This chapter demonstrates that the same type of analytic models used to investigate and predict performance issues are also useful to analyze availability considerations. In fact, availability and performance are important complementary aspects of the QoS of a computer system. The joint analysis of these two metrics leads to what is known as performability analysis. This chapter illustrates the usefulness of Markov models and QNs as modeling approaches to investigate design issues of data centers. Capacity planning is conducted in terms of the number of machines and in terms of the quality and size of the maintenance staff in order to achieve certain availability goals. |