Found an interesting piece out of the New York Times last week while catching up on my feeds:

The New York Stock Exchange is investing heavily in x86-based Linux systems and blade servers as it builds out the NYSE Hybrid Market trading system that it launched last year. Flexibility and lower cost are among the goals. But one of the things that NYSE Euronext CIO Steve Rubinow says he most wants from the new computing architecture is technology independence.

This is a semi-important story for open source in my opinion. Lots of companies have Linux servers running here and there, but in many cases these are the result of pilot projects running non-critical applications. It's nice to see an organization so synonymous with high performance, mission-critical technology drinking the Kool-Aid.

The article goes on to say:

One technology that the NYSE isn't adopting so eagerly is server virtualization, which comes with a system latency price that Rubinow said he can't afford to pay. In a system that is processing hundreds of thousands of transactions per second, virtualization produces "a noticeable overhead" that can slow down throughput, according to Rubinow. "Virtualization is not a free technology from a latency perspective, so we don't use it in the core of what we do," he said.

I worked on a trading floor system in 1999-2000, and given what I know about the architecture, its debatable whether these types of applications would make them a good candidate for virtualization anyway. It's far more likely that they'll be running any given trading application on a bunch of blade servers in parallel, and since a lot of these systems are more or less stateless, they'll simply load balance requests across the pool.