Allocation-phase aware thread scheduling policies to improve garbage collection performance

Abstract

Past studies have shown that objects are created and then die in phases. Thus, one way to sustain good garbage collection efficiency is to have a large enough heap to allow many allocation phases to complete and most of the objects to die before invoking garbage collection. However, such an operating environment is hard to maintain in large multithreaded applications because most typical time-sharing schedulers are not allocation-phase cognizant; i.e., they often schedule threads in a way that prevents them from completing their allocation phases quickly. Thus, when garbage collection is invoked, most allocation phases have yet to be completed, resulting in poor collection efficiency.We introduce two new scheduling strategies, LARF (lower allocation rate first) and MQRR (memory-quantum round robin) designed to be allocation-phase aware by assigning higher execution priority to threads in computationoriented phases. The simulation results show that the reductions of the garbage collection time in a generational collector can range from 0%-27% when compare to a round robin scheduler. The reductions of the overall execution time and the average thread turnaround time range from -0.1%-3% and -0.1%-13%, respectively.

Exploiting FIFO Scheduler to Improve Parallel Garbage Collection Performance

“…Xian et. al [24][25][26][27] design several types of schedulers for Java applications, including object allocation phase aware and contention aware schedulers. The allocation phase aware scheduler schedules the threads according to their allocation rate or preempt threads with the memory consumption instead of the running time which is used in the CFS.…”

Exploiting FIFO Scheduler to Improve Parallel Garbage Collection Performance

“…The number of threads that can run at the same time is often determined by the number of CPUs. The scheduler decides when to run which threads, and the order of the threads' execution has been shown to affect the overall performance of an application [19,25,27].…”

Modeling heap data growth using average liveness

“…Neither does it transitively discover the death times for objects that are pointed to by earlier dead objects. Other techniques such as [22,25] proposed to use object lifetime profiles to improve GC performance. This paper uses Merlin and extends the use of OLP to include the seven new metrics of average liveness.…”