Spreading the Low-Latency Love

After speaking with a number of buy- and sell-side CIO recently, one popular project on nearly everyone's plate is expanding the benefits of low-latency infrastructure to other parts of the enterprise. When prodded for details on where they saw the non-algo trading benefits, no one really would cite specifics.

My initial thoughts where that the industry might adopt more event stream processing into the data management space and replace the current extract, transform and load (ETL) tools that most firms have in place. But speaking to a couple of vendors on the topic, they're seeing a few clients doing this, but it currently is more the exception than the rule.

While the industry continues to bleed out latency when it comes to execution, we've seen a similar speed revolution happen in batch processing. It's doesn't have to do with changing messaging protocols or co-location, but the adoption of grid and cloud computing's parallel processing capabilities. Include other technological innovations like graphical processing units (GPUs) and multi-day calculations have been reduced to hours and overnight calculations have been reduced to minutes.

Minutes might not sound like they're in the same league and microsecond order acknowledgements, but for some processes in the trade lifecycle it is fast enough. If you look at the auction phase for the work up of a single trade of an electronically traded over-the-counter (OTC) instrument, could last more than a minute. But a minute in the equities, futures or foreign exchange (FX) is a life time.

To reduce latency in the middle and back office, I've a feeling that it's going to be more about improving batch performance than ripping and replacing it with streaming architecture for the most part.