1 Million Messages A Second for Trading System
July 3rd, 2008 By: Rich Miller
A financial trading system has executed 1 million messages per second, establishing a new speed record. The landmark was announced jointly by Sun Microsystems and Intel, which designed a low-latency system for the Thomson Reuters Market Data System, a platform used by Wall Street financial houses to aggregate data and complete trades.
“We’ve broken the 1 million messages per second barrier using off-the-shelf Intel-chip hardware and Solaris 10,” Ambreesh Khanna, CTO for Sun’s financial services division, told Network World. “Reuters Markets Data System is widely used to aggregate information, and algorithmic trading is one reason market data has more reason to focus on latency.”
Timely market information is crucial in algorithmic trading, which now represents 22 percent of all trading on Wall Street. Financial firms using computerized trading systems measure success in milliseconds and microseconds, which is how long it takes to connect to exchange systems to execute a trade. A microsecond is one-millionth of a second. That kind of speed requires serious horsepower.
The Benchmark was conducted using Intel-based Sun Fire X4150 servers with dual Quad-Core Intel Xeon X5460 3.16 GHz processors, running Solaris 10 and using a 1 Gigabit Ethernet network infrastructure. This benchmark was done using the Thomson Reuters standard topology.
We did some tests with Intel last year which achieved better throughput than this.