SmoothSpan Blog

I’m reading with interest as Dare Obasanjo and Tim Bray try to make the case that there is lock-in with Amazon Web Services and Google’s AppEngine.  The lock-in argument from Bray’s end basically boils down to the API’s being so heavily embedded in your application that it’s hard to switch out of them or to the unavailability of commercial alternatives ready to go now.  Obasanjo doesn’t add much to the argument except to point out that any cloud apps you use (such as Google Apps or Zoho, competitor’s to his Microsoft Office apps) need to be able to ship their data without loss to other cloud apps, a point I do agree with.

So what’s up with all this?  Is the Amazon Cloud fraught with lock-in, or not?

Frankly, I don’t see Bray or Obasanjo’s arguments as holding much water.  I recently wrote that lock-in can be avoided for Amazon if you focus on buying virtual hardware and leave the more proprietary aspects of the service out of you plans.  Neither of these new posts brings anything of substance to that earlier discussion.

It took my company, Helpstream, a grand total of an afternoon to make our app run on Amazon.  The only real missing piece vis a vis our datacenter implementation is that we hadn’t pulled the trigger to point the DNS at the Amazon implementation.  It’s pretty hard to call that locked in.  Perhaps if we had focused too much on Amazon’s SimpleDB or other more proprietary services, we would have a problem, but we specifically have not.  As I explained in the other post, we want to take maximum advantage of commodity computing, and that means we have to avoid lock-in.

There is another argument that has been raised that even if the API’s are portable, it doesn’t matter until a real commercial alternative is ready to go today.  I see this as equally groundless.  Not only are the API’s easy to retarget (they are so simple you don’t need transparently compatible API’s), but there are commercial alternatives available today.  They range from other cloud vendors such as IBM and Sun, both of which offer virtual machines ala EC2 and virtual SAN ala S3, to various alternatives provided by hosters.  Or, you can just go back to having your own data center.  It would be expensive, but that’s what you’re stuck doing anyway if you don’t adopt the Cloud, so why are we holding that as a point of lock-in?