Brendan Gregg is one of the most influential practitioners of computer performance engineering, known for inventing the flame graph and for turning systems performance analysis into a teachable, repeatable discipline. His own bio traces a career across several of the companies where large-scale performance work happens. At Sun Microsystems he was a kernel engineer counted among the top performance experts, where he worked on ZFS storage and developed the ZFS L2ARC. He then led performance engineering at Netflix, doing large-scale performance design, evaluation, analysis, and tuning, and later worked as an Intel fellow on AI and datacenter performance.
His best-known contribution is the flame graph, the stacked stack-trace visualization he created and released in December 2011. On his own flame graphs page he describes it as a visualization of hierarchical data he made to visualize stack traces of profiled software so the most frequent code paths can be identified quickly and accurately. The technique spread to virtually every profiling toolchain and observability product, and it is the artifact most people first associate with his name.
Gregg also pioneered eBPF (extended BPF) as an observability technology, building and popularizing collections of tracing tools that let engineers ask arbitrary questions of a running Linux kernel and applications with low overhead. Many of those tools ship by default on major platforms. This work, together with the flame graph, established a style of performance analysis grounded in directly measuring production systems rather than reasoning about them in the abstract.
Much of his lasting influence comes through writing and teaching. He is the author of “Systems Performance” and “BPF Performance Tools” (both Addison-Wesley), texts that are recommended or required reading at major technology companies, and his methodologies for approaching a performance problem systematically are widely taught. His personal site, brendangregg.com, remains a primary reference for performance engineers, hosting the original flame graph material, extensive notes on Linux perf and tracing, and the methodology that ties them together.