SLA (Service Level Agreement) and its management have become more and more important for both service providers and customers, and the measurement of service level is fundamental for other management processes. The paper discusses the differences between service measurement and traditional network measurement from the perspective of sampling, and then deduces a theoretical lower-limit on sampling frequency. To reduce the number of samplings and keep high accuracy of service level data at the same time, the paper presents an
adaptive algorithm based on estimation method. The random function is used to determine the actual sampling time in each sampling period, which is divided into a number of small even intervals. Then the auto-regressive model is chosen to estimate the current sampling value
based on the previous sampling results. If the estimation result is within the range of probability accepted, sampling at this interval is given up. Thus, this algorithm reduces sampling's influence on the measured service by decreasing the sampling numbers. The experiments verify the effectiveness of this algorithm.
Chengdong Zhao, Chengdong Zhao,
"An adaptive method for service-level monitoring using autoregressive technique", Proc. SPIE 5245, Internet Quality of Service, (8 August 2003); doi: 10.1117/12.505037; https://doi.org/10.1117/12.505037