Here are the top seven developments we saw in public cloud computing in 2012: The three biggest setbacks, and the four biggest wins. We'll start with the setbacks.
Setback #1: Outages Plague Amazon And Others.
When you're trying to convince big companies to bet their business on cloud operations, the worst thing that can happen is for the cloud infrastructure you're thinking of using to suffer an unplanned outage. Amazon Web Services didn't have an outage in 2012 that rivaled the hit it took over the Easter weekend in April 2011, when multi-availability zones in one of its data centers went down. However, Amazon was nevertheless buffeted at its big East Coast complex by service outages on June 14 and June 29, due to power outages.
[ Check out the InformationWeek buyers' guide to IaaS. Infrastructure-As-A-Service Options. ]
The June 29 outage came after the region suffered a series of violent electrical storms, and the outages were contained to one availability zone inside the data center. Amazon didn't say why the battery and generator backup systems of a supposedly highly available cloud didn't keep services running. The outage only disrupted important customers for a few minutes, but they were some of Amazon's most prominent customers, including Salesforce.com's developer cloud Heroku, Netflix, and social networking firms Instagram and Pinterest. On Oct. 22, Amazon suffered an outage of its Elastic Block Storage service for a few hours, making it impossible for some companies to update their websites or retrieve data, even though the sites remained on display.
None of these incidents was a crippling event for Amazon customers or for Amazon's cloud infrastructure business. A few customers likely dropped the service, as WhatsYourPrice did after the online dating service got a flood of complaints during the two-hour June 29 outage. However, the ongoing problem of outages gives cloud skeptics ammunition against moving essential applications there, even as the likes of Amazon, Google, Rackspace and Microsoft win the debate around other performance factors, such as load balancing and database services.
The periodic outages suggest just what a complex beast infrastructure-as-a-service (IaaS) is. Its supposedly redundant, resilient architecture keeps coming up with unanticipated ways of crapping out during predictable events, or from the occasional human error. Cloud providers do have a better track record for uptime than the average enterprise data center, but cloud architects still have their work cut out for them to reduce the odds of high-profile outages that hurt customers and damage the public cloud's reputation.
Setback #2: Virtual Machine Snooping Threat Gets More Real.
It's still only a theory, but a researcher this year published a disturbing example of one virtual machine spying on another on the same physical server. The possibility of such a risk torments virtual server users, who thus far had been coached that the parameters of virtual machines are hard boundaries that couldn't be breached. Cloud computing relies heavily on the multi-tenant, virtual server host, where one physical server is used by multiple companies and customers.
It's important to note that no known breaches using this technique have occurred in the wild. The researchers at the University of North Carolina, University of Wisconsin and the RSA unit of EMC said it was difficult to execute such an attack even in a lab setting.