A Simple Way to Improve the Uptime Service Level

Whether we're calling them cloud computing, hosted software, or SaaS agreements, one of the most important aspects of each relates to the system availability service level. System availability is a critical service level in hosted software agreements for a good reason: the customer is wholly dependent on the vendor to keep the system online and available for use by the customer.

There is nearly always negotiation between the parties on what the system availability service level should be. The vendor may provide an uptime guarantee of 99.5% availability, for example, which equates to roughly 3.6 hours of downtime per month or around 50 minutes per week, measured over a 24 x 7 availability window. But the customer may require a better availability guarantee than one that allows 50 minutes of downtime in a week, especially considering that the outage might occur on a Monday morning at 9:00 am.

Of course, the customer should be prepared to negotiate the availability service level percentage, but there is one simple way of allowing the customer to improve the system availability which doesn't require a change to the availability percentage target. To do so, consider defining availability not on a 24 x 7 x 365 basis, but instead tailor the availability window to the times when the customer will actually need to use the system.

For example, as shown above, a 99.5% uptime guarantee allows for 50 minutes of downtime per week without triggering a service level credit or other remedy. But what if the software is used by the customer to generate management reports? While it may be considered critical to have the software available during normal business hours, the customer may be willing to take more of a risk of unavailability during off hours or over the weekend in order to improve the availability during those business hours when the lights are on.

System Availability During Working Hours

If we narrow the measurement window to working hours -- say, 7:00 am to 7:00 pm (leaving it, for now, at seven days a week), our window is cut in half; effectively doubling the service level. Instead of 50 minutes per week (doing the math, 0.5% downtime x 10,080 minutes in a week), with the smaller measurement window, if the system is unavailable for 25 minutes in a week (0.5% downtime x 5,040 minutes), credits and other remedies will be triggered.

System Availability Excluding Weekends

If we next modify the measurement window to also exclude weekends, the system downtime in any given week can reach just 18 minutes (0.5 downtime x 3,600 minutes) before the contractual remedies are available.

Minutes of Downtime Based on Measurement Window

Therefore, if the system in question is one that is primarily used during normal business hours, we've reduced the weekly allowable downtime from 50 minutes to 18 minutes without needing to negotiate the percentage target. Vendors are often willing to accept this sort of change more readily since it still allows them significant maintenance windows overnight and on weekends that will not impact their service level guarantees.

Of course, each customer will want to consider the right measurement window for its own purposes and the type of software in question. In some cases, 24 x 7 availability is the proper measure, but in many cases narrowing the measurement window will lead to improved service levels without a great deal of drama.

Kenny Hoeschen