Case Study: Implementing SLOs for a New Service
summary
Squarespace's Ceph-backed S3-compatible object store implements SLOs using synthetic probers instead of instrumenting every client, collects four weeks of SLIs to set baselines, and uses a Ceph durability calculator to justify a 8 nines durability SLO per year. Availability SLO is 99.9% over four weeks (≈43 minutes/month error budget), latency SLOs include p90 ≤300 ms and p99 tracking, and durability checks validate cross-site replication and integrity; result: better alerting, clearer prioritization, and published SLO docs with quarterly review.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:55What is Squarespace
- 1:24Agenda
- 1:57Object Storage
- 2:46Applications
- 3:43API Availability SLO
- 5:06SLO Implementation Process
- 6:01Step 1 Determine SLOs
- 6:13Step 2 Understand Users
- 6:37Step 3 Understand Components
- 7:29Step 4 Define SLOs
- 7:44Step 5 Define SLOs
- 8:04Our SLOs
- 10:39Collecting as Allies
- 11:42Latency SLO
- 12:27Performance SLO
- 12:47Availability SLO
- 14:09Error Budgets
- 14:29Hellos
- 16:14SLO graphs
- 16:33Alerting
- 17:18Lessons
- 18:22Tips
related talks
24mSLOs for Data-Intensive Services
28mEngineering Data Reliably Using SLO Theory
48m8ms/99th write percentile latency - is it fast?
45m8ms/99th write percentile latency - is it fast? (...)
32mCeph Durability: How Safe Is My Data?
54mPrimary storage for SaaS solutions
45mChasing unicorns : 0 dataloss and 99.999% availability
34mSailing the Database Seas: Applying SRE Principles at Scale