DevOps results - conference season 2014

Wrapup from Velocity, DevOpsDays and DevOps Enterprise 2014 at DevOpsTO

Transcript

1. @steveElsewhere Steve Pereira
2. DevOps Results November 2014
3. Conference Season Recap
4. DevOpsDays Toronto 2014 Chicago and Belgium Velocity Full Stack Toronto
5. DevOps Enterprise
6. Puppet 2014 State of DevOps survey results 30x more frequent deployments 8000x faster lead times vs peers 2x change success rate 12x faster MTTR 2x more likely to exceed profitabilty, marketshare, productivity goals 50% market cap growth over 3 years
7. Big Results
8. Ticketmaster 17 disparate ticketing systems > 1 Teams reorged around delivery - devs on call 6 week deploy > 6 min 109 deploys /month 30% revenue increase
9. Target > 30 APIs 80 deploys /week - 1884 jenkins jobs 1.5 billion visits /month < 10 incidents /month 270 cookbook repos 134 engineers + 30 teams
10. Intuit 165 discrete experiments during peak season Conversion up 50%
11. Macy's Development cost down to 55M /year from 100M 140% increase in products under development Capacity for innovation from ~5% to ~40%
12. Nordstrom 63k employees 1800 tech Annual planning to sprints Features /month up by 50% Bugs /month down by 50% 22 week releases to and semiweekly
13. IBM Life Insurance client Dev productivity up 1600% 75% of teams converted to agile from waterfall 11 months early developing major product
14. Nationwide 5 year span 80% reduction in critical defects 86% reduction in high defects 90% on time delivery vs 60% 70% increase in product availability
15. Orbitz 10+ year old codebase Unified VCS, JDK versions and Frameworks Switched to product teams Dev/Prod parity Reduced overall server count by 25% Reduced search server count by 90%
16. IDC Global 1000 forecast for 2yrs Avg deployment frequency will double 20% faster time to market 50% have DevOps focused task force
17. Blackboard 6 month lead time to 1-2 weeks 3 month testing cycle to < 30min acceptance test time 6 hours value time from 70 hours total = 64 hours of lag
18. CSG 40 dev teams 50M customers 100k call centre seats 6B transactions /month 20 tech stacks 50 apps
19. CSG Batch size down from 28 to 14 weeks 10 days recovery to 2 days Redeploying reduces recovery to 1 day 70 deployments to Dev/Test prior to Prod
20. Mitre 15 days to 9 days total cycle time 2 mins deployment 65% reduction in manual deploys 38% increase in thoroughput Freed up 2.5 deployment staff
21. Huawei Dev: 20min > 2min test per 10M LOC Prod: 300min > 10min test per 10M LOC Regression 240min > 60mins Full suite 24h > 6 hours Feature cycle time 30 days > 7 days
22. Huawei 480K code reviews / day 100K builds per / day 100M test cases run / day 10K releases / day
23. Atlassian Before Manually maintained unique servers Started with Puppet
24. Atlassian Actions Started using Vagrant + Puppet-lint BDD + Continuous Integration AWS for branch integration testing ChatOps integration and automation
25. Atlassian 90K deploys a month 43K tests just for JIRA 1000 automated build servers
26. US DHS Compliance drives testing requirements 1 year release cycle for a simple change
 to Running Chaos Monkey
27. US DHS Running Chaos Monkey
28. Holistic Results ▲ Increased job satisfaction ▲ Increased visibility + estimation ▲ Increased accountability + confidence ▲ IT as a competitive advantage ▲ Increased alignment
29. Holistic Results ▼ Decreased MTTR/Cycle time ▼ Decreased disruption ▼ Decreased risk
30. Observations • Rock bottom required • Successful starts with business objectives • Fear of satisfying reg and compliance • DevOps is the only way to tackle legacy • Agile and Lean foundation • 3-6 year process
31. Common Resources Books The Goal The Phoenix Project Continuous Delivery Flaw of Averages Lean Startup / Lean Enterprise Toyota Kata Online 10 Deploys A Day - Dev and Ops collaboration at Flickr Deming’s 14 points Agile Infrastructure
32. DevOpsDays 2015
 Bigger + Better
33. Thanks! @steveElsewhere Steve Pereira