The cost for meeting SLA dependability requirements; implications for customers and providers

A Service Level Agreement (SLA) describes the service, the service-level objectives (SLOs), the price the customer should pay and the compensation if the SLOs are not met. There is a trade-off for the provider between the costs for improving the deployed service quality vs. probability of paying compensation. We propose how to estimate the provider's optimal service deployment. We show that the optimal deployed service quality is dependent on the SLOs, deployment cost, compensation and observation interval. A service deployment based on cost optimization results in targeted dependability objective values that are significantly better than stated in the SLOs. The proposed approach provides valuable insight for an aggregator, who buys services from other providers, to negotiate adequate SLOs, price and compensation from the providers to make a valuable offer for its own customers. - Highlights: • SLA with availability, number of failures, down time criteria and observation period. • Analysis of compound service delivery in the SLA context by a semi-Markov model. • SLA non-conformance risk dependent on the inherent/steady state system parameters. • Cost optimal service operation under non-conformance penalties (the safety margin).