One silver lining in the current gloomy environment, with its reduced margins and increased regulatory burdens, is the increasing appetite for the industry to work together to adopt common standards that lower costs for all participants. I have been thinking recently about just how far this process could go in the trading technology space, and have become convinced that there is tremendous room for additional harmonisation.
In this blog, I am going to take one simple example, related to common technical standards for distribution of market data to illustrate the potential:
Today, there is some consensus that the FIX Protocol can be used for market data distribution, supplemented by FIX Adapted for Streaming (FAST) for cases where high performance is required. Furthermore, FIX/FAST can be combined with a UDP multi-cast transport to allow fan-out to large numbers of subscribers, and this combination has indeed been adopted enthusiastically by many leading exchanges and market data suppliers.
Does this mean the benefits of using FIX/FAST over UDP for market data distribution can now be realised? Unfortunately not, or at least not fully. There are still multiple unresolved issues that mean the full benefits cannot be realised:
How should clients receive their initial data values?
How should subscribers handle lost/dropped/out-of-sync packets?
What fields should remain unencrypted in the header?
How to announce and distribute newer versions of the FAST template?
Whilst these may seem corner cases, they make an enormous difference to the ease with which we can implement common high-performance market data gateways based on a single common code-base. The end result is that many gateway implementations in the market today do not share a common code base, which results in higher maintenance costs and longer time to market for new features.
I think this example illustrates the un-tapped benefits that exist today that can be captured by continuing to extend technical standards. Of course, there are reasons why sometimes progress has been slow in these areas, and I will look at that in a subsequent blog.