? Will we buy every server computer and every piece of storage in the world? No, that isn’t going to happen. I don’t think that – if you just think about the level of capital investment that involves.

We need to permit the private cloud, and the kind of thing we’re showing, the kinds of things we’re doing with Windows Azure is about making sure there’s a public version and there’s a customer version, and there can be a government version, all based on the same core technology, and there’s some innovation to go make that happen.”

From:
Microsoft ‘All In’ on Container-Powered Cloud

Popularity: unranked [?]

Competitive pricing industry wide requires that data center providers find smart and creative ways to keep power costs low. Here are five approaches that can help.

More:
5 Ways to Reduce Data Center Power Costs

Popularity: 27% [?]

Skytap claims it is the first enterprise private cloud provider to offer network automation features to customers, enabling them to create multiple secure networks that support advanced virtual routing and IT policies.

Sundar Raghavan, Skytap chief products and marketing officer said the automation features, which the company has been working on for the past six months, gives Skytap a lead over other private cloud providers that are working on similar functionality. Cloud providers are currently only able to offer a single network in the cloud, Raghavan said.

“We enable customers to run their enterprise applications unchanged,” said Ian Knox, senior director of product management. The network automation features enable customers to build a computing environment with application images and complex networking topologies using firewall-based security policies. Access to the cloud is through self-service Web UI and APIs.

According to Skytap, examples of complex network topologies could be server machines with multiple network adapters; server clusters with fail-over configurations and shared services; and the ability to add virtual routers, firewalls, and gateways. The self-service network allows customers to save network topologies and virtual data centers as templates, and provides role-based access for users to deploy pre-packaged data center building blocks.

Customers pay for the virtual machines and computing hours they consume, which typically runs at $2,000 a month for 25 machines.
Customers are not limited to the number of networks that they build but Raghavan said customers typically build five or six networks. The networks are included in the price.

From:
Skytap: Automating Networks in Clouds

Popularity: 6% [?]

TorrentFreak has an intriguing story this morning about Twitter’s use of file distribution technology from BitTorrent to boost the performance of its back-end infrastructure, which has required constant attention to keep pace with the growth of the popular microblogging service.

“With BitTorrent, Twitter is planning to distribute files faster and more efficiently, saving time and precious resources and improving the scalability of Twitter’s operation,” writes Ernesto at TorrentFreak (via Glyn Moody).

Here’s the full description: “Twitter’s new project, codenamed ‘Murder’, will not use the bandwidth of Twitter users. Instead, it will transform the site’s servers into a large BitTorrent swarm that will distribute file updates using BitTorrent technology. The setup is pretty straightforward. Murder uses a ’seeder’ server where the new files will be distributed to thousands of ‘peer’ servers. Because all servers assist in the deployment of the files, it will only take a fraction of the time it would otherwise take when files are distributed from a central server. This server-to-server BitTorrent technology also explains the name ‘Murder’ which is used to describe a flock of crows.”

Read more at TorrentFreak.

Original post:
Twitter Using BitTorrent to Speed Servers

Popularity: unranked [?]

We often think of computers as a very modern phenomenon, but there were actually plenty of computers around 50 years ago. They just weren’t an everyman commodity, instead limited to goverment and corporate use. And they certainly weren’t small. Some of them had imaginative names like Whirlwind, Colossus and Pegasus, while others were slightly less poetic with names like Z4, AN/FSQ-7 and ENIAC.

Below we have listed as many as 19 examples of computers from the early days, pioneering efforts that although cutting edge in their day now look lovingly retro.

These computers didn’t use the same kind of components as we do today. The computers in the 1940s and 1950s were mostly based on vacuum tubes. Transistors showed up late in the game, and integrated circuits were just a distant dream and didn’t start showing up in computers until the 1960s, and then in very limited capacity. How tempting it would be to travel back in time and show the engineers of these computers a normal modern-day PC, just to see their reaction.

We have listed the completion year for each computer, although often work on them had begun several years earlier (they were huge projects). We’ve arranged them in chronological order, oldest first. Please note that these are just a sample, there are plenty we didn’t include (in order to make this a blog post and not a book ).

Z4

Year: 1944

Designed by the legendary German engineer Konrad Zuse, the Z4 was a follow-up to its pioneering predecessor, the Z3 computer he built in 1941 (the world’s first programmable, automatic computing machine). The Z4 used about 4,000 watts of power and ran at approximately 40 Hz. It had 64 32-bit registers, the equivalent of 512 byte of memory. One addition took 0.4 seconds.

Above: The Z4 computer, as seen in a German museum (in Munich).

Colossus

Year: 1944

Two generations of Colossus, the Mark 1 and Mark 2, were used by British codebreakers to decrypt coded German messages at the end of WW2. It processed 5,000 characters per second (it could process faster, but then the paper tapes holding the data would break). The existence of Colossus and other British codebreaking machines remained secret until the 1970s out of fear that widespread knowledge would encourage more efficient encryption algorithms.

Above, top: The Colossus in its heyday. Note the punched paper tape running on the right side. Above, bottom: A reconstructed Colossus.

ENIAC

Year: 1946

When the ENIAC was announced in 1946 the press immediately started calling it a “Giant Brain”. ENIAC was the world’s first general-purpose electronic, digital computer and is probably the most famous of the ones included in this article. It weighed 27 tons. Among other things, ENIAC was used for calculations for the creation the hydrogen bomb. Programming the machine could take weeks, since after the program had been figured out on paper you first had to manipulate the various switches and cables that controlled the programming and then follow that with verification and debugging.

Above, top: The ENIAC in all its glory. Above, bottom: Old-school programming?

Whirlwind

Year: 1951

The Whirlwind was the first computer to use video displays for output. The first version had 512 byte of main memory and could do 20,000 instructions per second, although a switch to a different kind of memory later doubled its performance and made it the fastest computer of its time.

Above, top left: The Whirlwind. Above, top right: Closeup of the circuitry. Above, bottom: The control room.

UNIVAC I

Year: 1951

An acronym for UNIVersal Automatic Computer, the UNIVAC I was the first US-produced commercial computer. It was designed by the inventors of the ENIAC. A total of 46 systems were built and delivered. It weighed 13 tons (29,000 pounds), ran at 2.25 MHz and could perform 1,905 instructions per second. The UNIVAC I cost up to $1.5 million per system.

Above: The UNIVAC I, built by Remington Rand (see their nifty logo top left in the image).

WITCH

Year: 1951

Short for the Wolverhampton Instrument for Teaching Computing from Harwell, the WITCH was also known as The Harwell Dekatron Computer. It was slow (a multiplication took 5-10 seconds), but this was justified by its ability to run long periods of time unattended. It could therefore be left on its own with a large amount of input data. At one point it was left running over the Christmas and New Year holiday and was still working when the staff came back 10 days later.

Above: The WITCH in use. Is it just us, or do they look a bit confused?

BESK

Year: 1953

Pingdom being Swedish, we had to include this Swedish computer from 1953. BESK stands for Binär Elektronisk SekvensKalkylator, which is Swedish for Binary Electronic Sequence Calculator. The main memory was 512 40-bit words, the equivalent of 2,560 byte. An addition could be performed in 56 microseconds, and a multiplication in 350 microseconds. For a short time it was the world’s fastest computer. Small aside, “besk” means “bitter” (as in taste) in Swedish, but besk is also an alcoholic beverage from the south of Sweden. The name was a pun sneaked in by the computer’s creator, who had previously had the computer name COGNAC rejected by officials.

Above: The control panel for the Swedish BESK computer.

IBM 702

Year: 1955

The IBM 702 had been announced as early as 1953, but the first production model wasn’t installed until 1955. It was a commercial computer that could be leased from IBM. The system could have a maximum of 11,000 7-bit characters of main memory, i.e. roughly 10 kilobytes. It could do 3,950 additions or subtractions per second, but multiplication and division was significantly slower.

Above: An IBM 702 installation.

IBM NORC

Year: 1954

The IBM Naval Ordnance Research Calculator was arguably the first supercomputer and was the most powerful computer of its time. It could perform 15,000 operations per second, and the first version had 2,000 64-bit words of main memory, roughly the equivalent of 16 kilobytes.

Above: Various angles of the IBM NORC.

IBM 305 RAMAC

Year: 1956

This computer is most famous for being the first commercial computer delivered with a hard disk drive. The hard disk drive could store a total of just under 5 MB and consisted of 50 24-inch diameter disks. The 305 RAMAC was one of the largest computers IBM ever built. (If you find ancient hard drives fascinating, check out our post about the history of computer data storage.)

Above: Yes, those huge units in the foreground are hard disk drives. Each storing a massive 5 MB…

Bendix G-15

Year: 1956

The Bendix G-15 weighed 450 kg (950 lb) and cost around $60,000. It had 2,160 29-bit words of memory, the equivalent of about 7.6 kilobyte. The G-15 has sometimes been called the first personal computer, although there are disagreements about this. More than 400 were made.

Above: The Bendix G-15. It looks like a very big tower desktop computer. Kind of.

Pegasus

Year: 1956

The British computer Ferranti Pegasus was designed and built to be cheap and reliable. It had 5,120 40-bit words of memory, the equivalent of 25 kilobyte, plus 56 words (280 byte) of fast memory. A Pegasus 2 from 1959 is still operational at the Science Museum in London. It is the world’s oldest working digital computer.

Above: A Pegasus 2 at the Science Museum in London. The cabinet was built by Rolls Royce, hence the use of car door handles for the doors. Note also the inserted clock at the short end.

AN/FSQ-7

Year: 1958

A successor to the Whirlwind, based largely on the design of the never-realized the AN/FSQ-7 was developed by IBM in collaboration with the US Air Force to be used with the SAGE air defense system. It is sometimes incorrectly referred to as the Whirlwind II. One computer took up 2,000 sqm of floor space (roughly half an acre) and weighed 275 tons. They are the largest computers ever built (52 of them were made). The AN/FSQ-7 could perform about 75,000 instructions per second.

Above, top: An installation of the AN/FSQ-7. Each cabinet a built-in phone to save time when calling in problems (seen here at the short end of the nearest cabinet). Above, bottom: SAGE control consoles. A sign of different times: each console had a built-in cigarette lighter and ashtray.

IBM 7090

Year: 1959

A typical IBM 7090 system cost $2.9 million and was designed for large-scale scientific and technological applications. Among other things, it was used by NASA to control space flights. A 7090 system is featured in the movie Dr. Strangelove. In 1961, a later version, the 7094, became the first computer ever to sing (the song Daisy Bell). This was the inspiration for a scene in 2001: A Space Odyssey.

Above:The IBM 7090. Trivia: The second man on the left is Smith DeFrance, founding director of the NASA Ames Research Center.

AKAT-1

Year: 1959

The Polish AKAT-1 was the world’s first transistor-based differential analyzer, designed specifically to solve systems of differential equations. It was never mass produced due to the country’s policies at that time.

Above: The AKAT-1.

Datasaab D2

Year: 1960

Never massproduced, the Datasaab D2 was a concept computer build in Sweden. It weighed “only” 200 kg and could be placed on a desktop. It held the equivalent of 15 kilobyte of memory and could perform 100,000 additions per second. It was a prototype designed to test the feasibility of computerized navigation aid in aircraft. Datasaab was the computer division of the aircraft manufacturer Saab, which made fighter jets for Sweden.

Above left: The Datasaab D2 in its entirety. Above right: Closeup of its control panel.

BRLESC I

Year: 1962

The name is an acronym for Ballistic Research Laboratories Electronic Scientific Computer. It was, as its name suggest, designed primarily for scientific and military tasks. It could do five million operations per second and had 4096 72-bit words of memory, the equivalent of 36 kilobyte.

Above: The awesome-looking console for the BRLESC I computer.

Honeywell 200

Year: 1963

The Honeywell 200 and its successors were introduced to compete with affordable commercial computers from IBM (specifically the IBM 1401). The native assembly language used to program the Honeywell computer was named Easycoder. Yes, at that time, assembly language was considered easy to code in. Honeywell ran an ad campaign over several years that they called the Liberator, using various very creative sculptures made from computer parts (one example available here).

Above: The H200 at work.

UNIVAC 1108

Year: 1964

The transistor-based UNIVAC 1108 supported up to three CPUs and up to 262,144 36-bit words of memory (more than 1 MB). The memory used integrated circuits (quite rare at the time) instead of the thin film core memory used in its predecessor, the 1107.

Above: A later model of the 1108 from 1969.

Final words, acronyms and the MANIAC

As you can see from some of the examples above, acronyms were highly popular. Some scientists were so fed up with this acronym mania that they started mocking it. There was for example a computer called MANIAC (I and II), which stood for Mathematical Analyzer, Numerical Integrator, and Computer.

We always think it’s fascinating to look back and see how things were in the early days of computing. It truly illuminates how far we have come. Today we have more computing power in our pocket than what would fit in entire buildings in the past. Our most modest smartphones widely exceed the performance and storage capacity of these early behemoths.

Data sources: Almost needless to say, Wikipedia was an invaluable starting point when researching this post, which in turn uses a ton of other places as data sources. Too many to mention here.

Image sources: Z4 by Clemens Pfeiffer; Colossus old image by UK government, rebuild image by MaltaGC; ENIAC images are from the US Army; Whirlwind top left from Computerhistory.org, circuitry by Dpbsmith, control room from MIT; UNIVAC I from US Army; WITCH from the Computer Conservation Society; BESK by Liftarn, IBM 702 from the US Federal Government;IBM NORC images from Columbia University; IBM 305 RAMAC from the US Federal Government; Bendix G-15 by unknown author (stored by Wikipedia); Pegasus; AN/FSQ-7 from US Air Force, SAGE consoles by Steve Jurvetson; IBM 7090 by NASA; AKAT-1 by Topory; Datasaab D2 images by Lars Aronsson; BRLESC I from US Army; Honeywell 200 image from unknown source; UNIVAC 1108 by the manufacturer.

Original post:
Retro delight: Gallery of early computers (1940s – 1960s)

Popularity: 6% [?]

Here is a roundup of some regional data center news around the United States:

• Midwest ISO seeks new home for backup data site. Major power consortium Midwest ISO is looking for a new home for their $35 million backup data site after the Boone County (Indiana) Commissioners voted down their rezoning request for land.  Due to a stipulation from the Department of Homeland Security, Midwest ISO was looking to build its 17-acre, $35 million data center in a rural area, away from major highways, railways and airports.  This is the second time they have been turned down in Boone county. Ten years ago they were turned down when they wanted to build a headquarters near Zionsville.  Midwest ISO recently announced expansion of their headquarters in Carmel, Indiana; investing around $110 million to upgrade computer equipment and adding 70 jobs over the next five years.
• Growth in Michigan for Online Tech.  CEO of Ann Arbor-based Online Technologies Yan Ness credits the cold Michigan temperatures for the continued growth of his business.  Online Tech has three data centers -  a 30,500 square foot facility in Genesee County, 10,500 square feet south of Ann Arbor in Avis Park and the newest facility, 5,000 square feet in down town Ann Arbor. Recently Online Tech won an Ann Arbor SPARK’s Fast Track award for a fourth consecutive year.  The award honors companies that maintain average annual revenue growth of 20% for three consecutive years.
• Texas data center demand outpacing supply. Sheila Bellinger, VP and Southwest director of the Grub & Ellis National Data Center Group spoke at Datacenter Dynamics Dallas Thursday on the topic of supply and demand for data center space in the Dallas-ForthWorth area.  Bellinger noted that about 1.5 million square feet of data center space was absorbed by the Dallas-FortWorth market over the past three years.  Although power costs in the area are expensive, other factors such as fiber-optic infrastructure, favorable regulatory environment and low construction costs are what drives demand.  Bellinger said that Texas is consuming more power than any other state in the U.S.

I am a little bit partial to this last story: Waterloo, Fastest (Internet) Town in the Midwest.  Stacey Higginbotham from GigaOm reports on a local newspaper story about Waterloo Iowa having faster broadband than even the most wired cities.  Mediacom selected Waterloo to (soon) have 105 Mbps cable Internet service on the downlink thanks to the town’s willingness to get involved in bringing better broadband to its residents.

Original post:
Regional Roundup: Indiana, Michigan, DFW

Popularity: 12% [?]

Amazon Web Services experienced an outage in one of the East Coast availability zones for its EC2 service early Wednesday due to power problems in a data center in northern Virginia. Failures in a power distribution unit (PDU) resulted in some servers in the data center losing power for about 45 minutes. It took several more hours to get customer instances back online, with all but a “small number” of applications restored within five hours.

The issues started at 4 am East Coast time Wednesday, and affected one of the three availability zones in Amazon’s East Coast operation. The zones are designed to provide redundancy for developers by allowing them to deploy apps across several zones.

“A single component of the redundant power distribution system failed in this zone,” AWS said in its status report. “Prior to completing the repair of this unit, a second component, used to assure redundant power paths, failed as well, resulting in a portion of the servers in that availability zone losing power. Impacted customers experienced a loss of connectivity to their instances. As soon as the defective power distribution units were bypassed, servers restarted and instances began to come online shortly thereafter.”

Amazon is known to operate a major data center in Ashburn, Virginia. EC2 previously experienced brief downtime in both June and July.

Original post:
Power Outage for Amazon Data Center

Popularity: 12% [?]

Open source content management systems are beginning to gather quite a reputation and customer base lately thanks to the recent announcement by the White House of its move from proprietary solutions to open source CMS heavyweight, Drupal.  Full service hosting provider NeoSpire has been a long time advocate of open source software and specializes in solutions incorporating Drupal.

Do It With Drupal Seminar

From December 9^th to 11^th, 2009 in New Orleans, the Do It With Drupal Seminar will take place and NeoSpire is not only one of its top sponsors but will be exhibiting as well.  The seminar promises to have some of Drupal’s top developers on tap to teach more about the program as well as information from community building experts.

Patrons of the event can expect to see successful Drupal sites and meet with like-minded professionals, developers, designers and more all intertwined within the Drupal community.  The event will allow interested participants the chance to see the actual workings behind the architect and building of a Drupal web site.  Topics such as building and growing successful web sites with Drupal will be covered.

The cost of this three-day event is $1195 until December 8^th.  Thereafter, walkup registration is $1295.

NeoSpire

NeoSpire managed hosting provides a number of hosting solutions designed to meet the demands of organizations needing a 100% guaranteed up-time.  NeoSpire has been in the web hosting business since 1999 and continues to provide extremely secure and reliable solutions for their hosting customers.  The hosting provider offers several managed hosting options, co-location services, application hosting and more.

One of the more interesting offers by this company is its Drupal CMS Hosting.  NeoSpire offers fully managed Drupal hosting that covers everything from simple web sites to load balanced and redundant set-ups.  The company even has a specially dedicated Drupal support team that are considered the best server admin and tech support people in their field.  Not many web hosting companies available on the Internet sport the same offering or specialized support team.  It is a testament to the security and reliability of both NeoSpire and Drupal.

Drupal

Drupal is an open source CMS program that was initially created in 2000 by an enterprising man, Dries Buytaert.  What initially began as a small news site allowing friends to leave notes to each other gradually grew into the full-blown content management system available today.

The Drupal CMS is extremely extendable with several hundreds of modules to add onto the main program and hundreds of themes to change the overall look.  As an open source project, many third party programmers are heavily involved with the project creating new and innovative modules almost every day.  Drupal users can expect to be able to find just about anything they require within the Drupal module download area.  Administration, e-commerce, file management, statistics, and spam prevention are just a small percentage of what can be found.

Conclusion

A world-class open source CMS program coupled with a world-class hosting solution provider is a dream combination.  NeoSpire and Drupal are, indeed, doing it.

Popularity: 12% [?]

Here’s one of the most unusual data center designs we’ve seen. The CLUMEQ supercomputing center in Quebec has worked with Sun Microsystems to transform a huge silo into a data center. The cylindrical silo, which is 65 feet high and 36 feet wide with two-foot thick concrete walls, previously housed a Van de Graaf particle accelerator. When the accelerator was decommissioned, CLUMEQ decided to convert the facility into a high-performance computing (HPC) cluster known as Colossus.

We first noted the development of the CLUMEQ site earlier this year when Marc Hamilton of Sun discussed its unique design, but offered scant details. Additional information about the design of the facility and its cooling system were discussed at the Sun HPC Consortium last month in Portland, Oregon.

The CLUMEQ Colossus cylinder features an interior “hot core” (as opposed to a hot aisle) in the center of the building and uses the outside ring of the facility as the cold air plenum. The cabinets are arranged in a ring on each floor, facing the outside of the silo. The floors supporting each ring of cabinets are comprised of grates rather than solid flooring to facilitate airflow through the facility.

The cooling coils and air handlers are located in the basement. Chilled air flows upward through the outside cold aisle and through the racks of servers. The waste heat exits the rear of the racks into the hot core, and is returned to the basement via the cold aisle.

The air flow pattern is maintained through differential air pressure – maintaining a higher air pressure in the cold aisle than the hot aisle. This keeps the air moving through the facility, which has a blowing capacity of 180,000 CFM and can cool up to 1.5 megawatts of electrical load. Up to 300 kilowatts of cooling capacity can be supplied by free cooling using fresh air from outside the facility.

“CLUMEQ silo totally blows up the paradigm of data center design,” says Nicolas Dube of Sun, who began work on the project as a graduate student at Universite Laval in Quebec. “The silo, by itself, is the CRAC (computer room air conditioner). The whole facility cools itself.”

As for computing horsepower, Colossus will have a peak of 86 teraflops of compute power. It’s equipped with a Sun Constellation HPC systems featuring 10 fully loaded Sun Blade 6048 chassis, 1 petabyte of Lustre storage and Sun J4400 storage arrays.

The data center racks are spread over three floors, with the switches on the second floor to keep the cable runs as short as possible.

For a full description of the CLUMEQ design, check out this video from Sun, which runs about 6 minutes.

Additional details are available in PDFs of presentations by Marc Parizeau of CLUMEQ and Nicolas Dube of Sun.

OTHER STORIES YOU MAY LIKE:

• Who Has The Most Web Servers?
• The Gallery of Exploding Servers
• Inside the ‘James Bond Villain’ Data Center
• Google’s Chiller-Less Data Center
• Inside a Cloud Computing Data Center

Original post:
Wild New Design: Data Center in A Silo

Popularity: 12% [?]