How City Storage Systems designed its Data Infrastructure to minimize toil
summary
City Storage Systems runs CockroachDB as the Tier‑0 production store for four years, deployed on Kubernetes with a home‑grown K8s operator that manages multicluster, multi‑region Cockroach clusters and automates CDC/table replication. Two cloud migrations include a rushed nine‑month move; most clusters use five replicas and deliberately run single AZ per region to reduce cost. OLTP is primary; Trino is used for analytics via a custom connector (one production conflict due to lack of AS OF SYSTEM TIME). Autoscaling CPUs and resizing disks minimises operator toil.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:10Defining toil
- 4:05Focus on invisible infrastructure
- 8:04Kubernetes and multicloud
- 13:21Managing Kubernetes
- 15:48Choosing workloads
- 18:17SQL flexibility risks
- 21:09Critical vs non-critical workloads
- 23:59CockroachDB feedback
- 28:15Closing
related talks
31mFuture-proofing operational data at scale
48mBuilding mission-critical applications with CockroachDB
15mJourney from PostgreSQL to CockroachDB: Scaling Cisco AI Defense
14mHow CockroachDB can enable Multi-Region Deployments and Data Locality 
23mWhy Squarespace adopted CockroachDB for better global performance at scale
33mWhat We’ve Learned Building a Multi-Region DBaaS on Kubernetes
15mScaling to support America’s #1 sportsbook
14mBuilding CockroachDB for performance at scale