The Dirty Secret about Cloud Service Level Agreements
Has your cloud service provider had an outage that affected your users or business? If so, you are not alone. Ask Netflix, Instagram, and Vine—just a few of the high-profile companies that have had major service disruptions due to cloud infrastructure failures in recent months.
If the cloud is supposed to be so reliable, can companies offset some of their losses through the cloud service provider’s service level agreement (SLA)? Based on a recent review of the fine print, I would not count on it because cloud service providers’ SLAs are notorious for being unreadable and unenforceable.
Neither hardware nor software is infallible of course. No matter how well designed an application is for reliability, failure of the underlying cloud infrastructure is always a possibility. For most consumers, cloud outages are an inconvenience, but for cloud based SaaS companies and on-line retailers, losses can easily run into millions of dollars an hour. Amazon itself suffered downtime on August 19, 2013, that lasted approximately forty minutes. That translates to an estimated income loss of $1,104 per second. Ouch!
Many companies that have suffered from cloud service outages have found to their dismay that they are unable to collect even a small amount in credits to offset their losses. Part of the problem is that many cloud providers simply do not bother publishing an SLA at all. For the ones that do, it is typically 99.95 percent of uptime (Amazon). The newly launched Google Compute Engine has at least made the SLA fairly easy to find on its portal—a feature not shared by many of its competitors (Terremark).
99.95 percent uptime might sound good to the naïve buyer but that still means that the service can be unavailable for an average of 21.56 minutes per month or more than eight hours per year before the SLA even comes into question. Once the lawyers add in clauses that cover the length of the outage and other complex details, you are left with unenforceable word salad, as Lydia Leong so nicely puts it.
OK, so one could argue that cloud services are priced cheaply, so it is not unreasonable to expect poor reliability as a consequence—you get what you pay for, right? It is not that the cloud companies cannot deliver reliable services at fair prices. To put it into perspective, the telecom industry has routinely promised and delivered more than 99.999% network reliability, or 5.26 minutes per year for decades. When was the last time you had to wait for a dial tone?
As companies increasingly turn to the cloud for its flexibility, the cloud service providers owe it to their customers to provide services to match reasonable expectations for quality and reliability.