If we quickly review the first two articles, I have discussed that in the Service Level Agreement (SLA) the service and its availability is defined as well as the clauses where the unavailability of the service is not on the shoulders of the service provider (cloud vendor). I will now discuss what happens when the unexpected (unwanted) happens and the service is disrupted and the cloud vendor is responsible. The downtime (or unavailability) of the service is lost business for the client and the client has the right to be compensated, which is clearly laid down in the SLA.
In order to determine the compensation, we first need to calculate downtime. I have given a rough calculation in the first article in the series about unavailability. To put the availability in contract terms, we need to determine a couple of things as:
In what time period is the availability calculated? Monthly or yearly? Usually this period is equal to the billing period.
Is the unavailability aggregated/rolled to consecutive/further periods? What if the availability is 95% in the SLA and the first month the availability is 100% and the next month 92%? Is it 8% unavailability for one month or 4% of average unavailability that is still inside the SLA limit?
What is the unavailability time that is not considered a disruption? Many SLAs do not consider disruptions shorter than 5-minute as unavailability.
Once the availability calculations are made and it is obvious that the availability is not met, the next step is calculating the compensation. The compensation is offered as “service credits” in almost all of the service level agreements. Personally, I haven’t seen any SLA where there is an actual money back guarantee. The reason for that is understandable: offering money back is simply a financial burden that the vendors cannot bear, which can result in a cash/financial crisis that can even result in bankruptcy (furthermore, the refunds will be reflected on the income statements as losses and will lower the value of the company as a whole).
Service credits are calculated as a percentage of downtime:
Pro-rated credit: The customer receives a credit of x % of the downtime. The percentage may be %100, which means that for 1 hour of downtime, the customer receives 1 hour of free service credit. In an extreme case, such as GoGrid, there is the 10,000% of service credit, which is for 1 hour of downtime the customer receives 100 hours of service credit. There is no need to say that the credits are not refundable.
Percentage credit: There is an x% discount on the next invoice if the availability falls below the SLA. For example Amazon discounts the next invoice by 10% if the availability of its EC2 falls below 99.95%.
There is a small note here: the compensations are for the companies who actively monitor the services, complain and ask for compensation. Therefore it is best practice to deploy a monitoring solution that can at least report availability to enable the customer follow what is going on with the services purchased. If not, the customer will have no comparison against the vendor’s figures and will be forced to accept whatever the vendor presents.
From another perspective, the service credits make the customer “captive” to the vendor. When the customer feels that there are “receivables”, he will continue to use the service from the same vendor. From the vendor’s side, since there is no loss and the customer already put in a captive “receiver” state, why would he provide a better service? The customer has to analyze the progress in detail and see if the downtimes are recurring and the vendor is uncaring. If so, the customer may (and probably will) be better off looking for a new vendor.
Service Level Agreement can be a sea hard to navigate unless you clearly define the terms. The compensation part is even harder because there are so many factors: company’s data being held at the vendor’s site, possible emotional dialogues, accrued expenses or other receivables, migration costs etc.. My personal belief is: any financial loss is better than captivity.
In the service level agreement series, I have covered the three major sections: availability, exceptions and limitations and compensation. I hope that what I shared in the articles will allow you to better shape your cloud service, set your expectations and finally arrange your contract accordingly, keeping a critical eye on the details.