OpenWLC: A Scalable Workload Characterization System

This paper discusses the design and development of a workload characterization system, known as OpenWLC, for large-scale computing systems. Large-scale hereafter implies computing systems with number of nodes ranging from thousands to hundreds of thousand. In such environment, it is critical to accurately monitor the system activities (workloads) and determine the potential performance bottlenecks that limit its effectiveness. Obviously, the challenge lies in providing a performance evaluation framework that can optimally and simultaneously meet the criterions of scalability, performance, extensibility, availability and minimum intrusiveness. OpenWLC is an attempt to meet this challenge. In particular, the OpenWLC framework employs a component-based multi-tier architecture coping with large amounts of monitored data during collection, storage, visualization and analysis stages. At the same time, OpenWLC aims to scale its framework up to and beyond 100,000 processors and to minimize its impact on system application performance by less than 1 % (on an average). It is important to note that the extent of OpenWLC goes beyond typical monitoring software. It is indeed an integral environment developed for collecting performance data, workload modeling and system evaluation specially designed for large-scale computing systems