Appro, which recently signed a contract to build three high-performance computer clusters for the U.S. Department of Energy, is eyeing a larger portion of the commercial supercomputer market for itself.
While not as well known in high-performance and supercomputing circles as IBM, Cray or Hewlett-Packard, the Milpitas, Calif. company has been trying to expand its HPC (high-performance computer) business beyond government contracts. Its targets are enterprises and even mid-market companies that need to run data-intensive workloads. At the same time, Appro is looking to show that its clusters can deliver the same amount of computing power as some of the other well-known supercomputers, such as IBMs Blue Gene/L systems.
To help make its case that it can compete again bigger vendors, Appro announced this week at the 2007 International Supercomputer Conference, in Reno, Nev., that it would expand its Xtreme-X Supercomputer series of HPC clusters with new blades. The first of these offerings is the Xtreme X1 dual-socket blade, which will be based on Intel quad-core Xeon processors, uses a Linux operating system and will be available to order later this year. A version based on Advanced Micro Devices chips, the X2, is scheduled for release in 2008. For the first time, the company is also embedding new management software to help better control and allocate resources.
While Appro is continuing the build out its HPC technology, John Lee, vice president of Appros Advanced Technology Solutions Group, said the key to the companys future growth is successfully designing HPC clusters that can scale for businesses that need massive computing power for a variety of large workloads, such as scientific applications.
Click here
to read more about Appros workstation business.
“The concept is [being] able to replicate the machines faster, so that each time we do a deployment, we get better at it,” Lee told eWEEK. “By the time we add the third or fourth cabinet to the cluster, we can deliver it all in six to eight weeks. Its really like building with Lego blocks. If one customer wants a cabinet with six teraflops of computing power, they can get it. If they want to scale to a petaflop, then we can add 150 cabinets.”
While building high-performance clusters is not new-406 of the worlds top 500 supercomputers are designated as cluster architecture systems-Lee believes that Appro can offer customers a better price and better total cost of ownership for small enterprises and mid-market companies compared to other offerings. HP is also expanding its HPC reach with a similar outreach to mid-market businesses.
Steve Conway, an analyst with IDC, said Appro uses commodity parts to build its clusters, which reduces its research and development budget and allows it to build HPC systems for less money. The recent $26 million budget Appro signed with the DOE shows that the company not only can deliver the right amount of computing power, but it can also deliver that power at a reasonable price.
“In a nutshell, Appro has become very good at building these types of big clusters,” Conway said. “While you dont have all of the features you would get with one of the more high-priced systems, you do get much of the same computing power for a much more affordable price.”
Since it can offer an HPC cluster at a lower price, Lee believes that several types of enterprises are ready for Appros Linux-based clusters, including the oil and gas industry, automobile manufacturers and semiconductor makers. There is also room to expand into the mid-market.
Conway said IDC reports show the supercomputing and HPC market is poised to grow about 9 percent between 2006 and 2011, and that much of that growth will be at the low-end of the market. Smaller companies, Conway said, are more open to using HPC to solve a number of problems.
Cray has introduced a pair supercomputers this year.
Click here to read more about the announcement.
“A lot of companies now have experience with HPC,” Conway said. “What Appro and HP are doing is developing turnkey systems that are can be built, delivered and are ready to go. Theres also price. About 10 years ago, these systems cost $10 or $20 or $30 million. Now, you have systems at the low-end that start at $25,000 and some even lower than that.”
Appros other goal, making its way up the Top 500 supercomputer list, could prove more difficult than expanding into the mid-market. When the new list was released Nov. 12, Appro had four systems listed in the Top 500, while IBM had 232 systems listed and HP had 166 supercomputers on the list.
Lee said Appro is looking to make up some of the ground next year with some newly installed clusters that could put the company as far up as the Top 10.
“We only had four clusters in the Top 500 and in the last few years, we have had some knocked off,” Lee said. “When the list launches again in June, I think one of our clusters will reach the Top 10 and this has been a real big push for us and one of the companys goals.”
Check out eWEEK.coms for the latest news, views and analysis on servers, switches and networking protocols for the enterprise and small businesses.
