White Papers

Clock Synchronization

The Jitters and Everything You Were Afraid to Ask About Precision Time

There are many approaches to, and techniques for, managing the performance of high-end applications, but the foundation they all must be built on is precise instrumentation. If you don’t know how your application is actually performing, then you can’t ensure it meets your needs; if you don’t know what the infrastructure it runs on is delivering, then you can’t ensure that it is properly resourced and provisioned to support the application. In the realm of low latency applications, such as high-performance trading, the foundation is precision timing: without knowing exactly when specific events occur, you can’t begin to measure, analyse, and manage latency.

The requirements on the precision of timing are driven by the timescales of the processing that happens in order processing and market data delivery. These have shortened dramatically over
the last few years: just five years ago, typical times taken by mainstream exchanges to enter orders in their books were of the order of tens or hundreds of milliseconds. Now several exchanges advertise execution times of hundreds of microseconds - a decrease by a factor of 1000 in latency. We have seen similar acceleration in market data delivery and processing, with the increase in available network bandwidths, and the development of streamlined protocols such as FIX FAST (FIX Adapted for Streaming) and high-performance feed-handlers.

This relentless lowering of latency in electronic trading systems has raised the bar for the precision of timestamps: whereas accuracy of 1ms (one millisecond) would have been more than sufficient previously, the basic precision that is required today is 1 μs (one microsecond), and we see pressure to increase the precision to hundreds of nanoseconds or better.

Request a full copy of this whitepaper.