"The Cloud" at Badoo a Year Later: Fixing Our Mistakes
summary
A retrospective of Badoo’s cloud-like infrastructure one year later, covering its overall architecture for load distribution and fault tolerance, plus how script logs are collected, indexed, and viewed. Describes use of Google App Engine as a cloud queue processor, problems encountered, and future plans including a PHP phproxyd, control logic in Go, and a reimplementation approach using Go and Tarantool.
AI-written summary · about
chapters (18)
- 0:00Badoo cloud intro
- 1:12Why we built cloud
- 3:40Legacy cron problems
- 5:26Cloud architecture
- 8:58Cloud scale & performance
- 10:25Scheduler rewrite in Go
- 11:52Load balancing algorithm
- 13:45Handling server failures
- 15:58Script execution reliability
- 18:24Task logging & monitoring
- 20:34Queue processing challenges
- 22:46Custom queue parser
- 24:58Production issues & downtimes
- 27:07Human errors in ops
- 28:34MySQL operational problems
- 31:44MySQL replication limits
- 33:09Future improvements & Go adoption
- 35:59Q&A
related talks
41mThe Evolution of Tools and Services at Badoo
49mDocker in Practice: A Look at Its Use in Badoo After a Year
49mServing 30,000,000 Requests an Hour in the Cloud
48mIncident Management at Badoo
28mCloud in your Cloud
19mEverything I learned about dynamic scaling of cloud apps.
39mHow GCP Helped Me Save My Production at Mamie’s
29mBowerbirds of Technology: Architecture and Operations at Less-Than-Facebook Scale