查看文章

This paper presents an energy-aware resource management for data centers with a heterogeneous server infrastructure. The main focus of the paper is on energy-efficient and performance-conserving resource allocation. The suggested resource management uses virtualization and consolidation techniques in order to achieve an energy-efficient operation of the servers in the data center. The resource management itself consists of a monitoring/controlling module, an analyzer module, and an optimizer module. The optimizer module uses power consumption models of servers to compute energy-efficient resource allocations. Service requirement models ensure that the computed resource allocations consider service requirements on CPU, RAM, harddisk, and network. The problem of an energy-efficient and performance-conserving resource allocation is described as a variant of the variable sized multi-dimensional bin packing problem. The result is an energy-efficient and performanceconserving resource management that goes beyond currently applied performance-based consolidation solutions.