In early May, AWS launched a new feature called the EC2 Fleet that makes it less complicated to assign a specific capacity to different EC2 instances. The fleet contains information that gives it the power to launch a group of EC2 instances with a single API call.
It can combine the Reserved Instance, On-Demand Instance and Spot Instance to launch more than one instance type across multiple Availability Zones. Developers don’t have to create custom codes when they need to monitor prices or discover capacities. This ability means that you can structure a fleet to suit the cost, performance and scale that satisfy your application.
Using The Fleet
The new feature from Amazon Web Services is an extension of Spot Fleets and does most of the heavy lifting. All you only have put in a request with the necessary capacity and instances then the fleet does the rest.
Besides instances, customers can define their required capacities as application-oriented units and vCPUs. For the fleet to complete the request successfully, you must indicate what number of EC2 Spot Instances are needed. With all the relevant information, the fleet will handle the launching, managing, tracking and scaling of EC2 instance types.
With this feature, you can accomplish a host of different tasks from genomic processing to running a big data cluster. EC2 provides considerable elasticity, which offers an opportunity to experiment with different pipelines on thousands of servers.
When sending requests, you don’t have restrictions on the number of instances that you can specify per EC2 Fleet. Both Spot and On-Demand purchasing options can be used to provision the instance types that you identify in your request. A user can also indicate the Spot price per instance, the Availability Zones and the Spot options for every fleet.
The fleet is designed to pick a combination of Availability Zones and EC2 instance types that offer the greatest cost-effectiveness. For Spot Instances, the fleet refers to the current prices while for On-Demand Instances, it uses the public prices.
During the fleet operation, if a Spot Instance fails or spikes in cost, causing Amazon EC2 to reclaim it, the Fleet picks a replacement from one of the instances that you specified. It helps to have a flexible resourcing strategy when it comes to instance specification.
For example, you can request that a particular fleet use On-Demand Instances as the primary capacity, and if necessary, supplement them with Spot Instances. However, if a customer has Reserved Instances, the fleet will use those rather than the On-Demand that you specified.
Take a look at the limitations of EC2 Fleet
EC2 spot Fleets are not without limitations and customers should be aware of them beforehand. Firstly, you can only access the feature form the AWS CLI or the API. For individuals who intend to use the feature in different regions, a separate fleet is required for every area.
Note that using the new feature from Amazon doesn’t result in additional charges; only the instances launched by the fleet. The fleet is a natural progression from EC2 Spot Fleets and negates the need to develop glue codes for multiple API calls because one call can launch a sizable number of instance types. All public AWS Regions can use this new feature.
This ebook could be of your interest: