Improving Performance Isolation on Chip Multiprocessors via an Operating System Scheduler

摘要: We describe a new operating system scheduling algorithm that improves performance isolation on chip multiprocessors (CMP). Poor occurs when an application's is determined by the behaviour of its co-runners, i.e., other applications simultaneously running with it. This dependency caused unfair, co- runner-dependent cache allocation CMPs. interferes 's control over priority enforcement and hinders QoS provisioning. Previous solutions required modifications to hardware. present software solution. Our cache-fair ensures application runs as quickly it would under fair allocation, regardless how actually allocated. If thread executes fewer instructions per cycle than scheduler increases thread's CPU time slice. way, overall does not suffer because allowed use longer. our implementation in Solaristrade 10, show significantly for SPEC CPU, JBB TPC-C.