Ben Trowbridge responds
Technology is great when it works. What does a buyer of cloud services do when something goes wrong, creating a service disruption?
That question became a reality for users of Amazon Web Services on April 20. Its Northern Virginia data center suffered a service outage for its hosting service at 10:41 p.m. EDT. This failure knocked thousands of websites off the Internet for 30 hours. Some of those websites are household names in the social media world including Foursquare, Hootsuite, Quora and Reddit.
What are the remedies for these buyers? Ben Trowbridge, CEO of Alsbridge, has negotiated cloud computing agreements with Amazon for Alsbridge clients. Trowbridge says Amazon Web Services’s standard service level agreement (SLA) on uptime is 99.95 percent for the year for computing (Amazon EC2). That means a client can expect to be down for six hours, 43 minutes and 12 seconds every 365 days.
“While this SLA target is possible to achieve, it takes great effort by the provider to attain this year after year,” he observes.
Since 30 hours is clearly over the SLA, the typical agreement calls for Amazon to pay the buyer a service credit equal to 10 percent of its monthly bill. Calculating the cost, if the client was paying $20,000 a month, then Amazon would credit the account $2,000.
Will that cover the lost revenue? While service providers generally do not offer an SLA credit covering direct damages, Trowbridge suggests some clients “should ask for increased penalties from their cloud providers. It’s a negotiation point. While many of the smaller players will not have the leverage to get Amazon to move on the calculation method, companies with sizable amounts of spend with Amazon should seek to negotiate with Amazon rather than accept the boilerplate SLA credit.”
Another issue: Sean Halverson, senior consultant at Alsbridge, points out at least three popular platform-as-a-service (PaaS) providers (DotCloud, Engine Yard and Heroku) went down, too, as a result of the outage at Amazon. Their customers were also impacted by the shutdown. The existential question here is who pays for the SLA misses?
Do the math
Can cloud users afford to leave the cloud in return for better reliability? Trowbridge points out cloud service providers can provide IT services for as much as 80 percent less than a traditional outsourcing service provider. He posits many of Amazon’s Web hosting clients “probably couldn’t exist without Amazon. Amazon is running a sophisticated IT environment that would be too costly if these buyers used a traditional provider or tried to build it themselves,” he says.
Salesforce.com has what it calls “service problems,” which is when the software runs interminably slowly. “But users are willing to stay with it despite the frequency of the problem because the application is so valuable,” Trowbridge says. He predicts the same will happen with cloud buyers, especially because of the economics.
Ask the right questions when buying cloud services
Trowbridge says many of the questions a buyer of cloud services has to ask are similar to the ones they should pose when sourcing a traditional ITO agreement. They include:
- What is the back-up plan?
- What is the plan for corrupt data?
- What is the recovery speed? This is especially important for large files.
- What are the SLA penalties?
One solution is not to put all eggs in one basket. Instead of putting all their data or hosting in one data center, buyers can use multiple data centers and mirror sites to increase the up-time. Another option is leaving the current provider and finding one with better metrics and performance. Switching, of course, requires payment in time and money.
All IT outsourcing involves risk, points out Trowbridge. Being an informed buyer has never been more important.
Finding the right cloud provider
Trowbridge has just written a book entitled “Cloud Sourcing the Corporation,” which will be available for purchase on May 18, 2011. It provides a holistic perspective on the evolving cloud services market and acts as a guidebook for IT executives on their journey to the cloud. Included in the book is the Cloud Sourcing 100 Index, an exhaustive review of the emerging cloud services providers complete with classifications and ratings. The Index is also online at www.cloudsourcing100.com.
There is always a silver lining in every cloud (pun intended). After this event, Amazon will probably work harder to prove its business case and come up with ways to improve its reliability. Cloud buyers will probably be willing to pay more for mirrored sites within their outsourcing agreement.
The bottom line: Achieving five nines is nearly impossible and, more often than not, cost prohibitive. So plan for service disruptions.
Have you experienced a service disruption? What happened? How did the service provider fix it? Please post a comment.