Derivation of response time service level objectives for business services

Abstract

Design of Service Level Agreements (SLAs) emerges as an increasingly important discipline in business-oriented IT management. In this work, we study utility maximization of contractual obligations of a business service provider for a typical SLA. We demonstrate that using service usage and performance data as well as IT performance data, routinely collected by enterprises, efficient automated derivation of optimal response time Service Level Objectives (SLOs) of an SLA is possible. This paper addresses a specific facet of the SLA design problem not sufficiently addressed in previous studies. One common approach is to calculate SLOs attainable for the given IT infrastructure by means of simple percentile analysis. However, this methodology is business agnostic and may result in sub-optimal SLOs. Another widespread approach addresses IT infrastructure (re)design, where the goal is to enable the IT to meet specified target SLOs. In contrast to these approaches, our work proposes finding optimal response time SLOs attainable for the given IT infrastructure in a business aware manner. We define the response time SLO optimization problem, propose an algorithm that efficiently solves it using linear optimization and percentile analysis, and evaluate our solution both analytically and experimentally. © 2007 IEEE.