Fast and Furious: Startup scaling with Kubernetes
summary
Twisto moves from one colocated server to a five-worker bare-metal Kubernetes cluster, deploying several times a day while handling tens of thousands of transactions across two countries. Tests show cloud VMs raise latency and raise costs, so the stack keeps dedicated servers plus S3/managed backups. Built: per-developer sandboxes from copy-on-write Postgres snapshots (half-terabyte replica), multiple on‑demand staging environments, CI that checks schema compatibility and a hotfix process to shorten fixes to under a minute. Monitoring: Prometheus, ELK, PagerDuty; outcome: resilience and faster team throughput.
AI-written summary · about
chapters (31)
- 0:00Intro
- 0:24Who am I
- 1:08Twister
- 3:00Poland
- 3:27Romania
- 4:31Benefits and risks
- 7:10New problems
- 8:48Initial reaction
- 9:15Performance testing
- 10:39Latency
- 12:07Memory costs
- 12:40Moving to Cloud
- 13:12Easy and Fast
- 15:23Kubernetes Cluster
- 16:33Benefits of Cloud
- 17:46Cloud Providers
- 18:38Team
- 18:55API
- 19:48Frontend
- 20:26Statistical models
- 21:39Parameters
- 23:09Microservices
- 24:21Agility
- 25:00Features
- 26:21Time correction
- 27:29Hotfix process
- 29:16Database testing
- 30:50Staging environments
- 32:01Tools
- 33:32Summary
- 35:38Questions and answers
related talks
46mWhy Kubernetes is Not Enough
47mDeploying Like a Boss: From 2 Servers to 20,000
34mEmbracing Upstream Kubernetes in Web Scale Organization
35mKubernetes journey
37mBalancing Power and Pain: Moving a Startup From a PaaS to Kubernetes
39mStrangling the Legacy with Kubernetes
53mImplementing Kubernetes on with confidence
42mDrupal, Kubernetes, Production