Per Second Billing for EC2 and EBS
On September 18th, 2017, [Amazon announced] (https://aws.amazon.com/blogs/aws/new-per-second-billing-for-ec2-instances-and-ebs-volumes/) that effective October 2nd, 2017 all Linux-based EC2 instances running as On-Demand, Reserved or Spot, as well as provisioned storage for EBS volumes, will be billed in one-second increments; per-second billing also applies to Amazon Elastic Map Reduce (EMR) and AWS Batch jobs.
Elastic Map Reduce (EMR)
With the ability to add more machines to the cluster, jobs can be completed sooner and will be more cost-effective.
AWS Batch
Most batch jobs complete in under an hour, so knowing the billing rate is down to the second and that AWS Batch uses spot instances to complete the tasks, using this service will be even more economical.
So what does this mean for customers who use EC2 and EBS for their day-to-day tasks; greater flexibility of their budget and potential cost-savings for short-lived batch jobs. Traditionally EC2 and EBS were always billed by the hour; meaning if an instance and an attached EBS volume were running for any time under an hour, the usage time was always rounded up to the next hour (i.e. if the instance and EBS volume were running for 1 hour and 5 minutes, they would be billed for 2 hours each).
With per-second billing, customers who run smaller, temporary jobs can leverage the ability to provision more EC2 instances to get the job done quicker since they don’t need to budget for the highest hour the job will take to complete. This greater flexibility means a customer could complete their tasks faster, which allows them to get to market sooner, which means less time working on jobs internally and more time engaging with their clients.
Example Scenario
Auto Scaling
You have an online shopping site that recently sent out a 30-minute flash sale notice to all your customers. The website starts to take on a heavy load, and you need to scale out to handle the traffic. With per-second billing, your profit margins will be higher because you will be billed down to the second for the auto scaling during the flash sale.
AWS Batch
You’re given a fixed budget to run batch jobs; which previously took slightly over 5 hours to complete on a m4.xlarge, but you were being billed for 6 hours due to the previous pricing model. You’ve calculated that the batch job could be completed in less than 4 hours if you scaled up to a m4.4xlarge, and now with per-second billing, you’ll save money. Meaning you get the job done sooner, and for less money.
Things to take away
This change is going to be applied to all AWS regions effective October 2nd, 2017, for all Linux-based instances currently running and newly launched, but note there is a 1-minute minimum charge per running instance. This billing change does not apply to any Windows-based instances or any Linux distributions with a separate hourly cost. The Dedicated Per Region Fee, AWS Marketplace products, and EBS Snapshots are still billed on an hourly basis.
All list prices and spot market prices will still be listed at the usual per-hour rate, but billing will be on a per-second basis. Bills will show the normal hours of usage, but now with added decimal digits to reflect the per-second rates (i.e. 384.115 hours @ USD $0.016 per hour).