User level Static and Dynamic Core assignment in Multicore System

There has always been a debate over the advantages of multi-core architecture over single core. It is obvious that the full advantage of multi-core can be achieved only if the programs are run in parallel [Amdahl's law] and the load is distributed evenly among all the available cores. In this context, processor or CPU affinity plays a major role in improving the performance of the system where time constraints are present. But, hard CPU affinity can also degrade the performance of the system in some scenarios. The alternative is to dynamically allocate the CPU core at application level. This paper studies these two techniques at application level to assign a particular CPU core once and stick to it, or switching among various cores dynamically by considering CPU usage.