Workload consolidation on virtualized datacenters is one common technique used to manage the power usage and energy efficiency of cloud platforms. However, current solutions limit the available insight into exactly how energy is consumed by the various cloud tenants.
In order to address this gap, this research develops PowerMeter ? a set of mechanisms and tools that make it possible to track energy usage in virtualized cloud infrastructures, on granularity of individual VMs, or sets of VMs corresponding to a specific cloud tenant, e.g., a customer or an application. PowerMeter is developed for a 3000-core fully instrumented datacenter prototype and a 1000- core system based on newest-generation technology, and is evaluated using representative workload mixes based on popular cloud applications and publicly available cloud traces. PowerMeter will provide currently missing capabilities for dynamically assessing in- datacenter/in-cloud energy usage, necessary to enable energy usage transparency and accountability to cloud customers, to understand the opportunities for improved energy efficiency in virtualized cloud systems, and for further development of management methods for energy- efficient datacenters.
