The speed and complexity of both algorithmic and systematic trading coupled with incremental growth in volumes of market data have necessitated the creation of ever more efficient data processing tools. This presents an ongoing challenge to financial institutions where the quantity of data being processed threatens to challenge and sometimes overwhelm trading and risk management applications.
The tests were conducted at Intelâs Low Latency Laboratory at Winnersh in Berkshire, England, and were carried out on calculations of aggregated values used in algorithmic trading strategies.
The latest performance test focused on the change in performance as a result of:
a) upgrading from StreamBaseâs CEP engine version 3.7 to version 5.0, and
b) upgrading from Intel Single Socket Dual Core processors to Intel Twin Socket Quad Core processors
âWe are delighted with the results that DSAL has achieved with our Core architecture in our low latency Labs. This underpins the match between Intel processor performance and DSALâs experience in Market Data and Analytics,â said Nigel Woodward, Director, Financial Services, Intel Corporation.
The results achieved were:
â¢ 48 % - increase in updates processed per second
â¢ 53 % - reduction in the standard deviation between updates processed per second
Anne Pounds, Director of Strategy at DSAL, commented: âBy undertaking these tests using the Intel Low Latency Lab we are able to compile true benchmark measurements. Many financial institutions appreciate the factual measurements that we provide as part of our benchmarking service as we are able to assist in pin-pointing areas of improvement, not easily identifiable in this âsuperfast-computational /low-latency arena.â
David Poole, Chief Operating Officer of ClientKnowledge, a research and advisory firm focused on the wholesale financial sector, commented: âEnsuring that the data being processed is current presents a major issue. Expanding the limits of processing enables a systematic reliance on the subsequent analysis and execution. Confidence in sufficient processing power will lead to further innovation, rather than focusing on defensive concerns."
In June 2007 DSALâs first performance tests achieved 200,000 calculations per second and 1.5 million updates per second running on Intel Single Socket Dual Core processors without any configuration changes to either the applications or the hardware used. The objective of the second round of testing was to measure the impact of recent Intel processor performance improvements and the impact of configuration changes.
DSAL, together with StreamBase, its partner in providing low-latency, high-throughput solutions based on the StreamBase CEP engine, provides super-fast processing. DSALâs goal was to benchmark StreamBaseâs CEP engine in order to ensure that it was considerably faster than others available in the market. DSAL was invited by Intel to put its applications through a series of stringent benchmarking and evaluation tests at Intelâs Low Latency Lab. DSALâs engineers, working with the Intel engineers, configured and tested a number of performance measurement models, with the performance results again surpassing all expectations.
âIn the new fragmented world Smart Order Routing is becoming more complex and the requirement for speed is ever present. It is therefore very pleasing to see that technology is once again stepping up to the challenge,â said Bob Fuller, MiFID expert and ex CEO of Equiduct.
Algorithmic trading performance was measured on a typical aggregation calculation over time windows of data. The objective was to establish a level of updates per second that could be sustained over a meaningful period of time. The testing was conducted with the default server settings for all the components â hardware, operating system, StreamBase CEP engine and the DSAL application. In a real life trading environment the performance could be significantly improved by optimising the configuration.