The low latency race may be showing signs of hitting a speed bump, as companies are finding that gaining an extra microsecond advantage may not be worth the high costs.
Rather than pouring millions into nanoseconds of improvement, firms are focusing on network service and design issues that deliver milliseconds of improvement at a fraction of the cost.
“It makes sense to arrive at a profitable intersection of speed/latency and available margin through scalability, across multiple regions,” said Ian Bowell, head of technology for the Americas at Options IT, a technology infrastructure provider. “In other words, rather than pursue the ultimate low latency platform, at any cost, firms should take the sensible approach that reaches an acceptable level of latency but still leaves room for profitable service and trading performance, and not get blinded purely by speed."
Instead, firms are looking at ways to speed up the risk management functions that support high-speed trading, and are allocating investment dollars accordingly.
“Latency management has evolved over the last couple of years,” said Donal O'Sullivan, president of product management at Corvil, a trade monitoring software group. “Latency management used to be about being as fast as you could possibly be, which meant trying to shave milliseconds, microseconds and even nanoseconds off latency performance.”
That has changed, however. Latency management now means “understanding how fast each part of the infrastructure is in quickly identifying and resolving problems”, said O’Sullivan.
“It used to be about the speedometer and accelerator,” he said. “It’s now about a full dashboard of temperature and fuel gauges.”
Corvil has launched a new range of “appliances” for its CorvilNet Latency Management System, with hardware based on Intel's Sandy Bridge architecture, which the company says will deliver greater than double the performance increase for the same cost.
By doubling the processing capacity, customers will be able to monitor greater volumes of market data and client trade flow.
“As data rates continue to grow exponentially, trade monitoring and surveillance is more important than ever,” said O’Sullivan.
The new appliance will enable Corvil’s customers to deploy latency management more broadly in their organization.
“Our appliances sit within the trading infrastructure and communicate with each other, either globally or locally, over a network,” said O’Sullivan. "Our customers are looking for enterprise-wide latency management solutions.”
The new CNE-7300 appliance can perform full decode and analysis on up to two million messages per second. Storage capacity is also more than doubled, allowing data to be retained and analyzed for longer periods of time.
“It presents alerts when it detects that something is going wrong, using visualization tools with green, amber and red dials,” said O'Sullivan. “It produces end-of-day and streaming latency data, so smart-order routers can use latency measurement in real time.”