Blog Post

Is Amazon the Hotel California of Web services?

Amazon(s amzn) CTO Werner Vogels had some words of wisdom Thursday for IT customers: “You should keep your providers on their toes every day. If we’re not delivering the right services you should walk away.”

No one watching the AWS Summit would disagree with that sentiment. But many questioned exactly how easy it will be for customers to walk away if they are fully engaged with higher-value-add Amazon services– the stuff that runs up the stack from the basic Amazon EC2 compute and S3 storage subsystems. And there are more higher value-add Amazon services seemingly every day (see chart.)

Vogels’ statement “has a good mouthfeel to it, but [walking away]  is sometimes easier said than done,” said Fred Holahan, CMO of VoltDB, a database provider that uses AWS. “Depending on the level at which you’re building software, you might be quite insulated from the details of the underlying infrastructure, but part of the objective of a platform vendor is to provide you with tantalizing services that lock you in.”

Holahan isn’t singling Amazon out — the same words apply to Microsoft(s msft), Google(s goog), Facebook — you name it. But Amazon, by virtue of its dominant position in public cloud infrastructure is worthy of special scrutiny.

As GigaOM has reported before, many of the newer AWS services including the AWS Storage Gateway, the Simple Workflow Services —  seem designed to tighten ties between the Amazon cloud and customer data centers.

Gavin Uhna, co-founder and CTO of, a maker of co-browsing software, agreed that the higher the level of integration, the tougher it is to move off of AWS.

“We started to notice this and began to make choices that would prevent us from getting stuck if we were to move to another provider (as we did). For example, we were deploying by storing a new release on an [Amazon Machine Image] and then launching new machines from it; We started to launch bare machines and run deploy scripts on them instead. Also, we stopped using the [Amazon Elastic Load Balancing] and started using a load balancer that we could deploy anywhere (we decided on HAProxy),” Uhna said via email. But, there was no problem at all moving off basic EC2 and S3 infrastructure, he added.

The overriding message to companies using any cloud infrastructure is: Be prepared. If you’re careful and smart,  you can stay ahead of the game.

“It’s not difficult to switch clouds if you plan appropriately,” said Marc Campbell, founder and CTO of, a maker of support plug-ins for mobile apps.  “If I was doing it today, I would get the new cloud up and running in parallel with AWS, writing data to both databases simultaneously.  When ready, it’s pretty straightforward to transition the new cloud to become the active data center,” he said.

11 Responses to “Is Amazon the Hotel California of Web services?”

  1. tnradmin

    Interesting read with one item I have been thinking about for awhile. I have implemented minimal AMIs – essentially the OS only based on my belief that provides for more flexibility. The one case where I find that belief tested is with auto scaling- whether using AWS auto scaling, our my own watchdog script to auto scale. The extra time required to perfom post launch installs and configurations negates the auto scaling to an extent.

    I’d love to hear other’s thoughts on this.

    • Mihir Patel

      Autoscaling gives you great flexibility for that. You just write a user-data script and hand it off to autoscaling, it would simplify post launch installations to great extent. I don’t know how sophisticated your watchdog script is, but if you want scheduling based scaling, policy based or metrics based scaling and saving some money, I think Autoscaling is the way to go for it.

      Mihir Patel

  2. Paul Calento

    Open vs. proprietary is an ongoing debate, particularly when moving from one provider to another … or between public-, private- clouds and back again. When you start using a public cloud service you often don’t think of the downside. Interoperability, central management, consistent security are all areas the OpenStack camp (and others) are addressing, but many folks don’t know what to look for until they need it. Lack of skills inhibits. For now, the takeaway is plan, plan, plan.