1 Comment

Summary:

New from AWS: the ability to change the size of your EC2 Reserved Instances. That along with an earlier change that lets you move RIs between zones show Amazon getting more flexible.

AWS: Reinvent
photo: Barb Darrow

If you don’t like restrictions on the use of Amazon Web Services’ Reserved Instances, wait a minute, things are changing.

Starting now, for example, Amazon will let users change the size of their reserved compute instances — within a size group at least. According to the AWS blog post announcing the change:

…you can change a reservation for four m1.small instances into a single reservation for an m1.large instance. With today’s launch you can take advantage of the pricing and capacity benefits of Reserved Instances even as you change from one EC2 instance type to another.

Before, if you signed up for a given size of Windows or Linux instance for the one- or three-year term of use, you were married to that size even if you no longer needed it or if you outgrew it. Or you could sell your unused RIs on Amazon’s spot market rolled out last year.

This change comes less than a month after AWS started letting users of RIs move them between availability zones (AZ), as long as the zone was in the same region. In the past, if your RI deployed in a given AZ, your RI remained in that zone come hell or high water.

AWS users said these tweaks are helpful and also show that AWS faces more competition as Google Compute Platform, Microsoft Windows Azure and other public clouds get more mature. And some of those competitors — Microsoft and Google, for example, offer sub-hour pricing on compute resources, whereas Amazon continues to charge by the hour. Microsoft charges by the minute, as does Google, although Google requires a 10-minute minimum.

Earlier this week, Microsoft also said it would give existing volume licensing customers discounts to use Windows, although details are scant. Microsoft, late to cloud, is starting to get its act together there now that it has IaaS capabilities analogous to what AWS offers.

That may be a long-term concern for AWS, although executives there have said they see no demand for sub-hour pricing increments. But as we know, never say never.

Joe Emison, CTO of Buildfax and a long-time AWS user and watcher, said the way Amazon, which he characterized as a great service overall, handles RI pricing is something that could be disrupted by a competitor if and when that competitor offers a comparable service

He’s not convinced that many current AWS customers would switch to any of the competitors now. But, “in the not too distant future” some of these rival clouds will be viable enough to pull customers away, he said via email.

One thing we’ve learned from AWS though, it doesn’t rest on its laurels, so stay tuned.

  1. This new flexibility enhancement is very good. The next logical step is to remove the RI number altogether and to let people express RIs needs in terms of “Compute power” (and instance generations to let AWS do it’s capacity planning).

    Regarding “Charging by the minute”, it’s very interesting for short-lived instances; but it becomes less and less interesting as the instance life duration increases.

    And for these short-lived instances, the EC2 Spot market can get you extremely good prices.

    So I guess it’s more important for Google and Microsoft to come up with a Spot Market alternative than it is for AWS to provide by-the-min billing (even if this would obviously be a welcomed change).

    Share

Comments have been disabled for this post