Optimal Temporal Partitioning of a Multicore Server Processor for Virtual Machine Allocation

Abstract

The problem of optimal temporal partitioning of a multicore server processor for virtual machine allocation in cloud computing is addressed as multivariable optimization problems and solved algorithmically and numerically. Analytical models for virtual machines are developed, i.e., partially available multi-server systems. The problem of optimal temporal partitioning of a multicore server processor for virtual machine allocation is formulated, where the overall performance (i.e., the average task response time) of a group of virtual machines is optimized. An algorithm is developed to solve the problem numerically. The problem of optimal temporal partitioning of a multicore server processor with power consumption constraint is also formulated and solved, where the overall performance of a group of virtual machines is optimized and the total power consumption of the virtual machines does not exceed certain available power. A virtual machine is treated as a queuing system with multiple servers, i.e., an M/M/m queuing system. The system performance measures are the average task response time and the average power consumption. Two core speed and power consumption models are considered, namely, the idle-speed model and the constant-speed model. Numerical examples are presented to demonstrate our methods.