Weekly Update

Why Google Fail ≠ Cloud Fail

It has been a rough week for the mighty Google, whose relatively brief outages have been analyzed at a level generally reserved for presidential elections – even bringing some conspiracy theories out of the woodwork. Not surprisingly, cloud computing as a paradigm continues to catch a lot of shrapnel from these attacks on Google. But if you ask me, that’s undeserved: Google’s basic suite of cloud services are not the equivalent of commercial-grade SaaS offerings, nor is Google’s underlying architecture a cloud of the ilk large companies would use, and they should not be talked about as such.

Google’s outages might have been annoying, but they were not proof that cloud computing is inherently unreliable. Sure, the first one reportedly caused a 5 percent drop in Internet traffic, but Google attributed the routing problem to human error – which is a problem in every in-house data center on the planet. And the word is Google is beefing up its support team as a result. Of course, failures of some sort are inevitable regardless the provider or the service, so the best bet for anyone buying cloud-based services is to figure out a strategy for mitigating the damage (e.g., a hybrid environment), lest you run the risk of “some monkey pushing the wrong button.” (If you want to see a real cloud-computing failure, check out the case of FlexiScale, which is just coming back online after a complete rebuild of its infrastructure.)

Some are even treating attacks on Google’s search results as evidence of the vulnerability of cloud computing. I have faith (possibly misplaced) that Google and other cloud providers know enough to cordon their customer-facing cloud infrastructures apart from their public-facing web infrastructures. While Amazon EC2, for example, might be based on what Amazon developed to optimize its retail business, Amazon.com does not share an infrastructure with Amazon Web Services. (Speaking of Amazon EC2, it just buried its competition, not to mention a slew of value-add services, by announcing monitoring, auto-scaling and load-balancing capabilities.)

Additionally, we shouldn’t expect the reliability of free services like search and Gmail to be equal to the reliability of paid services like App Engine or the application suite Google sells to businesses. I trust that Google, as well as its cloud cohorts, understands the importance of treating paying customers with the quality of service they deserve (and contract for). When the services for which Google takes a customer’s money go down (or paid services from Amazon or Rackspace or Saleforce.com … again) I’ll be more concerned.

Yes, Google – the company that believes it can save the world – deserves the flak it is receiving, but we don’t need to take aim at the distinctively separate, commercially-oriented cloud-computing market, too.

Question of the week

How will Amazon’s new management features affect the cottage industry of startups building tools to manage its products?