الفهرس | Only 14 pages are availabe for public view |
Abstract Attempting to reduce the consumption of energy without violating the service level agreement(hereinafter 2SLA3) poses an enormous challenge withinthe environments of Cloud computing. One of the useful approachesadopted to save the consumed energy and avoid violating the SLA is the so-called non-aggressive virtual machine consolidation. Dealing with the virtual machine consolidation problem involves a four-step strategy, which are: the host overloading detection, host underloading detection, virtual machine selection and virtual machine placement. In the currentthesis, the last two steps (virtual machine selection and virtual machine placement) are combined into a single step in order to eschewa possibly ineffective solution resulting fromperforming each step individually. In the steps of host overloading and underloading detection, the host status is categorized into five types: Over-Utilized, Nearly Over-Utilized, Normal Utilized, Under-Utilized and Switched off. Thereupon, a Naive Bayesian Classifier-based algorithm was carried outto detect the future host state and reduce the number of virtual machine migrations to a minimum.Accordingly, the low performance and consumed energy arising from the migrations will be decreased. Also an algorithm,which is built on the Random Key Cuckoo Search, was introduced in the steps of virtual machine selection and virtual machine placement. Additionally, evaluating the algorithm involved the use of 10 traces of real data to validate the proposed algorithms. According to the experimental results,applying the proposed algorithms can curb the energy consumption with average[24.24%, 9.73%, 19.12%] and the SLA violationwith average[88.35%, 89.55%, 72.91%] in the data centers, compared with the LR_MMT 2, UPAVMC and HSNBP algorithms |