As we near the end of 2011, we take a moment to reflect on the past year. It’s been a busy year for IT across virtually all verticals, from mobile and search to enterprise servers and cloud computing. When we attended HPC360 a few weeks ago, we had the pleasure to attend a keynote presentation by Addison Snell, CEO of Intersect Research in which he discussed the most important trends in high performance computing (HPC).
HPC is an exciting and growing industry that ICC has been moving into the past couple years. The traditional HPC space revolved around high-end research facilities particularly in science and engineering. However, with each year technological innovations and tailored systems such as our Supermicro GPU Simcluster have brought the realm of HPC closer to reality for many small/medium-sized business and organizations.
In this 2-part series we will look at the top 10 future trends in HPC from Intersect360’s research, coupled with our own analysis and thoughts. No better way for us computer nerds to close the year right? Let’s get started.
Top 10 HPC Trends for 2012 and Beyond
1. Accelerator adoption continues
We first heard about this in 2010 with increases in GPU computing, the release of CUDA 3.0 and OpenCL 1.1, and the forming of a consensus that co-processing accelerators were needed to play an integral role in approaching the exascale. Mr. Snell reported that about two-thirds of users were testing accelerators, up from from only 20% the year prior.
Today, NVIDIA holds about 90% of the current accelerator usage. Although AMD has entered the HPC space with its “APU Fusion” line, NVIDIA sees its real challenge as Intel, who plans to move into the GPU markets with MIC next year. MIC (pronounced “Mike,” standing for Many-Integrated-Cores) is based off of Intel’s discontinued Larrabee technology. Their first expected line is the much anticipated “Knights Corner.” But while Intel virtually dominates in the realm of processors, NVIDIA’s foothold with GPU technology, especially with the recent release of the M2090 Tesla series, will be sure to give Intel a run for its money.
2. Storage is the fastest growing product sector
Storage consolidation is intensifying competition among providers, and while it remains one of those things that “always gets cheaper” in the long-run, in the short-run we are of course faced with the consequences of the recent flooding in Thailand. The impact of this could potentially be that solid-state drives start to edge their way into the HPC space.
However, undoubtedly the most important storage aspect of HPC is the stabilization of the Lustre roadmap and parallel distributed file system made for large-scale cluster computing. Previously spearheaded by Sun Microsystems, Oracle gained possession of it during their acquisition of the former company. For a while there were fears that Oracle would abandon the technology. However, after much voicing and support from the open-source community and HPC startups, it now appears that Lustre will be moving forward in tandem with the growth of HPC.
3. The Uncertain Growth of Infiniband?
Infiniband has become the dominant technology for HPC, despite Cisco’s abandonment of the technology in 2009. With its premium on pure performance in environments with high interconnectivity requirements, it is almost without question the superior option to its competitors. That said, from a bigger picture, Infiniband remains a very small fraction of the traditional Ethernet technology. Indeed, one of the reasons Cisco dropped it was due to the declining costs of 10 Gb Ethernet.
Infiniband will inevitably grow with HPC, but in order to truly flourish it needs to find new usages and markets. And if it does, the question still remains over whether it will displace Ethernet as a dominant technology. For us, we deal primarily in Infiniband, but many distributors out there still deal exclusively in fibrechannel and ethernet. That said, providers of both are starting to look at ways of merging the capabilities of each together. That means a very interesting future for networking technology.
4. The Biggest finance application is NOT trading, but rather RISK MANAGEMENT
Perhaps because of the economic crash and the bailouts, the biggest use for high-end financial computation applications is now in risk management (trading remains the second largest.) As regulation continues to come down hard on banks, we should expect to see more of this in at least the short-term. In addition, with a majority of HPC-using banks comprising the high-end of the financial institution scale, smaller, local, and regional institutions who could benefit from HPC are currently going unserved. They represent an opportunity for growth in the HPC space.
5. Facilities continues to increase share of budget
Due to the high energy costs for powering and cooling IT systems, storage and network hardware now represents the highest operating expense for a data center. And there are no signs that it will change in the near future, with a projected 19% increase in the rate of data center energy consumption in 2012. Moreover, the sheer weight of racks adds additional logistical costs to owners and managers.
While advances in green technology and innovative environmental approaches (like Facebook’s) may help to offset these costs, it is uncertain how or when these will come into play. Moreover, such technology will almost certainly come at a higher price than existing units, so the budget will have to reconcile initial outlays with expected future savings.
Whatever happens in this area, the growing processing capabilities of computer clusters, GPU Simclusters and other supercomputing technology make energy efficiency one of the top priorities for HPC in 2012.