Critical lock analysis: Diagnosing critical section bottlenecks in multithreaded applications

Abstract

Critical sections are well known potential performance bottlenecks in multithreaded applications and identifying the ones that inhibit scalability are important for performance optimizations. While previous approaches use idle time as a key measure, we show such a measure is not reliable. The reason is that idleness does not necessarily mean the critical section is on the critical path. We introduce critical lock analysis, a new method for diagnosing critical section bottlenecks in multithreaded applications.Our method firstly identifies the critical sections appearing on the critical path, and then quantifies the impact of such critical sections on the overall performance by using quantitative performance metrics. Case studies show that our method can successfully identify critical sections that are most beneficial for improving overall performance as well as quantify their performance impact on the critical path, which results in a more reliable establishment of the inherent critical section bottlenecks than previous approaches.

A Java util concurrent park contention tool

“…Finally, critical lock analysis [8] could also be applied to our tool. The suggested method monitors only hot locks, i.e., locks that have critical sections on the critical path.…”

SyncProf: detecting, localizing, and optimizing synchronization bottlenecks

“…The first six programs are benchmark programs used by others to study concurrency and performance [9,25,42,43,49,64]. Only the first four benchmark programs have previously known bottlenecks [9,49,64]. The remaining programs are different versions of popular open source projects.…”

“…We use P IT,CS to denote the cost value associated with a particular CS, where CS refers to the unique static CS. Unlike existing contention measurement approaches [9,27,56] that focus on a specific metric, SyDGs provide a generic representation that enables SyncProf to compute multiple metrics that summarize the performance impact of CSs. Here, we introduce three metrics used by SyncProf.…”

“…Considering wait time can be effective at identifying synchronization bottlenecks, the results may be misleading when the synchronization bottleneck does not impact the completion time of a program. To address this problem, the CPWD metric considers the critical path 1 [6,9] of an execution while quantifying the performance impact of a CS. Applying CPWT requires isolating critical path graphs from other subgraphs in each SyDG: Definition 3: A critical path graph is a subgraph of a SyDG, where the last node of each connected component is from the last finished thread.…”

ParaShares: Finding the Important Basic Blocks in Multithreaded Programs

“…The first is to look for specific kinds of errors, sometimes within targeted program regions such as a program's critical path. For example, tools may identify load imbalances [4], long waits [16,8], lock contention [23,6], I/O blocking [18], or unnecessary I/O [5]. One issue with this approach is that each type of inefficiency may need its own tool or search procedure.…”