Hewlett Packard (s hpq)  today announced a new line of servers, a data center mapping program and some consulting and financing services aimed at companies that build out mega data centers. Potential purchasers of the new HP machines include those building cloud computing offerings and enterprise customers trying to build their own clouds or high-performance computing clusters.

Problem is, HP is late to the mega data center party. SGI (formerly Rackable) has built special-purpose machines for years, and keeps introducing new options for the mega data center. IBM (s ibm) even launched its own, highly proprietary iDataPlex hardware for the same market last year. Two years ago Dell (s dell) created a custom-order business aimed at serving this market while HP focused on better blades for enterprises and building custom setups for clients. HP is now combining some of the features its rivals’ cloud computing servers already offer, such as stripping out redundant power, sharing fans and making things easily interchangeable, as part of its new HP ProLiant SL server family. The resulting servers use 28 percent less power than standard rack-mounted servers, according to HP.

Creating some type of product portfolio for this market, however delayed, makes sense, and Christine Martino, VP and general manager of HP’s Scalable Computing & Infrastructure group says the company thinks it can compete with the existing offerings in the space quite well. She also stressed that both the data center mapping software (The HP Datacenter Environmental Edge), which helps cut costs by finding inefficiencies in the data center, and the consulting would be of use for enterprise customers that have the experience building out large-scale data centers that companies such as Google, (s goog) or Amazon (s AMZN) have.

SiCortex, a company that makes a green supercomputer using proprietary chips and some “Back to the Future” styling, is seeking to sell its assets by the end of June. (Check out what’s for sale here.) According to a story at HPCwire, SiCortex was seeking a third round of financing (it finalized a $37 million round last September), but one of its five venture backers pulled out. I called SiCortex CEO Chris Stone to get more information, but have not heard back.

Reportedly, the 5-year-old company was doing well, but in these hard economic times, it’s possible that a cash-strapped investor just couldn’t front SiCortex the money to continue. EETimes reported a similar capital crunch leading to the closure of video processing chip firm Ambric last November. However, there may be an industry trend working against SiCortex as well. In general, supercomputers have moved from being proprietary systems to open systems built using commodity hardware and open-source software. Supercomputers are now defined by their jobs, not their hardware. While processors such as IBM’s Cell (s ibm) chip and Nvidia’s (s nvda) graphics chips are being used to augment the x86 CPUs in some HPC systems, for the most part, specialty chips are a dying breed. For example, earlier this year, SGI, which made proprietary machines for the HPC industry, filed for bankruptcy and sold its assets to Rackable Systems (which has changed its name to SGI). So I wonder, is SiCortex’s lack of money a sign of a venture capital problem or a supercomputing industry problem?

John West over at insideHPC, tells me that he thinks the large upfront investment in SiCortex’s hardware that it needed to recoup was what ended up hurting it. He emailed that the company had sold 80 machines since launching its computers in 2007, and had a sales pipeline “tens of millions of dollars deep,” but wasn’t profitable. So it simply may have run out of cash.  In that case, its failure may be a sign of both the venture industry’s reluctance to invest in capital-intensive businesses, and the difficulties facing a specialty hardware company today.

Below I’ve embedded an old video interview featuring the SiCortex personal supercomputer. Unfortunately, I didn’t show its main product, which features a rack of machines that can be accessed through a DeLorean-style door that lifts up, rather than opens out. That, and the low power consumption, are pretty neat.

Rackable quietly purchased SGI from bankruptcy court for a mere $42.5 million. It’s a sad end for SGI, a once great computing pioneer that missed the x86 and Window’s revolution. For Rackable, this gives them access to SGI’s high performance computing products and research and development in the areas of power, cooling, visualization, and storage.

While Rackable is known for their servers and storage portfolio, they have entered the world of green datacenters with their Eco-Logical chassis. As the datacenter market is rapidly changing, the competitive walls are surely closing-in on Rackable. After-all, who would expect Rackable to survive a battle against Oracle, Cisco, HP, and IBM? At least that’s what the big boys want you to think.

Rackable’s portfolio compares quite favorably to the others in the marketplace; albeit without the marketing and services flash of their larger rivals. Additionally, Rackable has some unique products with their CloudRack and CloudRack C2 product lines achieving remarkable density and power/cooling ratios. Heck, Rackable even overs a data center in a container (ICE Cube).

It seems to me with all these goodies, Rackable is a legitimate takeover target. As of this posting, their market cap is around $161 Million. That’s a far cry from the billions Oracle Spent on Sun or Dell’s current $23 Billion valuation. If Brocade, Juniper, Siemens (Extreme Networks), Adtran, Turin, or more wanted into the “unified computing” space, then why not Rackable? Or, does private equity / venture capitalist firms show some vision and combine the assets of several companies together to create a viable challenger to Cisco, HP, and IBM?

M&A speculation is always a fascinating discussion, but for now I’ll say goodbye to a once great visionary company (SGI) while tipping the cap to Rackable for taking advantage of the current economic state of affairs.

Yesterday, Ashlee Vance reports for the New York Times, “Once-Mighty SGI Sold to Rackable for $25 Million.” (thanks to my friend Catherine for posting this to her Facebook page).

This is crazy.  Back in the day, the 1990’s, SGI were the hardware kings of The Valley.

To give you an idea of how little money $25,000,000 is, according to IMDB, Jim Carrey was paid that much to do Bruce Almighty (2003).

What this means is that Rackable could have gotten Jim Carrey to do standup comedy at their Holiday Party, or went ahead and purchase SGI.  They went ahead with SGI.

$25M for SGI is like finding out that the really good looking girl you wanted to go to the prom with in high school.  The one who was on the cheerleading squad, but she was cool because she knew who David Lynch was and listened to Nirvana.  Well, it’s like finding out that she’s now living in a van down by the river.

It’s jarring because it illustrates just how quickly fortunes in this business can change.  There’s a case study to be made (if someone hasn’t already written this up) about how Open Source software combined with the shattering of Moore’s Law brought about a platform that moved past SGI’s proprietary/high-priced solution at such shocking speed that they were too slow to react and their market share collapsed beneath them in a matter of quarters (for example: I seem to recall HP making some significant competitive winbacks at large digital animation studios with Linux in the early 2000’s).

And as they struggled to stay afloat for the past few years, what is left…speaking of which, what is left?

Well, whatever is left is now going to Rackable in their van down by the river.

Wait SGI is still around? I thought they died ages ago. Well I guess the aging workstation producer that started OpenGL and made XFS (a really good/fast file system) has finally had to file for bankruptcy and also is being bought by Rackable. Not a bad move for Rackable but who knows what it really means.

Links:

• Linux Journal

Earlier this week Silicon Graphics was purchased by Rackable systems for 25mill.  It also looks like Sun will be gobbled up to the tune of 7 Billion next week by IBM and Twitter the social tool could be purchased by Google.  Recessions are very interesting times!

If you travel all the way back to 1997, SGI was pulling in close to $4 billion in revenue a year. The company produced some of the flashiest computers on the planet for handling tough graphics jobs. SGI was the next big thing in Silicon …[More..]


But, even so, SGI is one of the “big names” that really came about as close to simply fading away as any I can remember. Just this past weekend, I got into a discussion with someone who asked if SGI was even around any more… and just …[More..]

OSNews.com informs you about the latest news on a vast range of operating systems, from the well-known mainstream OSes, down to small embedded (but also very interesting technically) ones.[More..]

Intel (s INTC) today unveiled its latest and greatest Nehalem chip for servers (now known as the Xeon 5500 series), setting off a round of announcements and articles comparing technical specifications across server vendors. And at 2.93 GHz (with certain tweaks it can get up to 3.33 Ghz), indeed, the chip is screamingly fast. Which is all well and good, but if you’re still unclear as to what all the fuss is about, we’ve broken down for you three areas where Intel’s Nehalem chip changes the game.

Better performance per watt: Just like a Ferrari with 12 horses cylinders will consume more gas than a Civic, high-power server chips have never been shy about gulping electricity. But with Nehalem, this is improving. While the processor can consume up to 95 watts in server applications, when measured in terms of what it can do with that electricity — a metric known as “performance per watt” — Nehalem is plenty sexy. Server vendors from HP (s HPQ) to Rackable (s RACK) have talked about how this particular metric (when tied to the cost of the chips) drives large-volume server purchases, which are becoming a bigger chunk of sales as companies like Google or Facebook build out their infrastructure.

Faster access to memory: One of the issues with multicore processors is that the cores can all process information really fast, but eventually they have to get more information to process from the memory. Under Intel’s previous architecture, they had to access that memory outside of the chip — and do it one by one. Intel has addressed this by including integrated memory on the latest Nehalem chip, and a transport protocol Intel calls QuickPath Interconnect. The integrated memory on the chip stores more information closer to the processor and the QPI manages the flow of information between the cores. Incidentally, AMD (s AMD) first offered this in 2003 with HyperTransport.

The bottom line is that by attacking the memory bandwidth problem instead of throwing more power-hungry cores at it, certain workloads can now run faster.

It has a future beyond servers: As part of the new server announcement, Intel also announced a line of Nehalem-based embedded processors for use in communications markets. As an Ars Technica article notes, the Nehalem architecture is so modular that it can be configured for a variety of different uses — in other words, beyond just workstations and servers.

Rackable (s RACK) announced today an update to its CloudRack servers. The CloudRack C2 servers can run at 104 degrees inside the data center, and they offload power supply to the rack to reduce energy wasted in converting AC electricity from the wall to DC electricity used by the box to 1 percent. Since these beasts can pack 1,280 cores, or 320 processors, into a rack, they’re not exactly in the power-saving category, but the design ensures that the electricity is going to power the processors rather than lost as heat or waste.

The updated servers feature a fan mounted behind the rack, rather than attached to each server, which also cuts power consumption for cooling to 8 percent, rather than 25 percent, of the total energy. Rackspace Rackable also announced that customers will eventually be able to build out servers in the CloudRack trays using Intel’s (s INTC) lower power Atom chips, which they can use for jobs that don’t need the full horsepower of the upcoming Nehalem-based Xeon chips. Customizing processors is one more way that data center operators are trying to boost efficiency.

The rising competition around designing power-efficient, heat-tolerant servers is being driven by a need to lower electricity and cooling costs in a data center contrasted with the need to pack as much computing into a box as possible to run web-scale application increases. Essentially, we need more computing but have less electricity to squander. Rackable can sell into corporate data centers, but its target market is the web world giants running thousands of servers.

It’s a market that’s growing increasingly competitive with Cisco (s CSCO) planning a new line of servers dubbed the Unified Computing system, Dell (s DELL) creating a seperate business unit just to deal with web scale customers, and HP has  a web-scale service design team as well. Intel estimates that 25 percent of its chips will go into web-scale boxes by the year 2012. I’m sure Rackable’s hoping that many of them will go into its servers.

As the need for fast, large-scale computing to power sites like Facebook or even computing clouds has grown, manufacturers such as Rackable Systems (s RACK) are taking notes on server design from Google (s GOOG), which builds its own systems. The goal of their mimicry is to provide more computing power in a smaller form factor while using less energy.

An article in EEtimes today details the emergence of these Google-inspired servers, which include features such as heat-tolerant processors to save on cooling costs, a focus on motherboards containing 12v-only power supplies for servers, putting two servers on one board and stripping out unnecessary parts.

These are all ways Google apparently modifies its boxes to deliver information faster and more cheaply. Rackable’s new CloudRack servers will offer dual servers on one board that crams more computing power into a smaller space, as well as 12V-only motherboards. The use of only 12 volts on a motherboard is supposed to make the power supplies more efficient by reducing the energy lost when having to convert electrical current to run at various different voltage levels.

IBM’s (s IBM) iDataPlex servers, designed for the cloud, have stripped away unnecessary hardware — a move aimed at reducing power-consuming components and saving space. Heat-tolerant processors allow a data center operator to keep air conditioning bills down, saving as much as 4 percent of total energy costs for each degree dropped. So as computing requires more scale, Google’s innovations influence other buyers and sellers of technology even as the search giant slows its own data center construction.

The folks in charge of the SC 08 conference being held in Austin, Texas, this week have trumpeted the phenomenal growth of the supercomputing show, with attendance up by almost 10 percent from the previous year, but I’m beginning to doubt that high-performance computing is driving this growth as much as the broad changes in the data center world. As Ori Aruj, CEO a GM of switch chipmaker Dune Networks, told me when I asked why he was at the show, “This is no longer about high-performance computing and research. This is now a data center conference.”

I’m inclined to agree with him, as there are a lot of networking and storage vendors here with really large and visible booths that seem outsized compared with the HPC market opportunity. There are also attendees here from companies that have little or no business in supercomputing, such as Dune Networks, Isilon or Rackable Systems (s RACK). Some of the 219 industry exhibitors (as opposed to the 118 research exhibitors) can’t possibly make enough in the HPC market to justify such a large presence at the show, although a product manager at Ciena (s CIEN), which makes networking gear,  pointed out that HPC installations can act as an effective advertisement for other business.

So here among the 10,764 attendees at the show one might be forgiven for occasionally forgetting that Microsoft (s MSFT), Intel (s INTC), Nvidia (s NVDA) and a host of other consumer brands aren’t here to talk about basic computing — but supercomputing.

I traveled down to Houston today to check out a data center geek’s version of paradise — the inside of the factory where HP builds their rack-mounted servers and high-value blade systems. I shot plenty of photos that show how a bunch of chips and boards gets assembled into a blade that I will put in a later post, but I also got to take a tour of the inside of HP’s containerized data center, known as a POD. “Tour” isn’t really the right word, since it’s hard to move around inside the 40-foot shipping container filled with racks, but it was pretty sweet to see all that processing power in one place.

I’ve embedded a three-minute video below with Wade Vinson, a thermal engineer with HP, and pasted a few photos of the outside of the factory. HP isn’t the only company filling shipping containers full of servers to save power and space; Sun Microsystems and Rackable are doing it as well. And IBM and Dell have indicated that they plan to get into the market as companies seek to place computing in remote locations or build out their data centers rapidly.

For the past few days we have been getting pinged by the press folks from Dell who want us to attend a joint event next week with Facebook, to announce a new cloud-computing project. That Round Rock, Texas-based Dell and Facebook of Palo Alto, Calif. are getting cozier shouldn’t come as a surprise. Facebook is seriously “server hungry” and has been on a spending spree to beef up its infrastructure. Dell, on the other hand, has been increasingly seriously about cloud computing, working with online companies and building bespoke solutions for companies like Facebook.

In my conversation with Michael Dell, he said: “Our view is that there is definitely enormous opportunity in cloud infrastructure. A few years ago, we were out there selling our servers and found that some of these new companies had unique requirements that were really different from the general-purpose servers.” Dell has been trying to get closer to Facebook. Dell has worked closely with Joyent to offer a cloud service that offers free services to Facebook app developers.

A few days prior to our Structure 08 conference, I met with Facebook’s VP of Technology Operations, Jonathan Heiliger, and we discussed the issues with current server designs and how today’s startups need a whole new class of machines. I am pretty sure the announcement is along those lines.

The pending news can’t be good news for Rackable, which had been banking on some of the dollars Facebook was spending on its infrastructure. Rackable’s 10-Q filings show that at the end of the second quarter 2008, Facebook accounted for less than 10 percent of Rackable quarterly revenues of around $76 million. A quarter earlier, Facebook contributed about 24 percent to Rackable’s Q1 2008 revenues of $68 million. Rackable is trying to streamline its operations and recently announced that it is divesting its Rapidscale clustered storage business.

Bonus reading: Check out Facebook’s blog where the company explains why it needed to build its East Coast infrastructure. It is a fairly elaborate description of its entire architecture and worth reading.

While it hasn’t yet decided to offer a cloud computing service, Hewlett-Packard today said it will combine its high-performance computing unit with it’s Web 2.0 and cloud computing infrastructure businesses to create the Scalability Computing Initiative, a name that will refer both to a business unit of HP and a set of hardware, software and services tied to scalable computing.

It also followed its competitors and introduced what HP believes will be the building block for the scalable data center, a new, two-in-one blade server. Like IBM’s iDataPlex, Sun’s Blackbox and Dell’s cloud computing efforts, HP is viewing the noise around cloud computing as a chance to sell more hardware — specialized, HP-built 10u racks of 32 blade servers containing 128 cores, to be exact.

I don’t know how important it is to build out scalable computing efforts with IBM’s iDataPlex or HP’s offerings rather than an array of commodity x86 boxes, but the merging of high-performance computing and cloud computing infrastructure is a triumph of the grid architecture running specialized software. It’s also the same trend that is leading Cray to work with Intel on designing the next generation of supercomputers.

HP’s blade servers are designed to save space in the data centers, but Paul Miller, a VP of marketing with HP, acknowledged that space was not at the premium that power is. Blade servers run pretty hot so it’s counterintuitive to think that cramming two of them in one blade makes much sense from an energy efficiency point of view. Miller said HP’s offering can be used with standard HP racks for water cooling or in conjunction with its Dynamic Smart Cooling technology. Without knowing how many watts are consumed, it’s hard to judge how energy efficient these blades are.

Update: HP has provided more information with regard to the server’s efficiency saying it ran in tests at 165 watts per server, and emphasizes that the servers are 60 % more efficient than stand alone boxes in part because they combine two servers into one shell, requiring one fan and power supply for double the compute power.

Fox Interactive Media is one of HP’s clients, so clearly there’s a market for HP’s brand of two-in-one blades, but HP will have to compete with existing hardware vendors such as Rackable and Silicon Metrics, who seem to be doing fine providing energy-efficient scalable hardware for Web 2.0 and cloud computing companies.

Otimização de todo o consumo de energia a cargo de um centro de dados que além da visão estatística dá um panorama geral do ambiente monitorado em 3D.

Objetivo: quanto maior o grau de quantificação de informações, principalmente dentro de uma grande empresa, maior o poder de gerenciamento e tomada de decisões a qualquer momento.

Isso é o que preconiza Scry, um novo produto da Microsoft ao ambiente corporativo. Outros concorrentes de mercado têm também suas vantagens, como o rebatizado Sun MD , Rackable Systems ICE Cube e novo Verari FOREST.

O centro de dados da Microsoft tem um diferencial na sua apresentação: é possível visualizar a infraestrutura física da distribuição do sistema monitorado provendo idéia geral das diversas instalações em visualização tridimensional. O programa permite percorre todos os setores controlados enquanto os dados são aferidos continuamente.

O Scry pode controlar todos os aspectos operacionais de tecnologia da informação dentro de uma empresa.

Itens como consumo de energia e suas variações ao longo das instalações, medição da quantidade de energia gasta com um único produto, por exemplo, para rodar o produto Hotmail dentro da Microsoft. Numa empresa poluidora, o nível de emissão de carbono ao longo do processo produtivo, controle de poluentes e degetos. Picos de energia e as contas derivadas bem como suas previsões, emissão de efluentes, refrigeração de água e controle de temperatura em várias ambientes de trabalho, gasto de energia em servidores, computadores, periféricos etc. Assim, o software reúne todas as informações num banco de dados capaz de trabalhar tudo de forma integrada. O programa pode apresentar o resultado de forma simplificada na sua interface inicial mas também pode pormenorizar analiticamente cada item em questão.

Michael Manos, Diretor Sênior dos serviços do Centro de Dados da Microsoft, apresenta em dois vídeos o novo produto. O primeiro, uma explanação resumida sobre as tarefas que o Scry pode controlar:

Aqui, a imaginação é associada à animação 3D que percorre instalações de uma planta controlada pelo Scry, deixando claro o espaço físico ocupado por cada componente; idéia do controle “ambientalizado”.

Produto recentemente exposto na conferência da AFCOM.

A complexidade da dinâmica do volume de informações num ambiente corporativo de médio ou grande porte tem que estar sob comando de um gerenciador de dados capaz de aferir passo a passo todas as informações operacionais importantes e, por conseguinte, permitir racionalização de energia ou, por falta desta, evitar paralisações, traduzindo-se em economia significativa.

Petrosky

Updated: Facebook these days is doing everything in its power to imitate Google, recruiting the search giant’s sales people, poaching its senior executives and — most importantly — using infrastructure as a competitive advantage. Like Google, Facebook has figured out that the right web infrastructure is the difference between user delight and dismay. And like Google, Facebook is finding out that it isn’t cheap.

I’ve been trying to get a handle on Facebook’s infrastructure for some time, but so far have been unable to get the company to open up. The last time I reached out to them, back in January, I was hearing that they had between 1,200 and 1,500 servers, along with storage and switches from EMC Corp. and Force 10 Networks respectively. As it turns out, those server numbers weren’t even close to the total servers used by them.

The company is running around 10,000 servers, according to Data Center Knowledge, citing comments made by Facebook VP of technology, Jeff Rothschild, at a recent MySQL user conference. (See video of the panel.) Of the 10,000 servers, 1,800 are from MySQL and around 805 of them are memcached servers. In order to house its sprawling infrastructure, Facebook has leased data center space from DuPont Fabros in Ashburn, Va., and Digital Realty Trust in Santa Clara, Calif., DCK reports.

How much is Facebook spending on its infrastructure? The company isn’t going to tell us, but there are clues. Server and storage company Rackable today reported first-quarter 2008 sales of around $69 million. Facebook is one of its largest customers, accounting for around 10 percent of Rackable’s sales (that number could be higher, but we’ll have to wait for Rackable’s 10-Q to get a clearer picture), so some quick, back-of-the-envelope math reveals $7 million in spending by the social networking company. A well placed source of mine just let me know that Facebook is going to spend over $9 million more on servers this year. That should be good news for Rackable. Next on my list is an estimate of Facebook’s bandwidth and data center costs.

The hardware spending by startups like Facebook will be a topic of discussion at our Structure 08 conference, where we are hoping to learn more about the infrastructure secrets of all of today’s top (and fast-growing) web players.

The future of the internet infrastructue is not pizza boxes or Big Iron

Om Malik poses a question: Will the inefficiency of “pizza box” servers push the internet infrastructure back to Big Iron? 

No way. 

What we’re seeing is the creation of a new market, with unique requirements.  What’s needed is Cheap Scale – thousands of servers, scalable at the drop of a hat.  And tons of storage.  

I had a great conversation with Stephen DiFranco at AMD on this exciting space just yesterday.  He sees a classic new market dynamic, where custom solutions have been developed in-house by web hosting firms.  Their technology is their ’special sauce’, but as the industry matures and grows, the solutions mature and standardize as well.  

Vendor solutions are already going beyond the pizza box.  Verari Systems and Rackable Systems have been leaders here, but the solutions are just beginning to evolve. 

Looks like Sun’s Project Blackbox project has a little competition — Rackable Systems has introduced (and is shipping) the Concentro. It’s a data center in a 40 foot shipping container. It’s tailored to Rackable’s server products, with half depth racks and a custom DC power system, but it’s intriguing none the less. With all the hype around “containerized” data centers, I’m pretty sure we’ll see a few others join the fray.

Whatever happened to Google’s rumored container project? Was that the genesis of Project Blackbox, or were all of these companies secretly building these things at the same time?