Appendix: Comparison of Blue/Green Deployment Techniques - Blue/Green Deployments on AWS

Appendix: Comparison of Blue/Green Deployment Techniques

The following table offers an overview and comparison of the different blue/green deployment techniques discussed in this paper. The risk potential is evaluated from desirable lower risk (X) to less desirable higher risk (XXX).

Technique Risk Category Risk Potential Reasoning
Update DNS Routing with Amazon RouteĀ 53 Application Issues X Facilitates canary analysis
Application Performance X Gradual switch, traffic split management
People/Process Errors XX Depends on automation framework, overall simple process
Infrastructure Failures XX Depends on automation framework
Rollback XXX DNS TTL complexities (reaction time, flip/flop)
Cost X Optimized via Auto Scaling
Swap the Auto Scaling group behind Elastic Load Balancer Application Issues X Facilitates canary analysis
Application Performance XX Less granular traffic split management, already warm load balancer
People/Process Errors XX Depends on automation framework
Infrastructure Failures X Auto Scaling
Rollback X No DNS complexities
Cost X Optimized via Auto Scaling
Update Auto Scaling Group launch configurations Application Issues XXX Detection of errors/issues in a heterogeneous fleet is complex
Application Performance XXX Less granular traffic split, initial traffic load
People/Process Errors XX Depends on automation framework
Infrastructure Failures X Auto Scaling
Rollback X No DNS complexities
Cost XX Optimized via Auto Scaling, but initial scale-out overprovisions
Swap the environment of an Elastic Beanstalk application Application Issues XX Ability to do canary analysis ahead of cutover, but not with production traffic
Application Performance XXX Full cutover
People/Process Errors X Simple process, automated
Infrastructure Failures X Auto Scaling, CloudWatch monitoring, Elastic Beanstalk health reporting
Rollback XXX DNS TTL complexities
Cost XX Optimized via Auto Scaling, but initial scale-out may overprovision
Clone a stack in OpsWorks and update DNS Application Issues X Facilitates canary analysis
Application Performance X Gradual switch, traffic split management
People/Process Errors X Highly automated
Infrastructure Failures X Auto-healing capability
Rollback XXX DNS TTL complexities
Cost XXX Dual stack of resources