insideHPC has confirmed that Marc Hamilton has been named by Sun Microsystems to lead Sun’s worldwide HPC sales efforts.
Marc worked with Sun co-founder Andy Bechtolsheim four years ago to architect Sun’s re-entry to the HPC market with Sun’s Galaxy line of x64 servers including the popular “Thumper” x4500 storage server, leading to Sun’s first big HPC win at the Tokyo Institute of Technology. At the time of its deployment in April 2006, the TiTech cluster was the fastest supercomputer in Asia Pacific, and the first Japanese supercomputer to surpass the Earth Simulator on the Top500 list. Marc and Andy then went on to work on the Sun Constellation System used to build the world’s largest x64 based supercomputer in the world at TACC.
For the last year, Marc has been responsible for all Sun systems sales in the Americas as Vice President of Sun’s Systems Practice; he will continue in that role as well as taking on the new responsibility for head of worldwide HPC sales.
“A lot of people think of TACC as a 1-off supercluster”, says Hamilton, “but we have sold over 2000 TFlops of Sun Constellation System clusters to customers around the world and despite the recent financial turmoil around the world, customers continue to invest in HPC systems”.
Marc told InsideHPC he is very excited about taking on the new role of leading worldwide HPC sales for Sun, and says readers should look for a number of new Sun HPC product announcements at the upcoming SC08 show.
I recently had the pleasure of interviewing Robert Graybill, CEO of Nimbis Services. Those of you closely following the HPC news know that Nimbis recently announced a partnership with the Ohio Supercomputer Center (OSC) to broker computational cycles and services. Before you immediately discount Nimbis as just another cycle house, I suggest you read on.
Nimbis was founded by Robert Graybill and Brian Schott. Those in HPC management might remember Robert from his days at DARPA and the US Council on Competitivenes. He spent the previous few years doing studies on the efficacy and utilization of computational resources around the US. He quickly learned that there is a significant gap in what Nimbis terms Digital Analysis Computing (DAC). DAC is essentially a broader term for high performance computing. Any problem requiring computational and storage resources larger than a desktop, is DAC.
Enter Nimbis, stage right. Nimbis has a created an environment and infrastructure in order to provide access to all aspects of Digital Access Computing resources. They’re essentially a clearing house for compute, storage, software (licensing) and consulting resources. Nimbis will provide access to anything from cycles on a cluster, to ad-hoc access to commercial software, to warm-body consulting resources in the form of domain experts. They’re targeting users that don’t typically require access to resources 365 days a year. One may need cycles for a specific project, customer or academic study. In these instances, the initial purchase and maintenance of HPC resources is simply not cost effective.
Business ventures such as this would typically require a huge capital investment for compute, storage and software. Not at Nimbis. Rather than hosting everything locally, Nimbis has ascertained partnerships with the likes of IBM, Amazon Web Services, R-Systems, OSC and Wolfram Research (and the list is growing). This allows Nimbis to remain agile and agnostic to the compute architecture and software payload for which they provide access.
So, how exactly does one get access to these resources? Nimbis has architected a mixture of custom developed software and currently available software packages in order to provide a more fluid user experience. According to Graybill, they will “rely heavily on virtual machine technology.” Nimbis is currently soliciting beta customers in anticipation of a production release in early 2009. For those who would like to see more, Nimbis will be holding demos at SC08 in Austin.
If you would like to read more about Nimbis Services or signup as a beta customer, head over to their website here.
Also posted in Enterprise HPC, HPC
The InsideTrack loves to read email from readers, especially when said email contains information like this.
When I interviewed Cray, Inc. for the HPCwire feature piece on their recently announced CX1, I asked who they were partnering with on the manufacture. I asked, but didn’t expect an answer, because companies rarely answer such questions. And understandably so.
But the InsideTrack is ever curious, and evidently it’s not the only one. An insideHPC reader was noodling around on Google the other day on an unrelated hunt for information, and stumbled upon this URL: cray.vxtech.com. Following that link takes you to the web configurator for the CX1, complete with shopping cart and all. The configurator works well, and covers all the bases, including optional support and a shipping case.
Their house brand product line includes small supercomputers, so they have some experience in the area. The vxtech.com-hosted configurator matches the one on Cray’s site page for page.
Who is VXTech you ask? They are a division of CIARATECH, a Canadian company owned by the Hypertec Group. According to the company’s marketing collateral
VXTECH designs, develops, markets, services, and supports a variety of server systems including NEXXUS4000® Personal Cluster, the acclaimed VXRACK® high density blade server, FUSION1200® SMP Server powered by ScaleMP™, VXR-3DT® Supercomputer Architecture, VXPRO® rack-mount/tower servers, VXSTOR® networked storage and GRAPHIXX® high-end workstations.
The company’s state of the art products are based on the Intel IA32 and IA64 architectures and utilize Linux and Microsoft operating systems.
The existence of all this Cray branding and product information on VXTech’s site implies that either: VXTech is Cray’s CX1 partner, or they were in the running at one time and got really aggressive building an incredibly convincing mockup for their proposal to Cray which they haven’t taken down yet. The latter seems like a worse business practice than letting the CX1 configurator sit out on the web unprotected, so my vote is that VXTech is Cray’s CX1 partner.
The letter linked below was sent from SGI to Linux Networx customers, and the InsideTrack secured a copy through its vast network of industry insiders.
The letter, from SGI Global Services VP Bob Pette, outlines what LNXI customers can expect. A few highlights of interest.
First, SGI’s take on what happened
Today SGI announced the purchase of certain assets of Linux Networx, Inc.. In conjunction with this transaction SGI has offered employment to a number of Linux Networx employees in the Engineering, Sales and Services areas and acquired LNXI’s spare parts inventory.
And then what it means for LNXI customers (note the honesty):
SGI did not acquire Linux Networx’s service contracts and as such, does not have a service contract in place with you. …During this period of transition, please continue to place your LNXI service requests as before by calling 1-800-459-7138 or online at http://support.linuxnetworx.com.
Very refreshing. Here is the whole letter as a PDF.
Also posted in HPCAnswers
Lots of interesting gossip and rumors floating around these days about Linux Networx and something big in the air for them this week.
I’ve been pumping my friends and irritating those I don’t know very well for the past 24 hours trying to find a quorum opinion on exactly what, and so far I don’t have one.
What I have found is reflected in a comment on this post left by an insideHPC reader on the web site today:
This week, they are going through some BIG changes, sell off/Merger or even a Complete shutdown [capitalization courtesy of the commenter]
I hear the same things. On the sale I hear that possible buyers include one of the “smaller” HPC companies, outfits like Rackable or Penguin.
I feel obligated to remind you at this point in the post that I have a terrible record with rumors. But then, so does everyone else.
John Leidel posted earlier today about the Linux Networx diskless system, but an industry insider tipped me off to a little inconsistency. There’s an interesting nugget on the interwebs today about the company hidden in this release.
If you look at the press release for this announcement on the company’s site, you’ll see it begins this way
SALT LAKE CITY, December 17, 2007 – LNXI (Linux Networx, Inc.), an award-winning provider of Linux-based…
But if you read the same release at Marketwire you’ll see this
SALT LAKE CITY, UT–(Marketwire – December 17, 2007) – LNXI (formerly Linux Networx, Inc.), an award-winning provider of Linux-based…
Note the addition of the word “formerly.” It appears that Linux Networx is at least considering a name change.
Companies tweak press releases all the time, and there are several other wordsmithy changes in this release. But none of them have the ramifications of this particular change.
The DailyTech has had several posts over the last few days chock full of tidbits from leaked memos and shady sources. Good stuff! I’ve gathered three here so you can amaze and astound your less-informed friends
First up, the DT has a leaked memo with details on the Phenom CPU and motherboard availability.
More details of the AMD platform launch that included the Phenom, the RD790 chipset and the RV670 graphics processor surfaced early in November. DailyTech reported early in November 2007 on the pricing structure of the Phenom processors ranging from about $280 to $330 USD.
An AMD internal memo shows the AMD Phenom 9700, 2.4 GHz processor is slated for mid-December availability. Allocation for the Phenom 9600 running at 2.3GHz has been pushed back to a Q1 2008 date.
Up next, also from AMD, are details on the tri-core processor scheduled for 2008 and 2009
AMD picked up big headlines the day before Intel’s Fall Developer Forum with the announcement of its upcoming tri-core processors.
AMD’s original release did not specify if this tri-core processor, code named Toliman, would be a totally new processor or merely a stripped-down version of the existing Agena core. The answer, it appears, is both.
And to round it out, we have leaked info on the forthcoming Intel Penryn mobile processor
Initial Penryn mobile processors will launch exclusively with a 35W thermal envelope. Existing Merom mobile processors (T7000-series) also fit inside the 35W envelope at standard voltage. However, with the Montevina Centrino refresh in the second half of 2008, new Penryn mobile processors will receive a 25W TDP rating instead.
Looking further into the future, Q3 will see the addition of another group of 45nm Penryn-based processors using Intel’s Montevina platform. These processors will feature 6MB of L2 cache in Core 2 Duo form and 12MB of cache for Core 2 Extreme models — both will feature 1066MHz FSBs along with support for DDR3 memory.
Dave Raffo at SearchStorage.com is reporting that BlueArc will hold off on its IPO until sometime in 2008. The NAS vendor filed for its public offering back on September 7.
According to Raffo:
A financial analyst who follows storage said investors “are doing more due diligences on IPOs now,” and said it could hurt BlueArc that the stock prices of storage systems companies Isilon and Compellent have dropped drastically since their recent IPOs. The analyst said BlueArc might want to show another quarter of solid growth to help its case.
Read the full blog entry here.
The InsideTrack was recently pointed to Nyquist Capital’s blog, and their speculation that Google is building their own big switches.
It is our opinion that Google (GOOG) has designed and deployed home-grown 10GbE switches as part of a secret internal initiative that was launched when it realized commercial options couldn’t meet the cost and power consumption targets required for their data centers.
How, you might ask, do they come to this conclusion? By following the parts, of course.
We were watching shipments of SFP+ components for 10GbE in the market but simply couldn’t account for their end destination – sort of an optical component dark matter problem.
Which leads them to this opinion
Through conversations with multiple carrier, equipment, and component industry sources we have confirmed that Google has designed, built, and deployed homebrewed 10GbE switches for providing server interconnect within their data centers. This is very similar to Google’s efforts to build its own server computers (excellent article here). Google realized that because its computing needs were very specific, it could design and build computers that were cheaper and lower power than off the shelf alternatives. The decision to do so had a profound impact on server architecture and influenced the market’s move to lower power density solutions that Sun (JAVA) , Intel (INTC) and AMD (AMD) now embrace.
Reader Paul Adams points the Inside Track to The Register’s Hardware site and news of Intel’s forthcoming 45nm chip, codenamed Diamondville. The news comes from Chinese language documentation posted by hardware manufacturer Asus, who plans to use the chip in a revved version of its ultra mobile Eee PC.
The chip is positioned as a low cost PC platform, so it’s not for HPC, but it’s worth at least noting.
Diamondville can operate without a fan – a cost-cutting measure as much as an attempt to get Shelton into small form-factor PCs – and will be soldered to the motherboard. Boards will be based on either Intel’s own 945GC integrated chipset or SiS’ SiS671.
And here’s the clincher that marks Shelton down almost certainly as the desktop Eee PC’s foundation: it’s designed to work with 2-4GB of Flash storage. That said, the chipset will also handle parallel ATA and SATA devices, and other peripherals through six USB ports. Early board pictures show a single DIMM slot and one PCI connector.
An HPC industry insider pointed the Inside Track to an interesting bit of news last night about SGI.
As the San Jose Mercury News reported yesterday (here), shareholder Southpaw Asset Management wants SGI’s board to find a buyer. Southpaw holds approximately 5.3% of SGI’s common stock, and disclosed their letter to SGI’s board in a filing with the SEC (which you can read here).
What’s got Southpaw’s knickers in a knot? SGI stock’s plummet, from a high of 30 this year to a low of 14.75 just a few days ago. From the letter
…the significant decline in the stock price over the course of the past ten months, coupled with the Board’s “business as usual” strategic direction, compels us to become more active. …For the Board and management to continue to take comfort in a turnaround story that is not being reflected in the stock price is unacceptable.
What does Southpaw want? A sale to another HPC company
The benefits to SGI’s competitors are substantial and encompass the expansion possibilities from the Company’s installed base and best in class technology, in addition to the synergies that could be realized. Anyone of the Company’s competitors could significantly increase revenues from SGI’s products and services through their substantially larger sales platforms, installed bases and superior resources. Furthermore, much of the Company’s SG&A could be cut by a strategic buyer to create substantial returns without any organic growth. With more than $150 million of annual SG&A, the
cost cutting opportunities are substantial for a strategic buyer and make SGI an excellent acquisition candidate on that basis alone.