Diagnosing Performance Bottlenecks in Production Systems

There are many tools available to capture system performance metrics. Within the context of live production systems, however, there is much to consider in regards to their practical use when isolating a system bottleneck. In this talk, we will review several approaches to performance analysis and benchmarking as they would be applied using real-world examples and constraints, such as the need to keep systems in production during diagnosis. We will use a framework of high availability systems and storage to explore cascading performance impacts across multiple layers of a production stack, where many system bottlenecks may masquerade as others. Attendees of this talk will gain a working knowledge of effective strategies for diagnosing bottlenecks and optimizing system performance that may be used within a broad base of infrastructure applications.