SR Labs Introduces Shared Memory Product for Low Latency Trading

SR Labs is releasing a new product this week that processes data in shared memory for financial firms integrating market data into high-speed electronic trading applications. The new product, known as Memory-to-Memory Data Replication Cache or MMDRC, allows both the market data and the trading applications to run multiple processes on the same machine.

“MMDRC allows a user on a multi-core server to process market data once and then distribute the data via shared memory to other applications on the same machine,” explains Samir Islam, a director at SR Labs in New York.

This builds upon the firm’s existing SMDS (Market Data Solution), which allows firms to run their proprietary strategies in-process with SR Labs market data libraries and co-locate the entire solution on a single server at the exchange. (SR Labs currently has 44 feed handlers ranging from NYSE Arca to Turquoise.) However, the new product is different because MMDRC centralizes the market data processing. For example, say a firm using the in-process solution had four trading applications running on a server, each packet from the Nasdaq ITCH (direct feed) would be processed four times and then used by each trading application on the machine. However, With MMDRC the data packet is processed once and that’s given to the 4, or 10 or however many trading applications are running on the system, explains Islam. “Because we use shared memory, it’s incredibly fast,” says Islam, noting that MMDRC is written in C++. Several hedge funds and proprietary trading desks at banks are using MMDRC in production, according to Islam.

Ivy is Editor-at-Large for Advanced Trading and Wall Street & Technology. Ivy is responsible for writing in-depth feature articles, daily blogs and news articles with a focus on automated trading in the capital markets. As an industry expert, Ivy has reported on a myriad ... View Full Bio

1 of 2