Monitorama was a two-day conference/hackathon that focused on tools and practices for monitoring/managing complex web applications and systems. The first day was 12 30-minute sessions from a variety of experts, while the second day had 6 more in depth workshops. You could also participate in a hackathon the second day, and I would guess about 30% of the people did that while 70% attending the workshops.
graphite / statsd
nagios (eveyone hates)
logstash
github
pager duty
new relic
Think of instruments / measurements as unit tests
Unit tests tell you how your system should work, instruments tell you how it is working in production
When an outage / failure happens, add a unit test
Instrumentation costs happen up front / while unit test costs typically happen over the long term
Data != Information, Models + Data = Information
KPI / "soul" metrics / context / audience
Monolithic Systems vs Composable Services
Dealing with Outages
Data Mullet (Oracle in the Front, NoSQL in the back)
Alerts should provide more context
Systems should learn over time
Don't try to remove the human from the system, figure out ways for them to interact better
Tech I wasn't familiar with, will save me money...
VMs + configuration management
Development Environments as a Service
Setup a graphite/statsd environment in 10 minutes.