Welcome to another essentials post with AWS Compute Architecture !! AWS Compute is about using provisioning EC2 servers for implementing functional aspects of any business requirement.
This post is about “How to get the maximum out of EC2 services but yet save huge on costs(given the workload can accommodate Spot interruptions) ?”
Need #1: Application hosted on AWS now meets with the business requirement but off late has gathered momentum and gets sudden demand with peaked requests. Existing design has to accommodate varying workload scenarios and should be capable of scaling up & down automatically
Autoscaling & Application Load Balancer together helps in making the compute capacity available for peak loads and ALB assures that traffic or requests are routed to healthy instances available inside an auto scaling group
Need #2: EC2 instance attribute types like OS/AMI or memory, storage, network parameters have to be rightly defined & used consistently
Launch Templates has to be used in order to retain multiple versions of Instance attributes combination(by using versions) and can be used for provisioning as needed along with Auto Scaling Group
Need #3: Capacity scales as needed during peak/off peak timelines with on-demand instances but we need best optimum compute usage to save costs
Creating EC2 Spot Fleet requests & spot instances(for workloads supporting interruptions & resuming) will make the implementation super successful & efficient by giving upto 90% discounts on compute usage
Need #4: Incoming traffic should be routed to all healthy instances that are available at any point in time
Definition & using of Target Groups will ensure the health check of the instance & redirects the requests from ALB as accordingly. Also helps in grouping mix of instance types like on-demand & spot definitions for a highly available & fault tolerant Application
For a focused explanation on spot & benefits rather an architectural blend, then visit Link
For AWS Compute architecture using all on-demand compute capacity please refer to Link