Continuously measuring critical section pressure with the free-lunch profiler

Abstract

Today, Java is regularly used to implement large multithreaded server-class applications that use locks to protect access to shared data. However, understanding the impact of locks on the performance of a system is complex, and thus the use of locks can impede the progress of threads on configurations that were not anticipated by the developer, during specific phases of the execution. In this paper, we propose Free Lunch, a new lock profiler for Java application servers, specifically designed to identify, in-vivo, phases where the progress of the threads is impeded by a lock. Free Lunch is designed around a new metric, critical section pressure (CSP), which directly correlates the progress of the threads to each of the locks. Using Free Lunch, we have identified phases of high CSP, which were hidden with other lock profilers, in the distributed Cassandra NoSQL database and in several applications from the DaCapo 9.12, the SPECjvm-2008 and the SPECjbb2005 benchmark suites. Our evaluation of Free Lunch shows that its overhead is never greater than 6%, making it suitable for in-vivo use.

Using Differential Execution Analysis to Identify Thread Interference

“…Analyzing the message rate can help detecting network contention [16], [17]. Several works show that lock contention can be detected by measuring the time spent waiting for locks [12], [13], [14], [15], [55]. Overall, these projects define metrics that identify a saturation, but they can not tell if the saturation is the cause of a performance bottleneck.…”

“…These tools give raw metrics, but they do not indicate if optimizing a block of code can lead to better performance. Typically, these profiling tools report metrics to know if a memory controller [1], [2], [3], [4], [5], a cache [6], [7], [8], [9], [10], [11], a lock [12], [13], [14], [15] or the network stack [16], [17] is inefficiently used. However, these metrics indicate a saturation, but a saturation does not necessarily lead to a large slowdown, either because the access to the saturated component remains efficient or simply because the slowdown caused by the saturation has only a marginal impact on the overall performance.…”

Parallel Execution Profiles

“…In [4] the authors present the Free Lunch profiler, a Java based tool for detecting lock contention. One of the most important contributions of this paper is the definition of what they call the free lunch metric, which is used to measure critical section pressure.…”

“…Understanding the runtime behavior of parallel programs is hard. Thus tools have been developed to understand parallel program behavior for performance debugging [1,2,4]. Existing works are aimed at specific metrics to identify performance bugs, e.g., normalized execution time, critical paths, slack etc.…”

Remix: online detection and repair of cache contention for the JVM

“…Previous systems that find and fix cache contention at runtime [27,28,34] have focused exclusively on false sharing in unmanaged C/C++ programs. Many existing profilers for managed languages [12] can identify some forms of true sharing like lock contention. However, code written in managed languages such as Java or C# can suffer from false sharing as well.…”