Introducing LatencyStats.com: A New Approach
Introducing LatencyStats.com: A New Approach
By Mark Schaedel, SVP Global Data Products at NYSE Technologies
Latency is one of the most used terms in today’s financial market structure debates but it is also one of the least understood. Latency truly has become an “emperor’s new clothes” issue in that we often compare latency performance figures and pretend that they are a useful benchmark when the truth is much different.
We all know that measuring time requires context -- namely identifying what exactly is being clocked. Yet we toss around latency numbers without this context rendering them meaningless. A race requires a starting line and finish line but we have neither – so what does a measurement of time duration mean without this context?
We have all heard the joke about the two men in the woods who somehow are taking the time to debate how they will evade a wild bear. One man questions the other as he laces up his sneakers, “you really think you are going to out run a bear?” The other man replies, “I don’t have to out run the bear, I just have to out run you!”
Speed is only useful in relative terms – from our customer’s perspective; it is not a matter of speed versus the clock but how fast they are relative to their peers. Using a clock as a benchmark introduces variables. Other than how fast they travel, even the two men running alongside one another will have variables impacting the outcome, such as the path they travel, their precise starting point, precise ending point (gulp!) as well as the precise location and travel path of the bear.
Of course, the world of trading involves many more variables including the network paths our customers travel, the message routing experience, their own network hand off, the technology they deploy to handle the message traffic, the market and network conditions, and so on.
As an exchange operator, The New York Stock Exchange is front and center in the latency arms race. To compete for order flow today, exchanges must invest heavily to ensure that they can provide their customers with a satisfying experience and be able to scale that experience consistently during activity peaks when latency becomes most critical.
Achieving this consistency is the most difficult part of the effort. Without a doubt it’s the most important aspect of performance for customers as it is a variable they cannot control and therefore must trust their previous experience to predict. NYSE Euronext has made substantial investments to remain competitive in this regard and will continue innovating to make sure that our customers get the best experience possible when they interact with our markets.
Transparency is the next logical step in the innovation process as it is the key ingredient in delivering a predictable experience. We believe that transparency also allows our customers to fully benefit from the investment we have made by eliminating other variables from the equation and allowing customers to dissect the experience they have with our markets and isolate performance obstacles. In addition, tracking latency often identifies the smoke before the fire and can provide a better means of detecting problems before they become more serious.
We kicked off our transparency initiative two years ago when we began quietly exposing our internal time stamps on our core market data feeds. Unlike other markets that apply time stamps when each message is published, we also inject the time the market event took place which is crucial to truly understand latency behavior as well as being useful in strategy modeling. Exposing one’s dwell time can be a risky proposition from an IT perspective but ultimately transparency allows us all to make better decisions and improves the trust between exchanges and customers.
Latencystats.com represents our latest step in this initiative and one of several areas of intensely focused innovation. We partnered with Corvil, the market leader in latency measurement to both provide the tools necessary for providing transparency and keep us honest as an independent third party. The result is full transparency throughout the market data distribution path. Together with our internal time stamps and the prolific order acknowledgement (“ack time”), this information improves visibility in an area that is usually opaque. Of course, the ultimate objective is to connect these measurements together as a full life cycle. Market data is perhaps the most opaque part of the life cycle and publishing the latency in real-time along with a precise blueprint detailing how and what was measured is the first step, going further than any market has gone before in latency transparency.
We welcome other markets to join and publish their latency performance figures in precisely the same way – Register your Interest. Corvil has offered to independently verify that we each play by the same rules. Until we do, latency figures are meaningless as a means of comparison across market experiences.
To those markets that are reluctant to match, consider this – your customers have already formed their own opinion about your latency performance and that opinion was formed from their own experience. It is likely that the truth lies somewhere in between what you consider as fact and what they have concluded from their own experience. Everyone will benefit from a central source for information about latency and we’re excited to get started!