Cloud providers may provide compensation to a customer when downtime occurs, but how is it spelled out in the SLA? Thoran Rodrigues explains how it works and what to watch out for.
In my last post, we started a discussion about the different components of service level agreements. The first element of most SLAs is the definitions of service levels and guarantees, which is what we covered the last time around. To give a quick recap: infrastructure-as-a-service providers usually have SLAs covering networking, hardware and VM uptime, but not the software that runs on top of the VMs; platform-as-a-service providers have SLAs that cover availability of the APIs used to perform actions on the platform; and finally software-as-a-service providers have SLAs that cover application and data availability. But what happens when the promises made aren't kept?
The second major elements of any SLA are the compensation clauses, which provide an answer to this exact question. These clauses detail what service providers are willing to offer to their customers in case they don't manage to keep the promises they made previously. Just as what happens with the other elements, these clauses follow a very similar structure regardless of who is your provider.
Service credits are not money!
The most important thing to remember regarding compensation clauses is that no provider will actually give their customers' money back if they don't deliver the promised service levels. Every provider offers service credits, which are discounts either on your current or your future bills. The reason for this is simple: if they offered money back, they would put their companies in serious financial jeopardy, running the risk of going bankrupt and closing the company just to pay contract fines. By offering service credits, they limit their financial risk to their operating costs (or whatever percentage of this the discount covers).
Receiving service credits from the service provider puts consumers in a strange position: on one hand, the service credit acts as a stimulus for further using the service, since it's a discount; on the other hand, the customer needs to consider if he wants to keep using a service provider that wasn't able to deliver on the promises made. You might give a provider the benefit of the doubt the first time some outage happens, but if problems begin to happen recurrently, maybe it's time to start looking for a new provider.
The amount of credits offered can offer some insight into the expectations and on the track record of the service provider being considered. A large amount of service credits being offered in case of downtime may mean two things: the service provider is very certain that it will almost never experience any downtime, and is willing to offer very high compensation because it will almost never have to pay it; or the service provider wants consumers to forget or overlook a poor track record since, if anything goes wrong, they will be more than fairly compensated for their troubles.
There are two central aspects to determining compensation: the way that availability is measured, and the way that compensation is calculated. Most providers calculate availability on a monthly basis, taking into account 5-minute intervals. This is important, since if the service is unavailable for less than 5 minutes straight, it wouldn't be considered as unavailable during that period. Due to this possible loophole, other providers measure availability simply by dividing the time offline by the time online.
One important detail when considering availability is the timeframe in which the availability is calculated. Most providers will consider either a monthly or yearly basis for their availability numbers. This means that availability numbers don't carry over from one month (year) to the next: if the provider delivers 0% availability on one month (year), it has to pay the compensation value for that month (year); on the next month (year); measuring starts over.
Once availability is determined, compensation gets calculated. Providers usually deliver a progressing level of compensation based on how unavailable the service was during a given month, such as 5% discount for every 30 minutes of downtime or some such scheme. Compensation follows availability in terms of how it is calculated and delivered: if availability is measured on a monthly basis, then compensation will be calculated for a given month; if the measurement is per year, the calculation is also per year, and so on.
A final point to remember regarding compensation is that most service providers will only offer compensation for those customers who actively complain and request the compensation. If your cloud server remains unavailable for the whole month, but you forget to ask for your service credits, the service provider won't give you any service credits for that downtime. While there is a period in which requests can be made, once the time has passed, so has any chance of receiving compensation for interruptions in service.
Next time, we'll talk about the exceptions and limitations that service providers impose on the service agreements that define the situations when the rules don't apply.