| With the number of services using virtualization and clouds growing faster and faster, it is common to mutualize thousands of virtual machines within one distributed system. Consequently, the virtualized services, softwares, hardwares and infrastructures share the same physi- cal resources, thus the performance of one depends of the resources usage of others. We propose a solution for vm load balancing (and rebalancing) based on the observation of the resources quota and the dynamic usage that leads to better balancing of resources. As it is not possible to have a single scheduler for the whole cloud and to avoid a single point of fail- ure, our scheduler uses distributed and collaborative scheduling agents. We present scenarios simulating various cloud resources and vm usage experimented on our testbed p2p architecture. |
Export this paper
