| Full text | |
| Author(s): |
Vila, Sergi
;
Lerida, Josep L.
;
Cores, Fernando
;
Guirado, Fernando
;
Verdi, Fabio L.
;
Malawski, M
;
Rzadca, K
Total Authors: 7
|
| Document type: | Journal article |
| Source: | EURO-PAR 2020: PARALLEL PROCESSING; v. 12247, p. 15-pg., 2020-01-01. |
| Abstract | |
Using virtualization, cloud environments satisfy dynamically the computational resource necessities of the user. The dynamic use of the resources determines the demand for working hosts. Through virtual machine (VM) migrations, datacenters perform load balancing to optimize the resource usage and solve saturation. In this work, a policy, named WPSP (Weighted Pearson Selection Policy), is implemented to choose which virtual machines are more suitable to be migrated. For each VM, the policy evaluates both the CPU load and the Network traffic influence on the assigned host. The corresponding Pearson correlation coefficients are calculated for each of the VMs and then weighted in order to provide a relationship between the values and the host behavior. The main goal is to clearly identify and then migrate the VMs that are responsible of the Host saturation but also considering their communications. Using the CloudSim simulator, the policy is compared with the rest of heuristic techniques in the literature, resulting in a reduction of 89% in the number of migrations, and thus reducing the use of bandwidth (5%), network saturation (20%) and over-saturated hosts (51%). Additionally, an improved VM allocation technique to reduce the distance the VMs must travel in order to be migrated is presented, obtaining an average reduction of 87% in the quantity of migrated data. (AU) | |
| FAPESP's process: | 15/19766-9 - Monitoring of physical and virtual resources in cloud computing environments and virtualized infrastructures |
| Grantee: | Fabio Luciano Verdi |
| Support Opportunities: | Regular Research Grants |