Uptime SLAs are not being written to protect the customer. They are designed to shield the service provider and avoid responsibility. It should be different.
I was just reading some stories of system downtime on the LinkedIn conversation “Skip Continuous Availability and High Availability. Have You Been a Victim of No Availability?” It struck me that service providers are failing to write SLAs that respond to the real costs of downtime.
And that’s critical. The adage is that one annoyed customer tells nine. So shouldn’t our “service level agreements” be service oriented? But it’s not happening that way in today’s cloud.
• One major hosting provider “guarantees” 100% uptime — but in the contract they disregard any outage of less than 30 minutes. And if such an outage occurs, they only rebate 5% of the monthly fee for the affected server. That hardly compensates the customer for the loss of business and productivity that may have occurred in the 30-minute outage — and does nothing to address any number of 29-minute outages. What kind of SLA is this, really?
• A SaaS provider’s enterprise-level SLA: “A service outage is covered by the SLA when service is completely unavailable or inaccessible for customer’s use for 10 or more consecutive minutes.” So a nine-minute outage means nothing. Their listed “key industries” include media companies, such as Universal Music Group, who would lose at least 630 files in a 9-minute outage, according to my quick calculations. I am not going to try to estimate the costs of an outage for their next key industry: medical.
• A giant web services provider guarantee only 99.5% uptime (which is just under 44 hours a year!). But they claim that their well-documented 5-day outage that ended on April 21st never violated any SLAs. I know of a remote heart-monitoring company that was unable to read electrocardiogram results for an entire day; I doubt they are happy with their SLAs today.
SLAs should have teeth. They should be written with the customer needs and expectations in mind. Compensation for violated SLAs should bite and draw blood. The compensation for the customer should more than compensate for the pain of their downtime.
There should be no amount of downtime too small to care about. Customers should be clear on how much downtime they can expect, since they feel the pain of business mistakes. The customer should feel compensated — and most important, that they can continue to trust your company.