Support Operations Engineering: Scaling Developer Products to the Millions
Talkat SREcon19 Europe/Middle East/Africa · on Wednesday, 2019-10-02· at 11:30· Track 2: The Liffey A·
summary
Cloudflare scales support by turning diagnostics into engineering: a pipeline that fuses internal logs, a helper bot layer and machine learning to auto-resolve tickets. Cosine string-similarity with 2-gram tuning replaces naive Levenshtein, cascaded binary classifiers push false positives to a fraction of a percent, and formal runtime contracts enforce 0% risky changes. Results include 57.3% deflection for free plans, 78% resolutions without humans, and fingerprinting a credential‑stuffing cluster with a 30.5% success rate used to block attacks.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:18What is Support Operations Engineering
- 2:07Building a curve
- 3:39Need for automation
- 4:04Customer tooling
- 6:16Novel approaches
- 7:49Overengineering
- 8:38Formalized runtime contracts
- 9:53Taxonomy classification
- 10:22Attack mitigation
- 16:23What we learned
- 18:24Questions
related talks
51mWhat We’ve Learned From Billions of Security Reports
34mPanel: LiveOps and Support Strategies to Shape the Future of Your Games.
22mScaling Culture of Resiliency in the Enterprise at Charter Communications
48mDevops@Scale
56mWhat We’ve Learned From Billions of Security Reports
47mHow did I Help to Save the Lives of More Than 1.5 Million People?
18mAutomating 90% of support tickets: How we built Vercel's support agent
25mBuilding a highly scalable system to assure Business Resilience