Noah Watkins talks on Push-based Query Processing with RT-DataPath

We consider the problem of providing predictable performance to an application-level workload consisting of many declarative queries with specific response time latency tar- gets. At a low-level we use existing techniques that provide predictable I/O performance, and avoid worst-case analysis by (1) extracting performance requirements from the application workload, (2) transforming these requirements onto the interface of the predictable I/O sub-system, and (3) intelligently scheduling I/O for efficiency.