A steep discount to AWS Fargate pricing aims to lure more container workloads to the cloud, anchor AWS' foothold in this maturing market and fend off rivals such as Microsoft Azure.
Amazon released the Fargate container management service in November 2017 for customers that want to deploy containers on AWS but don't need or desire control over the underlying infrastructure. It is similar to the likes of Azure Container Instances, which Microsoft made generally available in the spring of 2018.
AWS Fargate pricing is based on vCPU usage and gigabytes of memory per second. Amazon has lowered the costs of the former by 20% and 65% for the latter, according to an AWS blog post. Customers could save up to 50% overall, depending on the ratio of CPU to memory they use for applications, AWS said.
The industry is amid a hype wave around containers, so the hefty price cut is likely a tactic to appeal to customers to try its newer service rather than those of Azure or other competitors.
Owen Rogers
"AWS has slashed the price to encourage its own customers to experiment more with Fargate and also to show the market that AWS has container credentials that, it believes, beats its competitors," said Owen Rogers, an analyst at 451 Research based in New York.
The cost cut is significant, but most enterprises aren't at the stage to optimize for cost, added Gary Chen, an analyst at IDC.
"Right now, it's important for customers to start building container expertise," he said. "For most customers who are in large-scale production, [their] top concern is getting used to the technology, getting expertise on a new application architecture."
Initially, some customers questioned if the benefits of such a container management service would outweigh the costs. With the AWS Fargate price cut in hand, that analysis is worth revisiting.
Line blurs between containers and serverless
Containers have grown popular in recent years, and one reason is they are lighter-weight than virtual machines (VM) and can be spun up and down more quickly. But the container model strips out the hypervisor layer, which means containers don't benefit from the security and workload isolation features found in VMs. MicroVMs have a best-of-both-worlds approach, using a minimalist design to provide more speed while maintaining the benefits of VMs.
Gary Chen
AWS developed Firecracker, an open source technology that supports microVMs, for internal use underneath Fargate and its Lambda serverless compute offering, and released it at re:Invent in November. But Firecracker isn't the only technology of its kind -- Microsoft Azure has Hyper-V containers and Google offers gVisor.
Typically, cloud providers price their services based on how they think they can reduce the cost for users, through software optimization or via Moore's law, Chen said. Firecracker, basically an optimized container in a VM model, is the type of technological leap that indicated AWS has become more efficient, he said.
One thing to watch is whether AWS similarly slashes prices for Lambda, which lets developers write and execute code without the need to manage underlying servers.
Right now, it's important for customers to start building container expertise.
Gary Chenanalyst, IDC
In broad strokes, the objectives of Fargate and Lambda are higher cost savings, speed and efficiency. Thus, customers should expect these prices to converge over time, said Ryan Marsh, a software development trainer in Houston, who also works as an evangelist for software testing tools vendor Xolv.io.
"Eventually, the price difference will be so small that most organizations will decide based on architectural needs, rather than perceived cost," Marsh said. "What is a Lambda but a tiny container? To AWS or Google, it's just compute. The difference is in resource utilization and extracting the most rent from their data center hardware at the most competitive prices."
One might think functions have added resource management overhead relative to containers, but that's not necessarily true, Marsh added. "The fact is we don't know," he said. "Few people are running their own hosted container and hosted function services for us to compare."
Owen Rogers
Gary Chen
