Amazon Web Services’ new Spot Instances pricing model has created quite a buzz among commentators wondering what it means for the future of cloud computing. Some suggest it signals the advent of a commodity market for cloud resources, where CPU prices always are fluctuating based on what buyers are willing to pay. The consensus seems to be that cloudwide market pricing would be a good thing. Indeed, AWS spot pricing probably is the first step toward major innovations in CPU-buying – perhaps well beyond simple spot pricing – but it’s difficult to see significant change coming anytime soon.
The reason is that AWS holds a commanding share of the infrastructure-as-a-service market, and commoditization isn’t in AWS’s best interests if it wants to continue seeing its current profit margins. Commoditization requires, at the least, interoperability among clouds to the point where applications and data can migrate as the market price dictates, and Amazon is under no obligation to support any such standards when they do materialize.
Maybe it doesn’t have to; while VMs might be interchangeable, the feature sets that surround them are not. For example, GoGrid this week announced the availability of GigaSpaces-powered instances. Rackspace announced the availability of FathomDB’s relational database as a service on Rackspace Servers. Rackspace doesn’t have GigaSpaces images, and GoGrid doesn’t have a database as a service. AWS offers its own relational database service and offers GigaSpaces instances – as well as a slew of other homegrown features and third-party instances. This week, it announced streaming media support on its homegrown CloudFront service.
Until customers can apply the same capabilities to their applications on any cloud, it is difficult to see anything but the most static, low-performance workloads jumping from cloud to cloud. As long as AWS continues to offer the most features, all of which are designed to work with each other, it might add enough value to justify elevated pricing, even if competitors move to commoditize the market and drive per-CPU prices toward zero.
We might be wise to question whether a true cloud-computing commodity market even would be such a good thing. For starters, there’s the problem of budgeting for cloud resources when nobody really knows what they will go for at any given time. If a budget plans for standard cloud capacity in the $.01- to $.03-per-CPU range, for example, a jump to $.05 could have serious implications. It might be a tenuous analogy, but look at what has happened when the world got accustomed to cheap gasoline, or the way increased corn prices have affected certain geographies. By comparison, guaranteeing a maximum price with AWS (or any other provider) might not be so bad, even if it does involve a certain level of lock-in and even if spot pricing does trend toward the high end.
Another problem with commoditization is that it kills innovation. Unless cloud providers drastically evolve their business models, commoditization will slash profit margins to the point where providers stop investing in new features and focus all their energy on reducing delivery costs. When was the last time a better version of wheat hit the market? Perhaps innovation will live on in the PaaS sector (where some suggest AWS is heading), but those utilizing IaaS offerings will be out of luck. Unless, as noted above, AWS can continue to reap acceptable margins through a high-value IaaS offering, that market will be all about delivering as much volume as possible with as little overhead as possible.
The IaaS sector might well become a commodity in the years to come, but it will take some major changes to make it happen, and it might not be all it’s cracked up to be.