GEMFIRE REAL-TIME EVENTS™

In fiercely competitive markets like Financial Services, which are characterized by high volatility and reduced margins, sensing the pre-cursers of events often provide the ability not only to proactively capitalize on opportunities that would have gone unnoticed otherwise, but it also can identify an area of process vulnerability that a company can address before it becomes a material issue. Several other industries like federal intelligence, telecom and manufacturing/logistics have similar applications for such event-driven business models. GemFire Real-Time Events (RTE), a core component of the GemFire Enterprise Data Fabric™ (EDF) is a CEP engine that enables an enterprise
to facilitate agile, event-driven business models. It helps companies:

• Instantly identify business events that are relevant to them based on real-time information that is constantly changing.
• Immediately analyze these events to discern patterns and scenarios of interest, with the ability to correlate with other sources of information like historical data, customer data, reference data, etc.
• Intelligently distribute appropriate information to relevant clients and applications that have to react to these business events.

From a technology standpoint, RTE is a *main-memory based active data management system that can analyze thousands of events per second and distribute events of interest to hundreds of remote clients*. The system consists of one or more server processes that hold and process data for any number of remote client processes. The processes create and populate the RTE server, and clients run queries against the data. The clients can perform two types of queries: adhoc SQL queries, and GemFire Continuous Queries (CQ) that are registered with the server and automatically updated as the data changes. Traditional query processors utilize a request response paradigm whereby a user poses a logical query against data and a query engine
processes that query to generate a finite result set. If this finite result set is cached then it becomes stale quickly, especially when the underlying data is continuously changing, as in the case of streaming data sources.

In the continuous query paradigm, the users register logical specifications of interest over streaming data sources, and a continuous query engine filters and synthesizes the data sources that deliver streaming, unbounded results to users. RTE is a continuous querying engine implementation that acquires the incoming streaming data into one or more server-side caches and continuously analyzes the updates to the tables to determine how the different queries registered are affected. Subscribing clients in GemFire use SQL statements to express their data set of interest. All the delta events are automatically pushed or pulled to/from the client caches periodically keeping their state synchronized at all times.

TECHNICAL HIGHLIGHTS

• Intelligence for handling hundreds of Continuous Queries: RTE uses several algorithms and techniques for managing query predicates to quickly determine the queries that are affected by incoming events and also supports advanced joins and incremental client-view maintenance.
• Data distribution and notification services: All events captured as deltas on the RTE server are translated in real-time into deltas on views registered by clients and pushed to them. Data can be distributed to hundreds of clients in real-time. Clients are notified of in-coming data through simple call-backs.
• High availability: The RTE server can be configured to be highly available. GemFire accomplishes this by replicating data as it arrives at one or more servers. All meta-data that the server uses, such as information on active queries and schema information, is available to all servers. Clients are automatically routed to alternate servers if the server to which they are connected fails.
• Simple Intuitive syntax and semantics: RTE provides a very intuitive and easy to use interface for application developers. All events are captured using simple SQL Insert, Update and Delete operations. Applications express interest in data through views expressed also using SQL select statements.
• Load balancing: RTE can be configured to load balance connections across several replicated servers. Load balancing will result in the servers performing and scaling better. Clients will have the choice of using sticky or round-robin load balancing schemes.
• Automatic fail-over: When configured with multiple replicated servers, RTE will automatically detect unrecoverable errors in server connections and automatically delegate to alternate servers.