Attendees at the Supercomputing 2004 (SC2004) conference held earlier this month in Pittsburgh got a view of the latest in high-performance computing hardware and storage systems.
Most announcements were aimed at the general high-performance computing audience. However, most of the new products introduced at the conference would find use in any scientific computing or visualisation application environment and thus would be of interest to life science organisations that do genomic research, computational biology, or modelling of biological systems.
All of the major systems vendors, including Apple, Cray, Dell, IBM, HP, NEC, Sun Microsystems and SGI were at the conference showing their latest high-end systems. Among the notable trends were more higher-density 64-bit rack and blade server systems; many new dual, quad, and eight-way AMD Opteron systems; and increased support for Intel's Extended Memory 64 Technology.
Additionally, many of these same systems vendors and some new players such as Orion MultiSystems showed off the latest in high-powered workstation technology. Many of the products in this category are aimed at scientific computing and visualisation applications that are commonly used in life science research.
New systems on display at the show included:
RLX Technologies announced the SB6400, a blade server based on dual Intel Xeon processors. The SB6400 supports Intel's EM64T and is a sixth-generation RLX blade server that includes an 800MHz front-side bus, DDR2 memory and PCI Express support. RLX also announced two new 1U form factor servers, the RM1400 and RM1100, both of which support EM64T.
Microway announced the Navion-8, an SMP server that supports up to eight AMD Opteron 800 Series processors. The eight-way system is designed to run 64-bit compute-intensive applications, including scientific modelling and high performance database applications. The Navion-8 can accommodate five 3.5-in. SCSI or Serial ATA drives, and includes Microway Cluster Management Software (MCMS), the NodeWatch remote monitoring and control tool, and optional MPI Link-Checker Software.
Appro International unveiled its Appro XtremeBlade System, a blade server system that includes InfiniBand interfaces to external data and storage networks. An Appro XtremeBlade Cluster supports up to 72 blade servers in a single rack unit. The system is comprised of six sub-racks that house the blade servers. Users can combine two-, four-, and eight-way AMD Opteron-based servers in these sub-rack units. The XtremeBlade System, which will begin shipping in the first quarter of 2005, can include an integrated InfiniBand or Gigabit Ethernet switch.
One of the main themes of this year's conference was high-performance storage. In many of the conference sessions and on the show floor, there was growing recognition that the performance of storage systems needs to be taken into account in today's high-performance computing environments.
Many vendors, including 3Pardata, Archivas, BlueArc, EMC, IBRIX, Network Appliance and Storage Technology, used the conference to show their high-end storage systems. And some vendors used the conference to announce new offerings.
New systems on display at the show included:
Panasas announced the next generation of its Panasas Storage Cluster Platform. Elements of the announcement include the Panasas ActiveScale File System 2.0, which is a new release of the company's file system, new hardware bundles, and expanded services to help companies deploy and manage a Panasas Storage Cluster. New features of the ActiveScale File System include enhanced load balancing, file locking for improved application compatibility, parallel data access to speed data backup and restoration, and a media monitoring feature that analyses disks for high reliability.
Systems vendor Linux Networx introduced Xilo, a scalable storage system for its cluster systems. Essentially, Linux Networx is offering its cluster users an easy-to-manage, integrated storage system. A Xilo system can scale from 6TB to 224TB of usable storage. The system can support Dual Gigabit Ethernet, Myrinet, or InfiniBand high-performance interconnections.
Isilon Systems introduced OneFS 3.0, a new version of its distributed file system. One key feature of OneFS 3.0 is FlexProtect-AP, a data protection technology that can rebuild a failed 250GB disk in as little as one hour. Isilon claims that's faster than the time it takes to do a similar rebuilding of more costly Fibre Channel disks.
Additionally, FlexProtect-AP monitors and preemptively migrates data off at-risk components. Isilon notes that OneFS 3.0 tightly integrates with its SyncIQ replication software, which allows users to quickly make disk-based copies of critical files on other Isilon IQ clusters.