The Katalyst family of electronic offering, sales, and aggregation systems employ a multi-tiered architecture built on top of Kestrel's proprietary Distributed Messaging Infrastructure (DMI). In addition to providing messaging and storage abstraction facilities similar to RAPTr’s PDPI, the DMI also provides clustering, monitoring, goal-seeking failure detection and recovery facilities.

The Katalyst architecture is server-centric, with thin Java components that are responsible only for providing a presentation layer to the Katalyst server cluster. At the applications level Katalyst consists of one or more interfaces to customer side systems (i.e., trading systems or back-office systems), one or more channels (i.e., external ECN’s, native Java clients, or other presentation mechanisms) and a set of cluster-hosted applications that implement the pricing, market making and trade processing functionality.

Trading system interfaces are available to Bloomberg’s TOMS (via the Trading System and CRF protocols) and Kestrel’s RAPTr. Price feeds are available from Bloomberg’s BBCOMM, TradeWeb MDF, and RAPTr. Channel interfaces are available to Bloomberg’s MPF and AutoQuote/AutoEx.

Katalyst’s current implementation uses Oracle for persistent storage and TIPS for analytics although lower-cost alternatives are being implemented. In keeping with Katalyst’s economical implementation of low-cost hardware clusters, in the next release Oracle will be deprecated (but still supported) in favor of open source alternatives and TIPS replaced with in-house routines.