Optimizing Operation Scheduling for In-Memory Databases

A Computer system is a sophisticated combi-nation of software and hardware components that work closely together to solve today's problems. While algorithms within a software component itself can be implemented and optimized by a small group of developers, algorithms for an entire software stack are more complex and their development gets expensive. To speed up the development of in-memory database specific scheduling algorithms for oper-ating systems, we implemented a simulation model based on SAP's TREX in-memory column database. This simulation is used to optimize query scheduling and execution algorithms without the necessity of modifying every component within the existing system. To quantify the advantages and draw-backs of the different scheduling algorithms, we use the Star Schema Benchmark to measure the cost reduction per user with regards to the implemented operating system scheduling techniques.