Fuzzy modeling to predict performance of collocated virtual machines in private clouds

This research describes a general model for estimating performance of collocated virtual machines (VMs) on physical machines (PMs). This prediction is measured based on resource contention between several VMs. Using a Fuzzy model, we predict how VMs and PMs would be affected in different work scenarios. Base on the Model, the level of contention on the resource usage is quantified and can be used to predict VMs performance behavior toward more efficient resource allocation decisions. Experimental evaluations on resource benchmarks using several client/server side applications and relational database management systems (RDBMS) demonstrate that this modeling approach can detect and quantify the relationship between VMs and PMs for CPU, disk I/O, memory and network consumption profiles.