Summary

Linux has enjoyed great popularity, specifically with low-end and mid-range systems. In fact, Linux is regarded as a stable, highly reliable operating system to use for web servers for these machines. However, high-end, enterprise-level systems have access to gigabytes, petabytes, and exabytes of data. These systems require a different set of applications and solutions with high memory and bandwidth requirements, in addition to larger numbers of processors. This introduces a unique set of issues that may be orders of magnitude more complex than those present in smaller installations. For Linux to be competitive for the enterprise market, its performance and scalability must improve.

You should now have a sense of the strategy for improving Linux kernel performance, with special emphasis on eight-way SMP scalability. This includes an overview of several performance evaluation methodologies, including tracing, workload characterization, numerical analysis, and simulation. For benchmarking, the strategy includes understanding the hardware and software used, definining the run rules, and setting the benchmark targets. In addition, it includes measurement and analysis to find the performance issues, and proposed solutions to resolve them by creating improvements, through kernel patches to the Linux kernel, focusing on architecture-independent issues. Performance tools and benchmarks, selected to provide coverage for a diverse set of workloads and to address specific kernel components, are used to accomplish this task. Also presented are selected benchmark results, which quantify progress toward making Linux better for eight-way SMP scalability. This methdology is a means of substantiating that the performance of the Linux kernel can be improved by working with the OSC to address degradation issues. This will make Linux better and ready for the enterprise market.