HPCA '02 Proceedings of the 8th International Symposium on High-Performance Computer Architecture
Ensemble-level Power Management for Dense Blade Servers
Proceedings of the 33rd annual international symposium on Computer Architecture
Power efficiency for variation-tolerant multicore processors
Proceedings of the 2006 international symposium on Low power electronics and design
Proceedings of the 39th Annual IEEE/ACM International Symposium on Microarchitecture
Power provisioning for a warehouse-sized computer
Proceedings of the 34th annual international symposium on Computer architecture
ICAC '07 Proceedings of the Fourth International Conference on Autonomic Computing
No "power" struggles: coordinated multi-level power management for the data center
Proceedings of the 13th international conference on Architectural support for programming languages and operating systems
Multi-optimization power management for chip multiprocessors
Proceedings of the 17th international conference on Parallel architectures and compilation techniques
PowerNap: eliminating server idle power
Proceedings of the 14th international conference on Architectural support for programming languages and operating systems
Statistical profiling-based techniques for effective power provisioning in data centers
Proceedings of the 4th ACM European conference on Computer systems
Temperature-constrained power control for chip multiprocessors with online model estimation
Proceedings of the 36th annual international symposium on Computer architecture
SHIP: Scalable Hierarchical Power Control for Large-Scale Data Centers
PACT '09 Proceedings of the 2009 18th International Conference on Parallel Architectures and Compilation Techniques
SHARP control: controlled shared cache management in chip multiprocessors
Proceedings of the 42nd Annual IEEE/ACM International Symposium on Microarchitecture
Power routing: dynamic power provisioning in the data center
Proceedings of the fifteenth edition of ASPLOS on Architectural support for programming languages and operating systems
Joint optimization of idle and cooling power in data centers while maintaining response time
Proceedings of the fifteenth edition of ASPLOS on Architectural support for programming languages and operating systems
Safe overprovisioning: using power limits to increase aggregate throughput
PACS'04 Proceedings of the 4th international conference on Power-Aware Computer Systems
Leveraging stored energy for handling power emergencies in aggressively provisioned datacenters
ASPLOS XVII Proceedings of the seventeenth international conference on Architectural Support for Programming Languages and Operating Systems
PGCapping: exploiting power gating for power capping and core lifetime balancing in CMPs
Proceedings of the 21st international conference on Parallel architectures and compilation techniques
Aggressive Datacenter Power Provisioning with Batteries
ACM Transactions on Computer Systems (TOCS)
Underprovisioning backup power infrastructure for datacenters
Proceedings of the 19th international conference on Architectural support for programming languages and operating systems
Hi-index | 0.00 |
Data centers attempt to maximize return on investment by achieving high levels of utilization. This means deploying the maximum number of servers possible within existing power supply capabilities. Therefore, a key problem is determining how many servers can be safely accommodated. Recently, a variety of power capping solutions have been proposed to safely allow oversubscription of available power, but they conservatively assume that peak power should never exceed the rated power distribution equipment capacity. Hence, the open question is: how much power oversubscription is indeed safe? In this paper, we focus on data center branch circuits and systematically study the tripping characteristics of their circuit breakers (CBs). Our results on a physical testbed show that instantaneous violations of the rated CB power limit are not necessarily fatal because CBs are designed to sustain a certain amount of power overload. Whether a CB trips or not depends primarily on the transient behaviors of a power overload, such as the magnitude and duration time, as well as ambient temperature. We propose two adaptive power control strategies that utilize the CB tripping characteristics to aggressively optimize the system performance without causing the CB to trip. Our extensive hardware results with SPEC CPU2006, SPECJBB, and LINPACK benchmarks show that the proposed CB-aware power control solutions achieve 38%, 75%, and 68% better average performance, respectively, than a state-of-the-art baseline. A key contribution of our work is to provide a practical upper bound of the server power oversubscription allowed on branch circuits. As a result, our solutions allow a data center to host three times more servers than traditional static power provisioning schemes and 54% more servers than the current power capping practice.