Cost accounting for shared IT infrastructures

Abstract

IT infrastructure, such as servers and networking equipment, accounts for a large proportion of the IT costs in modern organizations. Typically, this IT infrastructure is shared among multiple applications and customers. Cost allocation of shared IT infrastructure is difficult and regularly based on biased cost allocation keys, which often causes free-rider problems. Measuring usage is technically difficult and incurs high costs. In this paper we propose a method to derive adequate estimators for the resource consumption of a customer-oriented service. These so-called resource profiles can then provide a basis for cost allocation keys. The estimators are derived from a series of load tests, as they are typically done before an application is launched in larger organizations. Such profiles need to be unbiased and precise even in cases of varying workloads and in rather heterogeneous environments. We describe the results of a set of experiments in an infrastructure provided by the BMW Group, and illustrate how the estimation can be integrated into existing IT service management processes. In our experiments we use Queuing Networks to validate the estimated resource profiles under different workloads.