|The Leading Source for Global News and Information Covering the Ecosystem of High Productivity Computing / June 30, 2006|
At IDC's annual analyst briefing at ISC 2006, Research Vice President Christopher Willard, Ph.D. said clusters became the top revenue generator in the first quarter of 2006, garnering 52 percent of the high performance technical computing (HPTC) market. He predicted cluster growth would continue at a compounded annual rate of 13 percent through 2010, reaching $11 billion in that year and settling out over time at 75 to 85 percent of HPTC market revenue -- unless a major new disruptive technology comes along.
Cluster revenue growth has averaged more than 90 percent over the past four years. Cluster market share growth has been near-linear since 2003, according to Willard. In 2005, there were 20,000 cluster units shipped in both capability and capacity market segments.
Price/performance for throughput workloads continues to be the biggest driver for purchasing clusters. "Clusters are being used for both capability and capacity computing, and clearly are the best solution for embarrassingly parallel problems," he said.
Based on a study IDC completed in January 2006, the top system requirement for clusters is total system performance, but aggregate I/O performance has also become very important, Willard said, noting that people are having problems balancing system computational performance with I/O performance. High availability is also showing up as an issue. Node to node latency and bandwidth are not seen as major issues. IDC study results indicate very strong growth in storage requirements over the next five years, from a median level of 27 terabytes today to 500 terabytes in five years. Cluster buyers are also starting to ask for some specific premium technologies in clusters.
Willard said he expects to see new technologies move into clusters over the next few years, include accelerator technologies and new architectural approaches. These innovation could send the market in new directions.
By way of guidance, he said clusters are becoming a standard method for managing workloads in technical computing centers. Users are willing to pay premiums for superior system integration, ease of use and certain other capabilities. Vendors should look to the grid managed computing environment as the potential next step beyond clusters. To survive and thrive in the highly competitive cluster market, they should not abandon technical excellence.