An overview of energy efficiency techniques in cluster computing systems

Valentini, Giorgio; Lassonde, Walter; Khan, Samee U.; Min‐Allah, Nasro; Madani, Sajjad A.; Li, Juan; Zhang, Limin; Wang, Lizhe; Ghani, Nasir; Kołodziej, Joanna; Zomaya, Albert Y.; Xu, Cheng‐Zhong; Balaji, Pavan; Vishnu, Abhinav; Pinel, Fredric; Pecero, Johnatan E.; Kliazovich, Dzmitry; Bouvry, Pascal

doi:10.1007/s10586-011-0171-x

Cited by 179 publications

(94 citation statements)

References 26 publications

(45 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Valentini et al [18] present the state of the art of the above mentioned technologies and examine them regarding energy savings and performance slumps. Energy saving potential varies significantly in relation to application, workload, cluster system, and scheduling strategy.…”

Section: Related Workmentioning

confidence: 99%

Overhead-aware Load Distribution and System Shutdown for Energy-Efficient Computing

Lenhardt

Schiffmann

2015

IJNC

View full text Add to dashboard Cite

The energy consumption of server farms is steadily increasing. This is mainly due to an increasing number of servers which are often underutilized most of the time. In this paper we discuss various strategies to improve the energy efficiency of a datacenter measured by the average number of operations executed per Joule. We assume a collection of heterogeneous server nodes that are characterized by their SPECpower-benchmarks. If a time-variable divisible (work)load should to be executed on such a datacenter the energy efficiency can be improved by a smart decomposition of this load into appropriate chunks. In the paper we discuss a sophisticated load distribution strategy and extend it by an adaptive power management for dynamically switching underutilized servers to performance states with lower energy consumption. Of course, also transitions to higher performance/energy states are possible if required by the current load. We introduce a new time slice model that allows a reduction of the switching overhead by means of a few merge and adjust cycles. The resulting ALD+ strategy was evaluated in a webserver environment with real Wikipedia traces. It achieved significant reductions of the energy consumption by the combination of load distribution and server switching by means of the time slice model. Moreover, ALD+ can be easily integrated into any parallel webserver setup.

show abstract

Section: Related Workmentioning

confidence: 99%

Overhead-aware Load Distribution and System Shutdown for Energy-Efficient Computing

Lenhardt

Schiffmann

2015

IJNC

View full text Add to dashboard Cite

show abstract

“…According to [13], one important research topic for getting energy efficiency by applying DPM techniques is to schedule powering on and off computer's components (the whole server in most cases) to adapt to the workload. The survey [2] also explains some DPM works from other authors that would get idle resources, but the reviewed works usually assume that those idle resources are powered on or off automatically and do not consider any scheduling strategy. Most of them are basically job schedulers that would substitute the existing schedulers or cluster management middlewares and would obviously modify the way that users interact with them.…”

Section: Related Workmentioning

confidence: 99%

“…Based on this information, the scheduler determines if new nodes must be switched on, or if there are nodes that can be switched off, and acts consequently (2). When a job is submitted to the resource manager (3), a request for nodes is made to CLUES by means of the resource manager connector (4).…”

Section: System Descriptionmentioning

confidence: 99%

“…This problem is especially important in clusters that are underutilized, either because they form part of large scale distributed systems (grids or clouds) [1], where load can have important variations, or because the clusters have been in production for several years and their usage has decreased in favour of other more modern systems. However, in the last years there have been advances in the energetic efficiency of HPC clusters, which have come as a result of two different approaches: Static Power Management (SPM) techniques that use low-power energy-efficient hardware to reduce energy usage, and Dynamic Power Management (DPM) techniques that are based on the knowledge of resource utilization and application workloads to reduce energy usage [2].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An energy management system for cluster infrastructures

Alfonso

Caballer

Alvarruiz

et al. 2013

Computers & Electrical Engineering

View full text Add to dashboard Cite

This paper presents a general energy management system for High Performance Computing (HPC) clusters and cloud infrastructures that powers off cluster nodes when they are not being used, and conversely powers them on when they are needed. This system can be integrated with different HPC cluster middleware, such as Batch-Queuing Systems or Cloud Management Systems, and can also use different mechanisms for powering on and off the computing nodes. The presented system makes it possible to implement different energy-saving policies depending on the priorities and particularities of the cluster. It also provides a hook system to extend the functionality, and a sensor system in order to take into account environmental information. The paper describes the successful integration of the system proposed with some popular Batch-Queuing Systems, and also with some Cloud Management middlewares, presenting two real use-cases that show significant energy/costs savings of 27% and 17%.

show abstract

“…in parameter sweep applications [2,3], the scheduling problems in Computational Grids (CGs) and in Data Grids (DGs) is dealing with in a separated way. Much of the current efforts are focused on scheduling workloads in a data center or schedule movement of data and data placement [42] for efficient resource/storage utilization or energy-effective scheduling in largescale data centers [41], [8], [33], [18], [48], [51], [57], [7], [10], [16], [17]. A recent example is that of GridBatch [44] for large scale data-intensive problems on cloud infrastructures.…”

Section: Introductionmentioning

confidence: 99%

Data Scheduling in Data Grids and Data Centers: A Short Taxonomy of Problems and Intelligent Resolution Techniques

Kołodziej

Khan

2013

Transactions on Computational Collective Intelligence X

Self Cite

View full text Add to dashboard Cite

Abstract. Data-aware scheduling in today's large-scale heterogeneous environments has become a major research issue. Data Grids (DGs) and Data Centers arise quite naturally to support needs of scientific communities to share, access, process, and manage large data collections geographically distributed. Data scheduling, although similar in nature with grid scheduling, is given rise to the definition of a new family of optimization problems. New requirements such as data transmission, decoupling of data from processing, data replication, data access and security are to be added to the scheduling problem are the basis for the definition of a whole taxonomy of data scheduling problems. In this paper we briefly survey the state-of-the-art in the domain. We exemplify the model and methodology for the case of data-aware independent job scheduling in computational grid and present several heuristic resolution methods for the problem.

show abstract

An overview of energy efficiency techniques in cluster computing systems

Cited by 179 publications

References 26 publications

Overhead-aware Load Distribution and System Shutdown for Energy-Efficient Computing

Overhead-aware Load Distribution and System Shutdown for Energy-Efficient Computing

An energy management system for cluster infrastructures

Data Scheduling in Data Grids and Data Centers: A Short Taxonomy of Problems and Intelligent Resolution Techniques

Contact Info

Product

Resources

About