Auto Scaling Guided by Code
0 likes588 views
This document discusses auto scaling in cloud computing environments. It explains that auto scaling allows resources to scale up or down based on demand to provide an elastic architecture. It then outlines the key steps to implementing auto scaling, including: 1) creating a launch configuration with parameters like the AMI and instance type; 2) creating an auto scaling group to define scaling conditions and load balancers; and 3) creating scaling policies and alarm metrics tied to CloudWatch to determine when to scale instances up or down based on metrics like response time. The document emphasizes that auto scaling is guided by code through tools like Chef and Puppet to provision instances automatically.
Auto Scaling Guided by Code
- 1. Auto Scaling Guided by Code Pablo Cantero @pablocantero http://pablocantero.com 19/09/2012
- 2. Cloud Computing • Pay-as-you-go • Scale Resources up or down based on demand (Elastic Architecture) http://aws.amazon.com/
- 4. Instance provisioning • Manually • Image (preconfigured) • Automated • Chef • Puppet • CloudFormation http://aws.amazon.com/
- 5. Create a launch configuration. The launch configuration are parameters to launch instances, such as image (AMI), instance type, security group, key pair and user data.
- 6. Create an auto scaling group. The auto scaling group defines the auto scaling conditions, minimum and maximum number of instances, available zones, default cool down and load balancer (ELB).
- 7. Create a scaling policy. The scaling policy defines how the auto scaling should scale up or down, if it is by size or percent and the cool down.
- 8. Create an alarm metric. The alarm metric is the metric used to decide when to increase or decrease the instances. Usually by response time. The metrics are provided by CloudWatch.
- 10. Create an alarm request. It is the last step to setup the Auto Scaling. The alarm request defines how to measure the metrics defined by the alarm metric, for example the time and period to evaluate the metric.
- 11. Talk is cheap. Show me the code.
- 12. Some advices
- 13. Don’t name your instances You should change your instances as you change your underwear. Lucas Teixeira
- 14. Don’t forget to remove your tests
- 15. Use Auto Scaling, avoid Cloud Smells http://pablocantero.com/blog/ 2012/09/07/use-auto-scaling- avoid-cloud-smells/
- 16. Thank you Pablo Cantero @pablocantero http://pablocantero.com