Platform as a service and redundancy: Nagging questions

The recent Amazon Web Services outage along with other cloud mishaps of late is raising questions about platform as a service maturity, lock-in and customer recourse.

The recent Amazon Web Services outage along with other cloud mishaps of late is raising questions about platform as a service maturity, lock-in and customer recourse.

Since the AWS outage, I've been pondering platform as a service. With AWS, which is essentially infrastructure as a service, it's not-too-difficult to imagine having another provider as a backup. Commodity APIs mean it's at least theoretically possible to swap in Rackspace for AWS. Or in the future it's HP's public cloud. With the right cloud management tools it's should be a relatively easy swap.

As you move up the cloud stack the outage worries pick up. For instance, let's lay out the following scenario.

  1. An enterprise picks a platform as a service provider---Microsoft's Azure or Salesforce.com's Force.com for instance.
  2. That platform works fine most of the time, but...
  3. It happens to blow up for a few days just like AWS did.
  4. Enterprise is locked into a platform that doesn't work.
  5. Platform as a service providers to date can't talk to each other. These platforms are proprietary. You can just swap Azure to Force.com. And guess what you probably never will. You're betting on a platform, a language and that company's ability to keep running.

BMC CTO Kia Behnia says that the scenario outlined above does highlight the immaturity of platform as a service today. "Infrastructure as a service is a bigger market and use case," says Behnia. "PaaS is growing up. It will have its place in focused areas but you can't just tell enterprises to part with their code and trust the vendor. PaaS is in its infancy."

Okta CEO largely agrees. Okta survived the AWS outage because it was architected to fail over to other parts of AWS. With IaaS it's clear that reliability depends on the provider. "As you move up the stack it's not as clear," says Todd McKinnon. "What is the SLA for platform as a service?"

Those nagging questions continue. A bevy of enterprise execs at SAP's Sapphire said that they were wary of platform as a service, but they had no worries about infrastructure in the cloud. Perhaps these IT buyers will come around. Today, they remain wary.

Ultimately, developers will vote with their code, but platform stability will matter big time. And PaaS vendors will have to demonstrate they have an extremely reliable system.