The major public cloud vendors have been throwing around the “Only pay for what you use” tagline as one of their main selling points since the concept was first introduced.
So when both Google Cloud and Amazon Web Services announced over the last few weeks that they’d be switching to a per-second billing model for some of their major services, many businesses were understandably excited. While Google Cloud and Microsoft Azure have offered per-minute billing up to this point, AWS have been a bit behind the 8 ball, only offering per-hour billing.
So, how can your business benefit from this?
- Dev/Test Workloads – chances are, these don’t need to be running around the clock. You may have already configured auto shutdown/startup for these virtual machines, which is great, but with per-second billing you truly are paying only for what you use – no more paying for 3 hours of compute if your EC2 instance is only running for 2 hours and 1 minute.
- Batch processing workloads – to use another AWS example, say you have a EMR cluster of 5 nodes and a job that runs for 10 minutes. Under the old per-hour billing, you would pay for 5 hours. With per-second billing, you only pay for 50 minutes – saving over 80%.
You can see how the numbers on this are looking more and more attractive.
It is worth noting, however, that this new model does not apply to all AWS and Google Cloud Services. Here’s a quick breakdown (per vendor) of the services making use of the new model.
EC2, EBS, Amazon EMR, AWS Batch, Elastic GPUs, and Provisioned IOPS for EBS volumes are all being offered on a per-second basis as of the October 2, 2017.
There are a few gotchas and things to note here, however. The first of these is that EC2 instances making use of Microsoft Windows Server and licensed Linux OS images like RHEL and SUSE do not offer the per-second billing. The official blog does say that this is not “currently available” – which does seem to suggest that, like Google Cloud, they will be offering this further down the line. The second thing to note is that Dedicated Per Region Fee, EBS Snapshots and instances launched from products in the AWS Marketplace also do not offer per-second billing.
Some services, like Permanent Disks, committed use discounts and GPUs, have been offered on a per-second basis for quite a while. As of September 26, 2017, these are joined by Compute Engine, Container Engine, Cloud Dataproc, and App Engine flexible environment VMs. What’s important to note here is that, unlike AWS, this does include virtual machines running licensed OS images like Windows Server, RHEL, and SUSE.
What’s most interesting about these changes is that they are likely to not only offer significant savings to customers but also to bring about a change in how we think about provisioning resources in a public cloud environment. We truly are entering a time where we “only pay for what we use”.
If you have any questions about what this new billing model could mean for your business or how to implement these services get in touch with us today.