HP Polishes Its Clusters
Adds to its clustering story at the International Supercomputer Conference in Germany
June 21, 2005
Hewlett-Packard Co. (NYSE: HPQ) will unveil the latest additions to its long-term clustering strategy at the International Supercomputer Conference in Heidelberg, Germany, on Tuesday.
The vendor is fleshing out its Unified Cluster Portfolio (UCP), a pre-packaged collection of hardware and software launched last November. This included Integrity servers and HPs StorageWorks Scalable File Share (SFS) technology (see New HP Clusters Simplify HPC).
With Tuesday's announcement, HP will add a range of dual-core Advanced Micro Devices (NYSE: AMD) Opteron processor-based servers to the UCP mix, starting with its Proliant DL585 device. Ed Turkel, marketing manager for HP’s high performance computing clusters, tells NDCF that two more Opteron-based servers, the DL145 and DL385, will join the family later.
The exec also confirms that HP will be adding blade servers to the cluster. HP’s BL35p Opteron-based blade server is first in line, says Turkel, and will be available sometime in August. He adds that HP is also looking to add its BL25p and BL45p blades into the mix, although this is “dependent on customer input.”
HP is not the only company polishing its clusters. Last week IBM Corp. (NYSE: IBM) announced a raft of new technologies for its pre-packed 1350 cluster. Rather than deploying clusters piece-by-piece, the idea is that users can buy a pre-tested set of cluster technology from an established vendor (see IBM's Cluster Bluster).For both IBM and HP, blade technologies are at the forefront of their clustering strategies. “Blades are a rapidly growing part of the high-performance computing market,” explains Turkel. “They provide simplified management, scaleability, and performance,”
HP will also add to its StorageWorks SFS technology for building Linux clusters. Basically, SFS is a data management technology for sharing files across clusters of servers and storage devices.
Version one of SFS was launched at last year’s International Supercomputer Conference. It offered up to 12 Gbytes of throughput per second and 256 Tbytes of storage capacity. The new version will offer more than 35 Gbytes of throughput and 512 Tbytes of capacity, according to HP.
Some in the industry say that, overall, IBM’s clustering strategy is more tightly focused on high-performance computing than its rival HP's, although Turkel disagrees with this. The exec points to a list of organizations that have so far signed up for HP’s SFS technology, which includes the Sharcnet research network in Canada, the Center for Computational Sciences at the University of Kentucky, and Indiana University.
Nonetheless, Turkel like his counterparts at IBM, believes that clusters are now starting to break out of their supercomputing niche, driven in part by more clustering applications written for popular enterprise database platforms such as Oracle Corp.'s (Nasdaq: ORCL). ”We certainly see cluster technology showing up in some non-traditional high-performance computing environments, such as financial services,” he says. “The type of thing that will flick the switch in the enterprise is Oracle."— James Rogers, Site Editor, Next-Gen Data Center Forum
You May Also Like