Autoscaling is table stakes for public cloud providers. Now AWS customers can apply it to their EC2 Fleet instance groups for improved control over cost and performance.
Every AWS customer wants scalability, availability, capacity and performance -- all at the best price. They also expect AWS to continuously improve on all four fronts.
Amazon EC2 Fleet, released in May, helps users assemble collections of EC2 instances that meet specific workload and cost demands. It's geared toward busy IT shops that want to automate as many routine tasks as possible. The service's predecessor, Spot Fleet, involved only Spot Instances, but with EC2 Fleet customers can add On-Demand and Reserved Instances to the mix.
The introduction of AWS Auto Scaling Groups for EC2 Fleet this week expands the service's appeal for AWS customers. They can now create fleets that scale up or down as workload demands change, and stick with the most economical mix of instances, said Jeff Barr, AWS chief evangelist, in a blog post.
Users generate fleets from the AWS Command Line Interface, EC2 Console or API, and specify and prioritize the desired combination of instances. They can control the type of on-demand instance -- the priciest flavor offered by EC2 -- and when they need to scale up. They can also specify availability-zone diversity for Spot Instances, which are generated from spare capacity and can cost much less than on-demand or reserved instances.
Customers can tweak settings at any time after fleets launch. AWS' autoscaling capability also provides lifecycle features, such as health checks, which scan instances for problems periodically. The system can automatically terminate and replace problematic instances.
You never use the same machines for everything. [Autoscaling] allows you to orchestrate more independently.
Holger Muelleranalyst, Constellation Research
The autoscaling functionality within EC2 Fleet is available now in all AWS commercial regions at no additional charge, Barr said. Autoscaling support for CloudFormation, AWS' infrastructure-as-code management tool, will be available shortly.
The ability to customize a fleet based on instance size, and balance on-demand versus spot pricing, helps improve cost efficiency, said Kevin Felichko, CTO of PropertyRoom.com, an online auction site based in Frederick, Md. "If an upgraded instance type is available at an ideal spot price and can reduce the overall job size, then we can scale back any on-demand instances which were fired up," he said.
The addition of autoscaling to EC2 Fleet isn't revolutionary, but it's the type of functionality modern IT shops need, said Holger Mueller, an analyst at Constellation Research in Cupertino, Calif. Resource elasticity and cost reduction are main drivers for companies that move newly built applications to public clouds.
"To achieve that, DevOps pros need tools, and this makes autoscaling a key feature," he said.
More control over public cloud resources is also crucial given the diversity of cloud customers' workloads. Mueller cited the example of an IoT implementation that monitors and collects data from a wind turbine, where workloads change dramatically as weather conditions vary. In much the same way, internal corporate applications can see usage spikes as employees arrive at work and log on, and leave at the end of the day.
"You never use the same machines for everything," he said. "[Autoscaling] allows you to orchestrate more independently."
