How the Sony PlayStation Network Does SRE
summary
PlayStation Network runs on AWS and Kubernetes, uses production game-day and load tests tied to PS5 sales/logistics, and shifts 24/7 monitoring in-house; traffic spikes in March 2020 hit about 1.5× concurrency versus Christmas and some APIs see 30–100× traffic after hits, yet the PS5 launch records zero red critical incidents during Nov 12–22. Four SRE models coexist—embedded, centralized, no-SRE and hybrid—so tradeoffs are concrete: scaling, task prioritization, team count (Tokyo embeds ~20–30 SREs across ~200 services) and cross-team coordination.
AI-written summary · about
chapters (19)
- 0:00PlayStation Network SRE intro
- 1:07PSN history and growth
- 3:54COVID-19 traffic impact
- 6:28PS5 launch challenges
- 9:22Handling PS5 traffic spikes
- 11:15Tokyo SRE evolution
- 13:02Early PSN operations
- 14:49PS4 cloud transition
- 16:43PS5 embedded SRE model
- 19:12Global SRE team structures
- 22:27Supporting teams beyond SRE
- 24:33Embedded SRE pros and cons
- 28:04San Diego Commerce SRE model
- 30:10San Diego Account SRE model
- 32:44San Francisco no-SRE model
- 34:34Comparing SRE models
- 37:25PSN SRE future opportunities
- 38:52Hybrid SRE model
- 40:16Conclusion
related talks
30mThe Path to SRE
25mExpect the Unexpected: Preparing SRE Teams for Responding to Novel Failures
21mSRE: An incomplete guide to cultural Narnia
35mSRE -- culture and strategy @ TrustingSocial
30mWhen Is SRE Right For You?
30mDifferences in SRE Implementations across Companies
27mMean Time To SRE: A DevOps story from adidas
25mSRE in the Third Age