SLA's in the Cloud

26th October 2011

High profile power outages, such as those suffered by Amazon, Microsoft and Google in recent months are continuing to mar the Cloudcomputing industry. But many Cloudservice providers are attempting to lessen the frustrations of the thousands who are faced with significant downtime by pledging to pay out compensation to all affected customers. Indeed, many providers are starting to guarantee a "four nines" (99.99%), or even a "five nines" (99.999%), SLA with regards to uptime. Such SLAs state that if the end-user receives less than 99.99 or 99.999% uptime, they will receive an agreed amount of compensation.

But before end-users hurriedly sign on the dotted line, convinced that they are covered for all instances when their applications are unavailable, due to maintenance, human error or natural disasters, we must question just how likely it is that a customer will ever receive a payout? It is after all, no great secret that whilst SLAs look good for the customer on paper, they are always designed to work in favour of the service provider. How then, does the end-user ensure that their Cloudcomputing contract has been drawn up with them in mind?

Uptime cannot be guaranteed

It has been stated time and time again, but regardless of what anyone says, uptime cannot be guaranteed by any Cloudservice provider 100% of the time. And any Cloudvendor that argues otherwise cannot be trusted! For those that rightfully do not guarantee 100% uptime, they usually put in place SLAs that state that if their agreed continuous uptime of say, 99.99% is not met, the end-user will be financially rewarded. How this clause is worded in the contract however, is something that needs to be closely watched. Take for instance, the two scenarios below:

Scenario One

Situation – Customer A experiences continuous service downtime of 20 minutes, leaving him unable to access his email account, applications or data. Productivity is reduced; meaning business critical work is lost.

Result – Customer A is compensated as continuous downtime of 20 minutes in a single day breaks the agreed 99.99% continuous uptime SLA in his contract.

Scenario Two

Situation – Customer B experiences downtime of five minutes for four days in a row. On four separate occasions, he is unable to access his email account, applications or data, disrupting the productivity and damaging the business on four separate occasions.

Result – Customer B receives no compensation as downtime of five minutes per day does not breach the agreed 99.99% continuous uptime SLA in his contract.

Sounds unfair doesn't it? Both customers experience downtime of twenty minutes, but written in the small print of the contract, is "continuous downtime" – twenty minutes of continuous downtime is actually very different, in the eyes of the contract at least, to four days of five minutes of downtime.

Has this maintenance been scheduled?

Another obstacle that many Cloudend-users are coming across is the issue of "scheduled maintenance." In most cases, scheduled maintenance will be excluded from the SLAs, so that the service provider is able to arrange maintenance windows for times when access to business-critical data and applications is not needed – typically, weekends and evenings.

There are two points to be made here: firstly, if the client's business is a global, 24/7 operation, those "weekend" maintenance windows will be heavily impacting end-users who are trying to access applications from the other side of the world.

Secondly, end-users must be wary of the SLA clause "Excluding non-scheduled emergency maintenance windows as required." Granted, excluding scheduled maintenance from the SLAs is one thing, but excluding "non-scheduled emergency maintenance" from the contract is a whole different ball game. Putting this into the contract, the service provider has nothing to lose. Suddenly, every power outage will become a "non-scheduled emergency maintenance window," and the customer will receive no compensation whatsoever.

Choosing a Cloudservice provider is never easy, but something that can be used as a differentiator between competitors is how willing they are to negotiate on the SLAs to be placed in the contract. Indeed, smaller providers tend to be a lot more flexible than the Cloudgiants, offering terms that the customer has requested rather than simply forcing the customer to abide to their rulings. The bigger companies on the other hand, can afford not to negotiate, and end-users will therefore find it more difficult to amend the terms of the contract if opting to go with these providers.

It is also of critical importance that before any contract is signed, the end-user fully appreciates that what the agreed SLAs look like they mean can be very different to what they actually mean. And if the end-user does not feel entirely comfortable with exactly what the SLAs mean, it is important that he has a lawyer present with him at the signing of the contract who is au fait with CloudComputing jargon. This way, the lawyer is able to look over the contract, and let the client know whether he will receive any payout if faced with downtime, continuous or not, or if faced with maintenance, scheduled or not!