The paper details the factors that affect a trading venue's latency - such as complexity of business model, whether and how transactions are safeguarded, hardware infrastructure etc - and what can be done to minimize it. It also elaborates on the relationship between latency and throughput. An important point made in the paper is that an exact definition of the points between which latency is to be measured is fundamental to a precise evaluation of claims regarding latency.
Cinnober is an independent technology provider serving leading exchanges such as Borsa Italiana, the Chicago Board Options Exchange and the London Metal Exchange, as well as alternative marketplaces such as Alpha ATS, Markit BOAT and Turquoise. Performance is at the very heart of the company's solutions.
In June 2009 Cinnober disclosed performance benchmarks of the, at that point, latest release of the TRADExpress(TM) Trading System revealing outstanding low latency with an average door-to-door latency of 286 microseconds, and throughput capacity of over 800 thousand quote updates per second in a setup with five matching engine partitions. Door-to-door latency is the time lapse from the moment an order enters the marketplace, is then processed in the matching engine, and the subsequent response and result are sent back to the participant.
The now released whitepaper presents performance benchmarks of Cinnober's TRADExpress Trading System showing the different configurations in which they were established, and reveals focus areas in the company's roadmap to reach a door-to-door latency of less than 80 microseconds within a year and 25 microseconds in 18 months.
'With this paper we are taking discussions on latency to a new, more relevant level', says Nils-Robert Persson, Cinnober's Executive Chairman. 'It has become very popular to publish performance figures making specific claims for latency but there is seldom enough accompanying information to make them relevant and useful for trading venue operators. We believe that hard numbers should always be supported by hard facts. At Cinnober we will continue our open and transparent approach to this topic, not only with regard to the measurement of latency but also to the publication of supportable facts regarding system performance.'