Latency is important, but consistency is vital.

By Mats Andersson, Chief Technology Officer, Market Technology, OMX 

In a market where speed ultimately determines the winners and losers, the quest for ultra low latency has become the new Holy Grail. Less than 10 years ago, latency – the delay between receiving data and acting on it, or submitting an order and executing it – was often measured in seconds. Today, with algorithmic and program trading fast becoming the norm, latency is measured in just a small number of milliseconds. And, some industry gurus are suggesting that the new standard should be microseconds, even though a latency of one millisecond is still out of reach for the vast majority of marketplaces.

Yet, even with this intense focus, there is still much confusion, even among exchanges, regarding the consistency of latency, how low is low enough, and perhaps most mystifying, how to accurately measure latency so that it can be compared among various institutions.

The sad truth is that it is nearly impossible today to compare latency figures. The industry has not yet established a standard benchmark for latency, and there is surprisingly little consistency in how these numbers are measured.

There are many different ways to configure software, hardware and networks. And, when you consider all the other variables that can come into play – communication protocols, message data formats, number of instruments traded, trading patterns, actual transaction load during measurement, geographical distance, use of reliability and disaster recovery functions, to name a few – it can be next to impossible to state with certainty how different latency numbers should be compared. To some degree, industry adoption of the FIX and FAST standards for trading and market data messages will simplify the creation of benchmark definitions.

In addition to establishing a clear benchmark for measuring latency, the industry also needs to determine just how low is low enough. The benefits are not equal to all. Low latency is a key competitive advantage in the North American market for equities and equity derivatives, for example, where the marketplace with the fastest response time will presumably get the order. With the upcoming implementation of MiFID, European markets are seeing a similar need. The low latency requirement is not as clear in markets that are currently less competitive, in Asia, for example, or in the trading of asset classes that may be less quantitative or that are not subject to competing, alternative execution venues.

The answer, however, to how low is low enough is a simple one. Marketplaces need to be fast enough to stay ahead of their competition and, as a key enabler of speed, low latency is vital, but it does not stand alone. Exchanges need to view their investment in ultra low latency in the context of their overall trading environment and larger business plan. The cost of speed to the business, and not just the monetary cost, needs to be considered. Risk control and reliability should never be compromised in the hunt for microseconds. Latency is important, but consistency is vital. Traders need to know precisely how much time it takes to quickly withdraw their quotes.

So, predictablility needs to be a key element of the latency equation. While predictability can only be ensured through rigorous testing in a variety of market conditions against standard benchmarks, today no standard benchmarks exist. OMX is working towards developing measurable latency standards. However, that isn’t enough. We as an industry need to band and together to develop and agree on the appropriate benchmarks. Only then will we be able to accurately compare latency figures within the industry.


Photo: Johan Bergling

MarketView 2007:2