ELB and ASG
We previously briefed the “cattle in a herd” analogy in cloud computing. In this section, we will explain the actual implementation using ELBs and ASGs and use an example to illustrate the mechanism.
An ELB automatically distributes the incoming traffic (workload) across multiple targets, such as EC2 instances, in one or more AZs, so as to balance the workload for high performance and high availability (HA). An ELB monitors the health of its registered targets and distributes traffic only to the healthy targets.
Behind an ELB, there is usually an ASG that manages the fleet of ELB targets—EC2 instances, in our case. ASG monitors the workload of the instances and uses auto-scaling policies to scale—when the workload reaches a certain up-threshold, such as CPU utilization of 80%, ASG will launch new EC2s and add them into the fleet to offload the traffic until the utilization drops below the up-threshold. When the workload reaches a certain down-threshold, such as CPU utilization of 30%, ASG will shut down EC2s from the fleet until the utilization rises above the threshold. ASG also utilizes a health-check to monitor the instances and replace unhealthy ones as needed. During the auto-scaling process, ASG makes sure that the running EC2 instances are loaded within the thresholds and are laid out across as many AZs in a region.
Let us illustrate ELB and ASG with an example. www.zbestbuy.com is an international online e-commerce retailer. During normal business hours, it needs a certain number of web servers to work together to meet online shopping traffic. To meet the global traffic requirements, three web servers are built in different AWS regions—North Virginia (us-east-1), London (eu-west-2), and Singapore (ap-southeast-1). Depending on the customer browser location, Amazon Route 53 (an AWS DNS service) will route the traffic to the nearest web server: when customers in Europe browse the retailer website, the traffic will be routed to the eu-west-2 web server, which is really an ELB (or Application Load Balancer (ALB)), and distributed to the EC2 instances behind the ELB, as shown in Figure 1.16.
When Black Friday comes, the traffic increases and hits the ELB, which passes the traffic to the EC2 instance fleet. The heavy traffic will raise the EC2 instances’ CPU utilization to reach the up-threshold of 80%. Based on the auto-scaling policy, an alarm will be kicked off and the ASG will automatically scale, launching more EC2 instances to join the EC2 fleet. With more EC2s joining in, the CPU utilization will be dropped. Depending on the Black Friday traffic fluctuation, the ASG will always keep up to make sure enough EC2s are working on the workload with normal CPU utilization. When Black Friday sales end, the traffic decreases and thus causes the instances’ CPU utilization to drop. When it reaches the down-threshold of 30%, the ASG will start shutting down EC2s based on the auto-scaling policy:
Figure 1.16 – ELB and ASG
As we can see from the preceding example, ELB and ASG work together to scale elastically. Please refer to https://docs.aws.amazon.com/autoscaling/ec2/userguide/autoscaling-load-balancer.html for more details.
