CANDEIA, David; http://lattes.cnpq.br/6490865184056059; MAIA, David Candeia Medeiros.
Resumo:
In the last decade we have seen the advent and growth of the Cloud Computing market. This market is based on offering three main services: (i) virtual on-demand computing resources such as storage and processing – this service is called Infrastructure as a Service (IaaS); (ii) platforms that facilitate the development of new applications – this service is called Platform as a Service (PaaS); (iii) software and data hosted on the cloud – this service is called Software as a Service (SaaS). SaaS providers can plan and build their Information Technology (IT) infrastructure making use of computing resources offered by IaaS providers. The business model used by IaaS providers states that computing resources can be reserved in advance by paying a reservation fee. A reserved resource has the advantage that its usage fee is lower than the usage fee of non-reserved resources. Capacity planning is one step in the IT infrastructure management performed by a SaaS provider that helps to estimate the amount of resources required to execute a future workload and thus establish good reservation contracts with IaaS providers. This dissertation presents two capacity planning heuristics: (i) heuristic based on resources utilization rates – UT; (ii) heuristic based on queue networks – RF. These heuristics consider an utility model based on the SaaS provider profit. The evaluation of such heuristics uses a simulation model that considers an e-commerce workload. In all scenarios evaluated UT and RF presented positive utility values and RF presented the best utility gains compared to a strategy that does not reserve resources. Finally, we discovered that according to the quality of the workload prediction that is used by a SaaS provider a better choice can be done among proposed heuristics.