14-02-2017, 02:13 PM
With the increase in energy consumption associated with IT infrastructures, energy management is becoming a priority in the design and operation of complex service-based systems. At the same time, service providers must comply with Service Level Agreement (SLA) contracts that determine revenues and penalties based on the level of performance achieved. This paper focuses on the problem of allocating resources in multi-topic virtualised systems with the aim of maximising revenue from SLAs while minimising energy costs. The main novelty of our approach is to approach -in a unified framework- the management of resources of the service centres by exploiting as mechanisms of action the assignment of virtual machines to servers, load balancing, capacity allocation, state adjustment Server power and dynamic voltage / frequency escalation. Resource management is modelled as a nonlinear programming problem for NP-hard mixed integers and is solved by a local search procedure. To validate its effectiveness, the proposed model is compared with the most advanced techniques of the state of the art. The evaluation is based on simulation and actual experiments performed in a prototype environment. Synthetic workloads are considered as well as realistic and a number of different scenarios of interest. The results show that we are able to make significant revenue gains for the provider compared to alternative methods (up to 45 percent). In addition, the solutions are robust at the time of service and variations in workload.
