Efficient Development Methodology for Multithreaded Network Application

Abstract

Multithreading is becoming increasingly important for modern network programming. In inter-process communication platform, multithreaded applications have much benefit especially to improve application's throughput, responsiveness and latency. However, developing good quality of multithreaded codes is difficult, because threads may interact with each others in unpredictable ways. Although modern compilers can manage threads well, but in practice, synchronization errors (such as: data race errors, deadlocks) required careful management and good optimization method. The goal of this work is to discover common pitfalls in multithreaded network applications, present a software development technique to detect errors and optimize efficiently multithreaded applications. We compare performance of a single threaded network application with multithreaded network applications, use tools called Intelreg VTunetrade Performance Analyzer, Intelreg Thread Checker and our method to efficiently fix errors and optimize performance. Our methodology is divided into three phases: First phase is performance analysis using Intelreg VTunetrade Performance Analyzer with the aim is to identify performance optimization opportunities and detect bottlenecks. In second phase, with Intelreg Thread Checker we allocate data races, memory leakage and debug the multithreaded applications. In third phase, we apply tuning and optimization to the multithreaded applications. With the understanding of the common pitfalls in multithreaded network applications, through the development and debugging methodology aforementioned above, developers are able to optimize and tune their applications efficiently.