Recap: Our 3rd Performance Engineering Meetup
September 4, 2018
Once per quarter, the LinkedIn Performance Engineering meetup brings together engineers passionate about performance and efficiency from across the San Francisco Bay Area. The meetup is a great way for engineers to meet, share notes, and learn from each other on topics like performance measurements, monitoring, optimizations, and culture.
In mid-July, we hosted the third event in this meetup series at our Sunnyvale campus. Thank you to everyone who attended the event and made it a success! For updates on the next event, please follow the @LinkedInEng Twitter account and join the LinkedIn Performance Engineering group. Looking forward to seeing both returning and new faces next time! For those who couldn't make the event, here’s a short recap of the programming.
"Balancing performance and security in Chrome while addressing the Spectre vulnerability"
Our keynote talk was by Googlers Andrew Whalley and Bradley Nelson, who worked on the Chrome browser to fix the Spectre/Meltdown security vulnerabilities. Their talk focused on a particular variant of the Spectre attack (Speculative Execution attack) and how Chrome addressed it while making sure the browser remained highly performant. One of the attendees noted that this was the clearest explanation of the Spectre attack they had ever heard. Software engineering is all about tradeoffs, and it was great to get a sneak peek of how browser developers handle tradeoffs in their world while still managing to keep the browser high-performing.
"Architecting for failure – Story of Factor Of Safety"
Jingjing Sun is a Staff Software Engineer at LinkedIn. She is part of the Voyager API team, which builds the API frontend for the member-facing www.linkedin.com web app and mobile apps. Jingjing’s talk focused on maintaining and building a highly-performant and highly-resilient service that can handle LinkedIn’s scale. She described an internal project at LinkedIn called “Factor of Safety,” which developed built-in safety mechanisms to throttle incoming traffic when load increases beyond safety. This feature has helped stabilize our API servers in the past year and enabled us to deliver fast and reliable experience to LinkedIn members.
“Heliosphere: LinkedIn’s browser performance testing framework”
Our last talk of the day was by Sobir Bazarbayev, who is a Senior Software Engineer on the LinkedIn Performance team. Sobir described LinkedIn’s internal browser performance testing framework, called Heliosphere, which is used to test performance of our web apps on browsers before a commit hits production. Heliosphere helps prevent performance regressions from making it into production by analyzing navigation timing data using statistical techniques. It also avoids any impact of variance in server processing durations by utilizing har remix.
In case you weren’t able to attend, here’s a video of the presentations from our LinkedIn engineers.