Looking to quench the never-ending need for data center capacity and bandwidth, Hewlett-Packard Co. and IBM Corp. are separately rolling out monster servers capable of handling the needs of high-end dot-com and enterprise customers.
HP this week announced its high-end HP 9000 Superdome server, which sports a number of mainframe-like features, such as partitioning. Partitioning will let users slice up the server to handle multiple functions or consolidate complex tasks to a server cluster - reducing the amount of management, configuration and administration associated with distributed network servers. For example, the HP 9000 Superdome can be split into partitions that might handle separate Enterprise Resource Planning and SAP applications.
Using resource partitioning capabilities, network professionals can use scripts or templates to carve up CPUs and establish policies based on how those processors should be used, says Brad Day, a senior analyst at Giga Information Group, in Cambridge, Mass. He says those policies can let customers allot how much CPU power they want to give to certain users and applications.
HP's Superdome systems are available now, according to Nam LaMore, a spokesperson for the company, with configurations starting at about US$400,000.
HP will also begin offering the Superdome based on Intel's IA 64-bit processors by the end of 2001, with support for Windows 2000 and Linux. Superdome now runs on HP-UX 11.i.
Day says with the announcement of Superdome, HP has a leg up on competitors Sun and IBM because it can also allow for disk I/O partitioning - which lets network managers assign disk I/O resources to specific groups of users and applications. Virtual partitioning allows for moving of applications and resources between partitions. For example, if a network manager discovers that one partition isn't enough for a workload, he can move them as workloads change through an on-the-fly management console.
For its part, IBM announced Blue Hammer, which essentially extends the management capabilities of the company's RS/6000 SP supercomputer to its enterprise Unix offering: the RS/6000 S80 - which supports clusters that range from two, 12-way systems to 16, 24-way systems. By porting its Parallel System Support Program (PSSP) management software, originally designed for the SP to the S80, IBM provides a set of tools that network managers can use to monitor and control all servers on the cluster from a single console. PSSP software is a powerful set of management tools that allows for configuration and other network server performance tasks to be done from a single console.
IBM's S80 cluster with PSSP software is available immediately. A base configuration of two 6-way clustered S80s - 12 microprocessors - begins at about $705,000.
IBM says it will have support for PSSP on its midrange systems - the M80 and H80 - by the end of 2001.
High-end Unix servers have become popular packages for serving up e-commerce, Internet and Web-related businesses. Sun is considered the market leader in this realm with HP and IBM in second and third place respectively. Published reports indicate Sun may try to one-up HP and IBM in the next few weeks with its oft-delayed UltraSPARC III chip-based workstations or servers.
IBM: www.ibm.com; HP: www.hp.com