Don't burst in Flames: 20 years of Performance Engineering with Martin Spier

·50 min·4 clips

How did a six-person team at Netflix solve intermittent CPU spikes with a heat map tool?

Hosts Brian Wilson and Andy Gravner introduce Martin Spier, a performance engineering veteran. Spier recounts his career start as a sysadmin and involvement in a performance improvement project that led him into performance engineering. He worked at Expedia before joining Netflix over a decade ago, where he celebrated the 20 million user mark and helped migrate to the cloud. At Netflix, Spier faced scaling issues and developed tools like Flame Scope and Flame Commander to visualize CPU profiles and detect patterns. He explains how a small team of six engineers built these tools due to a lack of existing observability solutions at scale. Spier discusses Netflix's decision to open-source these tools, considering competitive differentials and talent attraction. He highlights the potential for training models to automatically detect performance patterns from visualizations. After leaving Netflix, Spier joined PicPay, a Brazilian fintech with over 70 million accounts, where he leads foundation engineering. He aims to bring Netflix's lessons to improve engineering maturity and observability at PicPay. Spier describes the 'paved path' approach, offering a managed platform to streamline development without mandating usage. The conversation touches on performance testing strategies, shifting from pre-production load tests to canary releases and production monitoring. Spier emphasizes the importance of distributed tracing and metrics over logging for scalability. He reflects on the evolution from performance testing to engineering and observability, sharing insights on tool development and community contributions.

As heard by us

A grounded look at performance work, open-source tradeoffs, and the conference circuit.

Pure Performance builds on Martin Spier's performance stories to tie together the hosts' seven-year run, episode 193, and the practical appeal of why Netflix opened up its tools.

Read the full review in PlayNext →

Why you'd press play

Episode 193 leans into performance engineering, a long-running Dynatrace milestone, and the Netflix open-source question at the center of the chat.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from PurePerformance