A growing number of Service Level Agreements (SLA) between customers and Communication Service Providers (CSPs) are being expressed in terms of percentiles. This practice gives CSPs the opportunity to both refine the SLA collaboratively with their customer and also to monetize performance information obtained from their network. It also enables customer to choose from a wider pool of SLAs and adapt them to their specific needs. Some operators even pay penalties if they fail to meet the conditions of the SLA.
What are SLA percentiles and how are they used?
In the network environment, a percentile is a measurable percentage of performance either above or below an agreed level. Percentile measurements are commonly performed over an agreed interval, to ensure they adequately represent the performance of all the packets delivered over that network. A good example is packet delay. Here a SLA might specify that 95% of the packets need to be delivered within 10ms, providing crystal clarity over when a SLA breach occurs.
Calculating percentiles over a measurement interval, however, is resource intensive and, as a result, can prove cost prohibitive. One elegant way to bring the cost down is to obtain percentile measurements using Measurement Bins. A Bin is a kind of counter that increases every time a measured value falls within pre-determined boundaries. In their simplest form, two kinds of Bins can be defined: one for the measurement interval from zero to a threshold value, and a second from this threshold value upwards, infinitely. By just counting the number of values that fall within the first interval or Bin and dividing this value by the total number of values in the measurement interval, one can directly obtain the percentile of measurements that fall within the SLA. Performing this task is, comparably, very efficient, making it easy to verify whether or not the contractual SLA has been honoured.