StreamBase Systems last week announced a version of its data filtering and analysis software geared for financial services companies with particularly time-sensitive and gluttonous processing requirements.

The Lexington, Mass., start-up makes software that can process, analyze and act on data within milliseconds of when it arrives at a company's doorstep. It's tuned to deal with hundreds of thousands of messages per second. Typical uses of the technology include processing data from radio frequency identification (RFID) devices and other sensors, monitoring networks for intrusions, or tracking credit card transactions for fraud.

The StreamBase Processing Engine differs from typical store-and-query database systems from vendors such as TimesTen (which Oracle earlier this month announced plans to acquire) because it processes data on the fly without requiring it be loaded and indexed in a database. StreamBase runs all time-critical operations - including event processing, message routing, storage and execution of application logic - as part of one multi-threaded process, in a single process space. This approach eliminates the need for process switches among distinct software system components.

Database pioneer Mike Stonebraker - who was the main architect of the Ingres and Postgres relational databases and founded Informix - founded StreamBase in 2003 and today is CTO. The company released its first product in February after completing an $11 million second round of venture financing in January.

StreamBase 2.0 adds features geared for data-rich financial services businesses such as hedge funds, exchanges and management firms. In that industry, consolidated U.S. equity and options market data volumes can top 70,000 ticks per second at peak times, according to financial services industry research firm TABB Group.

New to the upgraded product is the ability to record and play back live datastreams. This will let users model market assumptions or a new trading strategy and see how it would perform against real, archived market data, for example. Alternatively, a network manager might try out a new intrusion-detection algorithm against archived system data to see how well it spots past incidents before deploying the algorithm in a production environment, says Bill Hobbib, vice president of marketing at StreamBase.