April was a stormy month for a lot of people. Besides the weather, it was also impactful on cloud computing. If you love the cloud, there was one particularly bad week. If you fear the cloud, there was a collective sigh of relief (at least temporarily).
On April 21, Amazon’s EC2 and associated services suffered a major outage in the Eastern Region, affecting several businesses that depend on Amazon Cloud Services to run their mission-critical applications. For some the outage lasted as long as four days. This was true even for some organizations that paid extra to run applications in multiple availability zones. Most of the outage severity and duration was caused not by the initial event (human error causing a network outage), but the programmatic recovery involved in supporting such a sophisticated and massive IT infrastructure. A detailed post-mortem on the Amazon incident is available here.
Amazon is not the only early cloud provider to suffer outages. Just as EC2 was getting back to “normal” operations, VMware’s Cloud Foundry suffered an outage, again exasperated by attempts to recover. A few days later, an analysis was provided.
Does this mean cloud is dead? Likely not. It's just another bump in the road. However, it confirms the cloud is likely not yet ready for most enterprise mission-critical applications. While enterprise data centers are not immune from issues (ask Sony about the Play Station “Cloud”) there is a sense of accountability for internal IT. Today, with public clouds, an SLA is usually limited to service credits. In fact in this incident, cloud proponents applaud Amazon for going well beyond its SLA and providing customers in the affected region a “10-day credit.” Imagine your business being down for days, writing a detailed explanation of why, and concluding with 10 days in services credits.
These events may temporarily curb the enthusiasm for some to jump into public cloud services. Not a week goes by when I don’t discuss with a CIO the role cloud may play into their IT infrastructure. The promise of cloud computing is extremely compelling. Let someone else worry about capital expenditures, infrastructure management, scale, etc., while you focus on core business functions. Yet some questions remain, recent events aside, that still require careful consideration prior to outsourcing IT to a cloud. Those include security, manageability, risk, services levels (i.e., availability) and ultimately, long-term cost once those concerns are addressed.
We maintain our view that cloud technology is very intriguing and is already significantly impacting corporate IT. For most of our customers, the best near-term approach is to quantify requirements for IT as a service, leverage selective public cloud and SaaS services to complement an internal “private cloud” strategy for mission-critical applications. Done right this will position an organization to take advantage of public cloud services for some applications/over subscriptions of internal infrastructure when (and if per Gartner) it makes sense.
Some of this is also reflected in a Gartner Report, "Private Cloud Computing: An Essential Overview."