Stochastic Dynamic Thermal Management: A Markovian Decision-based Approach

Jung, Hwisung; Pedram, Massoud

doi:10.1109/iccd.2006.4380855

Cited by 14 publications

(12 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several new controller designs have been proposed in the recent years for multicore DEM [Bartolini et al 2011;Jung and Pedram 2006;Wang et al 2009;Zanini et al 2009]. These controllers can be broadly classified based on their approaches, viz.…”

Section: Overview Of Proposed Steam Controllermentioning

confidence: 99%

“…statistical and control-theoretic. Jung and Pedram [2006] propose a statistical technique based on partially observable Markov chains to predict an optimal processor frequency setting. While their approach avoids the need for learning of the processor power or thermal models, their algorithm requires complex computations, and the number of DVFS state searches grow exponentially with the number of processor cores.…”

Section: Overview Of Proposed Steam Controllermentioning

confidence: 99%

See 1 more Smart Citation

Steam

Hanumaiah

Desai

Gaudette

et al. 2014

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

Recent empirical studies have shown that multicore scaling is fast becoming power limited, and consequently, an increasing fraction of a multicore processor has to be under clocked or powered off. Therefore, in addition to fundamental innovations in architecture, compilers and parallelization of application programs, there is a need to develop practical and effective dynamic energy management (DEM) techniques for multicore processors.Existing DEM techniques mainly target reducing processor power consumption and temperature, and only few of them have addressed improving energy efficiency for multicore systems. With energy efficiency taking a center stage in all aspects of computing, the focus of the DEM needs to be on finding practical methods to maximize processor efficiency. Towards this, this article presents STEAM -an optimal closed-loop DEM controller designed for multicore processors. The objective is to maximize energy efficiency by dynamic voltage and frequency scaling (DVFS). Energy efficiency is defined as the ratio of performance to power consumption or performance-per-watt (PPW). This is the same as the number of instructions executed per Joule. The PPW metric is actually replaced by P α PW (performance α -per-Watt), which allows for controlling the importance of performance versus power consumption by varying α.The proposed controller was implemented on a Linux system and tested with the Intel Sandy Bridge processor. There are three power management schemes called governors, available with Intel platforms. They are referred to as (1) Powersave (lowest power consumption), (2) Performance (achieves highest performance), and (3) Ondemand. Our simple and lightweight controller when executing SPEC CPU2006, PARSEC, and MiBench benchmarks have achieved an average of 18% improvement in energy efficiency (MIPS/Watt) over these ACPI policies. Moreover, STEAM also demonstrated an excellent prediction of core temperatures and power consumption, and the ability to control the core temperatures within 3 • C of the specified maximum. Finally, the overhead of the STEAM implementation (in terms of CPU resources) is less than 0.25%. The entire implementation is self-contained and can be installed on any processor with very little prior knowledge of the processor.

show abstract

Section: Overview Of Proposed Steam Controllermentioning

confidence: 99%

Section: Overview Of Proposed Steam Controllermentioning

confidence: 99%

Steam

Hanumaiah

Desai

Gaudette

et al. 2014

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

show abstract

“…The temperature of core is modeled with linear regression and found that it is linear function of frequency and workload in [10] and after modeling the regression variables are reduced by principal component analysis. The stochastic nature of temperature variation by modeling and accessing the uncertainty in core temperature is constructed in [11]. Using Markov decision process the thermal state was modeled to determine the next state and DVFS was used for thermal control.…”

Section: A Single Corementioning

confidence: 99%

Extraction of thermal workload signatures in multicore processors using least angle regression

Karn¹,

Elfadel²

2015

2015 International Conference on Communications, Signal Processing, and Their Applications (ICCSPA'15)

View full text Add to dashboard Cite

Performance counters (PCs) embedded in microprocessor are frequently used to characterize workload and predict thermal behavior for multicore processors. These PCs are required to be highly accurate, very compact, and tunable to workload changes in real time. Traditionally these PCs are selected using correlation map or some sort of statistical trial-error techniques. These techniques have the disadvantage of requiring the large PC set regardless of the workload type which is computationally burden when scaling number of cores in processor. In this paper, we use the more recent algorithm of least-angle regression to choose specific set of PCs for definite workload characteristic and validate its accuracy by thermal modeling. It include only those PCs most correlated with thermal behavior of workload. Such PCs are considered as signatures to predict workload characteristic and to apply specific thermal management action. The PC sets are trained and tested on model using workloads from the PARSEC and SPEC CPU 2006 benchmarks.

show abstract

“…0, for , (17) where , denotes the probability that the system will next come to state if it is currently in state and action is chosen. Constraints (15) -(17) capture the properties of a CTMDP.…”

Section: The Local Agentsmentioning

confidence: 99%

“…Each core in the server cluster is modeled using a continuous-time Markov decision process (CTMDP) [17] [18], in which actions are execution frequencies (and supply voltages) for processing the service requests. We know that a higher execution frequency will result in a shorter response time, but a significant increase in power consumption [12][13] [14].…”

Section: Introductionmentioning

confidence: 99%

Resource allocation and consolidation in a multi-core server cluster using a Markov decision process model

Wang

Chen

Goudarzi

et al. 2013

International Symposium on Quality Electronic Design (ISQED)

View full text Add to dashboard Cite

Distributed computing systems have attracted a lot of attention due to increasing demand for high performance computing and storage. Resource allocation is one of the most important challenges in the distributed systems especially when the clients have some Service Level Agreements (SLAs) and the total profit depends on how the system can meet these SLAs. In this paper, an SLA-based resource allocation problem in a server cluster is considered. The objective is to maximize the total profit, which is the total price gained from serving the clients subtracted by the operation cost of the server cluster. The total price depends on the average request response time for each client as defined in their utility functions, while the operating cost is related to the total energy consumption. A joint optimization framework is proposed, comprised of request dispatching, dynamic voltage and frequency scaling (DVFS) for individual cores, as well as server-level and core-level consolidations. Each core in the cluster is modeled using a continuous-time Markov decision process (CTMDP). A near-optimal hierarchical solution is proposed, consisting of a central manager and distributed local agents. Each local agent employs linear programming-based CTMDP solving method to solve the DVFS problem for the corresponding core. The central manager solves the request dispatching problem and finds the optimal number of turned on cores and servers for request processing, thereby achieving a desirable tradeoff between service request response time and power consumption. Experimental results demonstrate that the proposed near-optimal resource allocation and consolidation algorithm consistently outperforms baseline algorithms.

show abstract

Stochastic Dynamic Thermal Management: A Markovian Decision-based Approach

Cited by 14 publications

References 10 publications

Steam

Steam

Extraction of thermal workload signatures in multicore processors using least angle regression

Resource allocation and consolidation in a multi-core server cluster using a Markov decision process model

Contact Info

Product

Resources

About