- http://talkminer.com/viewtalk.jsp?videoid=T-Xr_PJdNmQ&q=
- Recap (http://www.youtube.com/watch?v=T-Xr_PJdNmQ#t=39m30s)
- Version control everything
- Optimize early
- automate, automate, automate
- Use configuration management... no matter how many servers
- Plan to fail
- Instrument everything
- Hardware, software, network, OS.... everything
- Discover trends and causes you never knew of
- Don't waste time on dumb stuff
- ODS
- Ganglia
- Active users (http://www.youtube.com/watch?v=T-Xr_PJdNmQ#t=36m26s) :
- per engineer: 1.1million
- per ops: 2.3million
HipHop for PHP
Memcached
300+TB live data in RAM
MySQL
Linux (CentOS 5)
CFEngine
Enigineering + Operations (no QA)
· Engineers write, test, an deploy their own code
o Quickly make performance decisions
o Expose changes to subset of real traffic
· No ‘commit and quit’
· Deeply involved in moving services to production
· Ops ‘embedded’ into engineering teams
o Help make architecture decisions
o Better understand needs of product
o Interface with other ops teams
· Change logging – log all code changes
IRC
BitTorrent
Small teams
Facebook has over 60,000 servers: http://www.infoq.com/news/2011/01/facebook-coding-practices
InfoQ: Infrastructure as Code [InfoQ]
Panel: Theo Schlossnagle – OmniTI; Luke Kanies – Puppet Labs; Adam Jacob – Opscode; Erik Troan – rPath
Infrastructure Automation with Opscode Chef [pdf]
Joshua Timberman (Opscode, Inc.), Adam Jacob (Opscode), Christopher Brown (Opscode),Aaron Peterson (Opscode, Inc.), Matt Ray (Opscode), Seth Chisamore (Opscode, Inc.)
http://www.planetdevops.net/?m=201006
