Cloud providers can use service-level agreements (SLAs) to help differentiate themselves with aggressive service guarantees. But how can providers know what metrics to model their SLAs around?
A typical provider's cloud computing SLA will focus on availability -- something most customers demand and expect. Customers take comfort knowing that their chosen network will be reliable and their applications will be constantly available -- thus, providers are usually prepared to make extensive claims in this area.
"When you talk to customers about cloud SLAs, the thing that concerns them the most is availability," said Tom Nolle, president of CIMI Corp. "What this means is that they expect some reasonable guarantees of how much application uptime they're going to get and, almost as important, what the remedy time is going to be if an app becomes completely nonfunctional."
Cloud computing providers generally have availability guarantees figured out. Some cloud providers, such as Rackspace and Terremark, are also willing to guarantee up to 100% network uptime -- boosting confidence in their ability to support the availability of any given customer's website and apps. Performance guarantees, on the other hand, are harder to find. These types of guarantees can be tricky business and involve uncharted territory.
Performance guarantees: A differentiator or more trouble than they're worth?
It is debatable whether it makes sense to include performance in SLAs because performance metrics include overall user experience -- something that is oftentimes subjective.
Moreover, trying to make performance guarantees can get a provider in hot water over something the customer never would have demanded in the first place, according to Nolle, and just bringing up the subject could have the potential to set a provider up for failure.
I would say performance SLAs are our next frontier.
"Every time you say, for example, "I promise you response times of 'x,''' you're creating a metric against which you will be measured that [a customer] probably would never have thought of and probably doesn't care about," he said. "But now they are going to measure you against it."
However, cloud computing providers such as Terremark are starting to brainstorm ways to create meaningful SLAs around performance in hopes of giving their company an edge over competitors. These types of SLAs are just beginning to be explored, however, and Terremark is still in the process of figuring out how to concoct a legitimate performance-based SLA for its cloud services.
"We are putting a lot of thought into how we can offer predictable, reliable and specific performance metrics to a customer that we can then [build an] SLA around," said Christopher Drumgoole, vice president of global operations at Verizon Terremark, a Miami-based cloud provider and division of Verizon Communications Inc.
Terremark employees have been looking into finding a way to enhance customer's experiences from a performance standpoint, opposed to just offering outstanding network availability. This involves making sure a customer is going to have the right network throughput, IOPS and availability as opposed to just availability, according to Drumgoole.
"I would say [performance SLAs] are our next frontier," he said. "It's something we've been starting to think about."
Determining guarantees based on the product and control
Read these cloud computing SLA-related stories
Don't make your cloud computing SLA a marketing gimmick
Prepare for the aftermath of an SLA failure
Offering a blanket performance guarantee may not be the best approach. A provider may want to evaluate the viability of cloud performance guarantees on a product-by-product basis. Cloud computing providers' SLAs can gradually offer guarantees beyond availability without taking on too much risk by only offering these types of guarantees for services in which the provider has a high degree of control over the customer's environment.
"Out-of-the-box, we always [guarantee] uptime and availability. That's kind of table stakes for us," said Drumgoole. "And then, depending on the product you purchase -- whether it is a kind of straight cloud, versus managed hosting, versus highly managed hosting -- we could then [guarantee additional] components beyond [that]."
This was first published in June 2013