Effective Performance Engineering at Twitter-Scale

Is performance engineering more craft than machinery? How do you scale something that seems to require both domain-specific context and comprehensive knowledge across multiple levels of the software and hardware stack?

Back in 2017, I was itching for a new adventure after returning from maternity leave. The result was a brand new team and function at Twitter — the Performance Engineering, aka the IOP Team. This small but mighty assembly achieved a lot over the next 5 years, saving Twitter over 100 million dollars, creating some of the largest data processing jobs owned by infra, driving multiple company-wide initiatives, and releasing multiple open source projects and publications. And we had fun.

In this talk, I will recap that experience while summarizing some key lessons learned about effective performance engineering. Spoiler— it was all about platform building.


Speaker

Yao Yue

Platform Engineer, Distributed System Aficionado, Cache Expert, and the Founder of IOP Systems

Yao Yue is a platform engineer with specialties in caching, distributed systems, and performance engineering. She worked at Twitter for 12 years, first led the Cache Team and later created the Performance Team. After November 2022, she co-founded IOP Systems, a company that is working to improve software efficiency and reliability via smart performance engineering, with her teammates from Twitter.

Read more
Find Yao Yue at:

Date

Tuesday Oct 3 / 03:55PM PDT ( 50 minutes )

Location

Ballroom A

Topics

Platform Engineering Performance Distributed Systems Staff Plus Engineering Microservices

Share

From the same track

Session Platform Engineering

Understanding Platforms: What They Are, Why They Work, When to Use Them, How to Build Them

Tuesday Oct 3 / 11:45AM PDT

Technical concepts are something that are thought of, approached, and understood differently across engineers, managers, and executives. Bridging the gaps and providing understanding to a complex and nuanced topic across all three groups can sometimes feel impossible.

Speaker image - Hazel Weakly

Hazel Weakly

Head of Infrastructure & Developer Experience; Director, Haskell Foundation; Infrastructure Witch of Hachyderm

Session Platform Engineering

Building Better Platforms with Empathy: Case Studies and Counter-Examples

Tuesday Oct 3 / 02:45PM PDT

Break out of traditional IT roles with your internal platform. Build a product based on customer empathy and real needs to achieve broad adoption.

Speaker image - David Stenglein

David Stenglein

Solo Consultant @Missing Mass, LLC with Over 28 Years in Systems, Software, and Consulting

Session Platform Engineering

How to Get Tech-Debt on the Roadmap

Tuesday Oct 3 / 05:05PM PDT

Only doing product-led work can lead to an unmaintainable system with lots of downtime. Unfortunately, getting time to work on the things that would prevent that can be challenging.

Speaker image - Ben Hartshorne

Ben Hartshorne

Principal Engineer @Honeycomb, Building Operable Systems with Resilience and Business Value

Session Platform Engineering

Building a Successful Platform: Acceleration, Autonomy & Accountability

Tuesday Oct 3 / 10:35AM PDT

Do we build a greenfield platform or do we incrementally centralize common foundations? Do we abstract away all complexity or provide upfront knobs and handles for composability?

Speaker image - Smruti Patel

Smruti Patel

VP of Engineering @Apollo Graph

Session

Unconference: Platform Engineering Done Well

Tuesday Oct 3 / 01:35PM PDT

What is an unconference? An unconference is a participant-driven meeting. Attendees come together, bringing their challenges and relying on the experience and know-how of their peers for solutions.