1 DevOp vs 1.000 servers - Amazon EC2 and Chef automation intro
- 1. DevOp1 Admin vs. 1.000 Serversphpunconf Hamburg 2011Thomas Metschke@tmetschke2011/09/10Peritor GmbH
- 2. Peritor GmbHConsulting, development and workshops since 20052DatabasesSelectionIntegrationMigrationAutomationDeploymentWorkflowsClustersRuby & RoRCode ReviewSecurity ReviewWorkshopsDevelopmentPerformanceAnalysisOptimization
- 3. 3
- 4. EC2 & EucalyptusCluster ManagementAutomatic ConfigurationSelf Healing ClustersAuto Scaling ClustersOne Click DeploymentOne Click Cluster CloningMonitoring & AlertingLogs & AccountingFirewalls & BackupsAccess & Rights Management4
- 5. Project life cycleMaturity &DeclineProduction& GrowthConcept & DevelopmentNiche /EOL5
- 7. 7
- 8. Amazon Elastic Compute CloudVirtual servers via API callPay per hourDifferent instance types in different regions 8
- 9. Regions and Availability ZonesRegionsMultiple Availability ZonesSlightly different costsAvailability Zones4 x US-east – Northern Virginia3 x US-west – Northern California3 x EU-west – Ireland2 x AP-southeast – Singapore2 x AP-northeast – Tokyo9
- 10. Server types10
- 11. Amazon EC2 additional services11Elastic Load BalancingVirtual Private Cloud Elastic IPCloudWatchSpot instancesElastic Block StorageAnd a lot more …
- 12. Amazon EC212
- 16. Amazon EC2 is16Infrastructure as a Service
- 17. What you can build fast17
- 18. But what if …18
- 20. Scaling up and downManualTimeAPILoad20Different triggers make it possible to scale any part of the infrastructure or create new clusters
- 21. Deployment21ClusterApplicationHandle deployments of applications out of source control systems or via http downloads. Allow update code on running servers. User management and GUI.
- 23. Healing
- 24. Scaling
- 27. ChefOpen Source Provisioning ToolConfigures your serversCookbooks & recipesRuby DSL24
- 32. Attributes29cookbooks/memcached/attributes/default.rbSet defaultvaluesfor variables to use in recipes & templates
- 36. /etc/chef/solo.json (configuration JSON)Chef Solo Run33Run Chef solo
- 37. Let’s talk about cute little Monsters34
- 38. Monster World1.6 million DAU8.2 million MAUPeak 130 servers~ 240.000 requests per minute on average35
- 39. Basic architecture36App ServerApp ServerEBS VolumeLoad BalancerApp ServerApp ServerDBsApp ServerApp ServerEBS VolumeApp ServerSlave DBsApp ServerAll configuration, management and deployment through Scalarium.No other tools needed.S3
- 40. Flexibility and Scalability1:5 facebook1:20 local social networks
- 41. Flexibility and ScalabilityCost savings of ~ 40%!
- 42. DEMO?39
- 43. Want to try ScalariumSignUp onhttp://scalarium.comand use it for free as long as you are in test & dev modeChef examples and guideshttps://github.com/scalarium/example-cookbookshttp://wiki.opscode.com/display/chef/Resources40