The need for speed in trading has not abated. Technology that minimizes latency is simply the "greens fee" to trade in liquid markets today. Without it, you just can't play.
Even though trading firms and venues have made tremendous progress over the past decade in reduce distances and maximize speeds over the distances that remain, the game has not ended. It has just evolved. In particular, as trading venues focus on minimizing the variability of their latency (in order to adhere to FIFO-based fairness policies), trading firms find that the return on investment investment increases for ever smaller latency improvements. Thus, trading desks at the leading edge of the latency race now care intensely about nanoseconds and are even beginning to measure some legs within their architectures in picoseconds. And for desks that don't rely on being first but can't afford to be last, applications must process information in just a few microseconds to stay competitive.
While many factors affect the end-to-end latency of a trading system, the network and the interaction of trading applications with the network are crucial. That's why the STAC Benchmark Council continues to focus on these issues. Advances in network technology continue to flow from established players, while new entrants drivie a significant amount of re-thinking.
The STAC Network I/O SIG has established benchmark standards for complete host network stacks including software and hardware (STAC-N1, and the proposed STAC-T0), oriented around applications that access the network API directly without intervening middleware. Other tests focus on switch performance. And the STAC-TS Working Group looks at the accuracy, capacity, and latency impact of network-level timestamping and capture solutions. (Some of the materials are reserved for paying members of the STAC Benchmark Council. For information on how to get access, click here.)