The recent STAC Summit saw Corvil showcase how measuring latency across each hop improves Tick to Order performance and outcomes.
Recent STAC Summits have highlighted the electronic trading sector’s growing interest in AI, but what was striking about last week’s Chicago event was a renewed interest in achieving performance advantage through ultra low latency. Two trends seem to be fueling this interest, the evolution from 10Gb to 40Gb and eventually 100Gb networks and the shrinking tick-to-order and tick-to-trade timeframes.
This was partly reflected in the vendors who attended Chicago, each offering different components in the electronic trading lifecycle. They are all targeting electronic trading firms with the promise of helping them improve throughput and performance with different solutions, from parallel file systems to FPGA, for optimizing their algorithms and using market data for competitive advantage.
The thought leadership sessions focused on ways to gain better market insight and improve trading outcomes. Improving one component is not the same as optimizing the entire trade plant. Corvil can help firms gain a performance advantage by providing transparency into the end-to-end performance across multiple hops within a single transaction.
By deploying our appliances at various points in the trade plant, firms can measure how successfully each component performs as well as understand the impact of each on the full order lifecycle, from market data tick, through trading algorithms, to the markets and back. This end-to-end, Tick-to-Trade analysis helps reveal the areas that deliver the best “bang for the buck” spent on new component technology.
As part of the event’s Innovation Roundup, Matt Davey, our Director of Product Management, gave a presentation about techniques and trends in instrumenting the trade plant for monitoring and optimizing Tick to Order latency.
Figure 1: Tick to Order latency - What is it and why measure it?
Note: There is some recent confusion in the use and definition of “tick to trade” and “tick to order.” In this blog we’re using the definitions shown in Figure 1, where “tick to trade” is measured from tick arrival to trade completion and “tick to order” is measured from tick arrival to order leaving the trade plant for the venue.
Matt discussed how the impact on trading success can now be measured in increments of hundreds of nanoseconds, and less, in some environments. The point here is that if your Tick to Order latencies are measured in milliseconds, then what you need to optimize will be measured in microseconds, and therefore your instrumentation will need to give you accurate data at those microseconds. Similarly, if you’re measuring in nanoseconds, you may need to optimize in picoseconds. So you need to carefully consider how your monitoring and analysis solution will accurately timestamp your events in order to calculate latency.
Digging deeper into monitoring and analysis challenges, Matt touched on several of Corvil’s innovations, such as flexible latency correlation, extended visibility into low-latency software through our App Agent, use of high-resolution timestamps from aggregation layer, techniques to handle traffic bursts on high-bandwidth networks, etc..
All of this research and development effort enables Corvil to provide a holistic view of latency across every hop even as trading volumes and speeds increase. We do this so our customers can measure latency accurately and understand its impact in order to effectively improve outcomes and optimize the trading business.