Lock contention analysis in multiprocessor systems
Lee W. Hoevel, O. Panfilov · 2002
An analytical model is described which deals with contention in acquiring exclusive spin/wait locks during access to shared data structures. It is assumed that after generation of a lock request a processor is waiting until the lock is granted. A performance speed-up factor is used as a performance measure in contention analysis. File partitioning is also considered as a means to reduce lock contention.>