Here’s some choice quotes from a whitepaper on StreamInsight:

Relational database applications typically acquire data and store it to disk before it can be analyzed. We therefore call analysis with traditional relational database systems query-driven. Query-driven analysis is well-suited for historical data. … To reach the necessary performance and scale, [some] applications need to analyze the data in near real time while it is being acquired from the source. We denote these applications as event-driven applications because new event data arriving at the system triggers the necessary analysis.

I think this is a great intro to help wrap yr head around the difference between query-driven and event-driven analysis.

Microsoft StreamInsight is Microsoft’s platform to build high-throughput, low-latency event-driven analytics applications.

That’s a great one liner.

With StreamInsight, business insight is delivered at the speed at which data is produced, as opposed to the speed at which traditional reports are processed or consumed.

Nicely worded value prop.

StreamInsight’s runtime performs calculations incrementally whenever possible. This means that the processing only involves the data for the current result and the new event. Unlike in traditional databases, updating a report with aggregates or KPIs with StreamInsight does not require to re-iterate through past data once a new event comes in. Instead, StreamInsight answers continuous queries with a single pass over all the data, which is an important capability for long-running, potentially infinite, standing queries. Incremental processing is one key performance benefit of StreamInsight.

That is rad engineering methinks.

StreamInsight automatically distributes the processing across the available processor cores on the system as well. Thread management and query parallelization are performed automatically by the system.

How cool is that!