NT Spawns Cluster Products

Learn about the cluster products that other vendors offer.

Barrie Sosinsky

October 4, 1999

3 Min Read
ITPro Today logo in a gray background | ITPro Today

Through a recent merger with Sequent Computer Systems, IBM aims to consolidate the company's lead in large clustered Windows NT systems. When a vendor controls the hardware and the software, tight integration between the two is possible. For example, IBM has large clustered solutions on both the RS/6000 and AS/400 platforms. However, now several companies are coming out with new products that will provide competition for IBM.

In June, Sun Microsystems introduced its High Performance Computing (HPC) ClusterTools software that can cluster 16 Sun Enterprise 10000 (aka Starfire) servers as one image. Each Starfire server contains 64 UltraSPARC-II RISC hot-swappable processors. Sun's solution can create systems in the supercomputer range, and the company hopes to get some larger customers, such as Sabre with its reservation system, to migrate legacy applications to these high-end UNIX systems. The industry will measure the Windows 2000 (Win2K) cluster against these UNIX systems and the upcoming Wildfire systems from Compaq.

Novell has StandbyServer for NetWare, a product based on Vinca's StandbyServer, that you can configure as either a one-to-one or a many-to-one cluster. In a one-to-one cluster, one server mirrors the other server, but in a many-to-one cluster, one spare server has a disk that mirrors the other members. Novell also offers the Novell High Availability Server (NHAS), which is a 2-node shared-nothing cluster similar to Microsoft Cluster Server (MSCS).

By far, Novell's most interesting product is Novell Cluster Services, based on the Orion Phase II project. Novell Cluster Services is a multinode solution that provides resource sharing, integrates with Novel Directory Services (NDS), and has manual load balancing that Novell's ConsoleOne manages (on an application level). At PC Expo in June, the Novell booth demonstrated Novell Cluster Services in 12-node form. This fall, Novell expects to introduce Novell Cluster Services with as many as 32 load-balanced nodes. According to Michael Bryant, Novell director of Advanced Services Marketing Group, the level of interest in the company's clusters is so high that Novell expects more than 80 percent of its customers to implement clusters.

Other vendors have also announced new cluster offerings. Network Engines is shipping WebEngine for NT and Linux. WebEngine is a thin server with ClusterControl management software. Data General and Dell have licensed NuView's ClusterX software to manage Microsoft Network Load Balancing (NLB) and 2-node MSCS. FullTime Software released FullTime Cluster 4.2 for NT and HP-UX on Alpha this past April. Stratus Computer plans to release a next generation Continuum II fault-tolerant server next year for Win2K and NT.

Some very large NT cluster implementations are beginning to appear, particularly in university environments where price for performance is a key consideration. One of the largest implementations came online at Cornell University's Cornell Theory Center (CTC) in early August thanks to the Advanced Cluster Computing Consortium (AC3) that Microsoft, Dell, Intel, and Giganet make up. The AC3 Velocity Cluster contains 256 processors in 64-quad Dell PowerEdge servers connected by Giganet's high-performance cLAN host adapters and cluster switches. The AC3 Velocity Cluster has the power of a supercomputer at a fraction of the cost, and researchers at Cornell University will use the cluster to solve scientific research problems. For more information about the AC3 Velocity Cluster, see "World's Largest Windows NT Cluster Goes Live," http://www.winntmag.com/articles, InstantDoc ID 7149.

The release of so many NT clustering products shows the growing vendor interest in NT clustering. At Microsoft's cluster lab, engineers are walking five companies per week through compatibility testing. Ultimately, the success of clustering depends on the availability of applications that run on it. When clusters gain an industry-standard API that runs on commodity hardware, vendors will be able to write applications that will cause an explosion of products in this area, and the technology will become widespread.

Sign up for the ITPro Today newsletter
Stay on top of the IT universe with commentary, news analysis, how-to's, and tips delivered to your inbox daily.

You May Also Like