New per-second AWS billing: making every second count

Amazon Web Services has now launched per-second billing, not only for their EC2 instances for also for EBS volume storage. What does that mean for you?

You might be jumping for joy – this could save you money after all – but it’s worth being wary. A wise man once told me to beware of money on the table; always ask yourself why it is there. Amazon won’t last long if the savings to customers end up eating into its own margins, so how can you make sure that you are one of the lucky ones to make the most of this change?

For long-running applications, the savings will hardly be noticeable and this, in part, is what Amazon will be banking on. With a one-minute minimum charge and an inevitable start-up overhead it’s still going to be good practice to group services that can be executed in sequence on the same instance.

The answer to making the most of this change is to stop going out to lunch.

To explain, I believe that a lot of cloud resources are wasted by the “going to lunch” effect. That’s when an engineer or service kicks something off that is not needed or finishes quickly and a manual step is needed to remove it.

Our customers usually find that storms happen on their compute clusters over lunch or at the weekend when less attention is being paid to what is being used. It’s hard to measure how much cloud spending goes to waste on unattended idle machines, but with reports of total wastage being about 45% there are clearly savings to be had.

There are plenty of solutions to identify long-running, underused compute and we are working on a new tool to optimise storage use, but this new billing policy from AWS opens up the forum to confident automated adjustment of resources that can spin down instances. This would save larger organisations significant percentages on their operational expenditure and reduce costs for smaller companies.

Even when your instance costs as much as a beer, turning it off is free – as in free beer!