Cisco Blogs

Unified Computing System: Ferrari Borrowed Our Codename

- April 17, 2009 - 2 Comments

As mentioned on the IPTV broadcast yesterday, the preliminary benchmarking for out new blade servers for our Unified Computing System is pretty darn good–something along the lines of 164% faster than previous-gen Intel-based two-socket systems. I think this not only makes a clear case for the upgrading to Intel Xeon 5500 Series processor but, the same way you would not put a Ferrari engine in a Cavalier, you also want to upgrade to a system that is designed to take advantage of that kind of performance, not just retro-fitted to deal with it. Here is a rundown of our preliminary results for some key industry benchmarks that cover a variety of workloads: image

The VMware VMMark® benchmark indicates a system’s capacity to scale when running virtualized environments. Preliminary benchmark results using a release candidate of VMware next version of ESX show a score of 24.14 running 17 benchmark tiles, an improvement of 164 percent over the prior top-scoring two-socket system based on previous-generation Intel processors. Cisco UCS performance on the SPECfp® rate_base2006 benchmark is 194, showing high performance for multiple floating-point workloads running in parallel, and demonstrating an improvement of 125 percent over previous-generation systems. Similarly, a SPECint® rate_base2006 result of 239 demonstrates high performance on integer compute-intensive workloads, delivering an improvement of 71 percent over the prior top-scoring system. The SPECjbb® benchmark demonstrates performance on Java software workloads that place intensive multithreaded workload demands on systems, indicating performance on multi-tier Web server environments and Web 2.0 applications. Cisco UCS performance of 556792 demonstrates breakthrough performance on Oracle JRockit running on the Microsoft Windows 2008 operating system. Cisco UCS B-Series Blade Server performance extends beyond enterprise applications and into high-performance scientific and engineering workloads. The SPEComp® MBase2001 benchmark is designed to test the limits of shared-memory, symmetric multiprocessing systems. The Cisco UCS B-Series Blade Server score of 43593 (currently under review) demonstrates industry leadership along with a performance increase of 154 percent over previous-generation systems. The key thing to point out is that while we are providing Ferrari performance, we are doing it with a Prius energy footprint. As I noted in my recent conversation with Intel’s Ed Groden, we see up to 9:1 consolidation with the Intel Xeon 5500 processors, so 184 single core servers could be collapsed into 21 Intel Xeon 5500 Series systems, cutting energy costs by 90%.

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. Nik:Agreed. Right now, we have comparisons to the existing benchmark winners. I would expect the topic of benchmarks to be an on-going conversation.Omar

  2. Saying you are faster than previous generation processors is hardly surprising. It might have been worth mentioning that your results are better than other Nehalem-based servers, not by much, but they are rack-mount systems, so having a blade beat them is quite something.