Machine Learning
Power provisioning for a warehouse-sized computer
Proceedings of the 34th annual international symposium on Computer architecture
Proceedings of the 2008 Asia and South Pacific Design Automation Conference
Power Management of Datacenter Workloads Using Per-Core Power Gating
IEEE Computer Architecture Letters
Blackbox prediction of the impact of DVFS on end-to-end performance of multitier systems
ACM SIGMETRICS Performance Evaluation Review
Power consumption breakdown on a modern laptop
PACS'04 Proceedings of the 4th international conference on Power-Aware Computer Systems
Hi-index | 0.00 |
In the past few years, we have seen the rising popularity of multi-core systems, including the 4, 6, and 8-cores present in i7 processors from Intel, and the 8 and 12 cores present in Magny-Cours processors from AMD. There is a general trend that newer processors have more and more number of cores. A study [6] showed that, in data centers, the CPU load is around 10-50%; thus, when multi-core processors are used in data centers, many of the cores will be unused for a majority of the time. Such a scenario is also true for a casual desktop PC user. As an idle core still consumes energy, from the perspective of saving energy, it is important to ensure that the idle cores are put in the lowest energy state and unnecessary wakeups for these idle cores are avoided. This will ensure the lowest energy consumption for a given set of tasks. The precursor step towards developing an energy efficient CPU scheduler requires an understanding of the relation between the type of tasks and their corresponding power profile. In this paper, we show that the power profile of a task is dependent on the type of processor cycles executed by the task. We further develop a model that can predict the power consumption based on the processor cycles executed by the task. We conclude our paper showing initial results on how such a model can be used to schedule tasks and save energy.