Adaptive Resource and Job Management for Limited Power Consumption

Georgiou, Yiannis; Glesser, David; Trystram, Denis

doi:10.1109/ipdpsw.2015.118

Cited by 14 publications

(9 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Powercap mechanisms have two major drawbacks: they may require high knowledge about the running applications (to tune DVFS or a similar technique), and it also delays some jobs. In our previous study [6], we found that only controlling the power increases the turnaround time of big jobs (as it is harder for them to "fit" in the powercap). This is why we focus here on energy budgeting, as we want to keep the benefit of controlling the cost of the cluster, while not discriminating against any type of job.…”

Section: A Controlling Power and Energy Consumptionmentioning

confidence: 91%

“…Readers may wonder why we do not take DVFS into account, since DVFS can be used to control the power consumption of a job. Previous studies [6], [7] have shown that controlling the energy consumption of jobs with DVFS is not trivial. Depending on the type of application, a given DVFS value may either increase or decrease the total energy consumption of the application.…”

Section: B Energy and Power-saving Techniquesmentioning

confidence: 99%

See 1 more Smart Citation

Towards Energy Budget Control in HPC

Dutot

Georgiou²,

Glesser

et al. 2017

2017 17th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)

Self Cite

View full text Add to dashboard Cite

Energy consumption has become one of the most critical issues in the evolution of High Performance Computing systems (HPC). Controlling the energy consumption of HPC platforms is not only a way to control the cost but also a step forward on the road towards exaflops. Powercapping is a widely studied technique that guarantees that the platform will not exceed a certain power threshold instantaneously but it gives no flexibility to adapt job scheduling to a longer term energy budget control. We propose a job scheduling mechanism that extends the backfilling algorithm to become energy-aware. Simultaneously, we adapt resource management with a node shutdown technique to minimize energy consumption whenever needed. This combination enables an efficient energy consumption budget control on a cluster during a period of time. The technique is experimented, validated and compared with various alternatives through extensive simulations. Experimentation results show high system utilization and limited bounded slowdown along with interesting outcomes in energy efficiency while respecting an energy budget during a particular time period.

show abstract

Section: A Controlling Power and Energy Consumptionmentioning

confidence: 91%

Section: B Energy and Power-saving Techniquesmentioning

confidence: 99%

Towards Energy Budget Control in HPC

Dutot

Georgiou²,

Glesser

et al. 2017

2017 17th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGRID)

Self Cite

View full text Add to dashboard Cite

show abstract

“…, T n } of n tasks T i . According to [SS09,GGT15], the power consumption varies from one task to another depending whether the task intensively computes or not. Each task T i is defined by its processing time p i and its power consumption ϕ i .…”

Section: Modelmentioning

confidence: 99%

Green Power Constrained Scheduling for Sequential Independent Tasks on Identical Parallel Machines

Philippe

Nicod

Rehn‐Sonigo

2019

2019 IEEE Intl Conf on Parallel &Amp; Distributed Processing With Applications, Big Data &Amp; Cloud Computing, Sustainable Com

View full text Add to dashboard Cite

Energy consumption is a major aspect to consider when designing large scale high performance computing HPC systems. Integrating renewable energy sources to the power supply of an HPC system is an efficient solution to lower its carbon footprint, but it is one faced by challenges since the power production of most renewable sources is variant while the power consumption of the system varies with the workload. Advanced workload management techniques in combination with powering down idle machines allow to increase the efficiency of this solution. We tackle here the problem of scheduling independent tasks on a multi-machine platform that is exclusively run with green energy. We propose different power constrained scheduling algorithms, and evaluate them through an experimental study on an HPC model that considers the possibility to switch machines on or off.

show abstract

“…In this paper we consider the power consumed by a task, its instant consumption, rather than its energy need. In [7] Glesser et al show that running a task results in an energy consumption that depends on the task type, either intensive or not. Based on that we define ϕ i as the power consumption ϕ i of task T i .…”

Section: Modelmentioning

confidence: 99%

Assessing the Use of Genetic Algorithms to Schedule Independent Tasks Under Power Constraints

Kassab

Nicod

Philippe

et al. 2018

2018 International Conference on High Performance Computing &Amp; Simulation (HPCS)

View full text Add to dashboard Cite

Green data and computing centers, centers using renewable energy sources, can be a valid solution to the over growing energy consumption of data or computing centers and their corresponding carbon foot print. Powering these centers with energy solely provided by renewable energy sources is however a challenge because renewable sources (like solar panels and wind turbines) cannot guarantee a continuous feeding due to their intermittent energy production. The high computation demand of HPC applications requires high power levels to be provided from the power supply. On the other hand, one advantage is that unlike online applications, HPC applications can tolerate delaying the execution of some tasks. Since the users however want their results as early as possible, minimum makespan is usually the main objective when scheduling this kind of jobs. The optimization problem of scheduling a set of tasks under power constraints is however proven to be NP-Complete. Designing and assessing heuristics is hence the only way to propose efficient solutions. In this paper, we present genetic algorithms for scheduling sets of independent tasks in parallel, with the objective of minimizing the makespan under power availability constraints. Extensive simulations show that genetic algorithms can compute good schedules for this problem.

show abstract

Adaptive Resource and Job Management for Limited Power Consumption

Cited by 14 publications

References 22 publications

Towards Energy Budget Control in HPC

Towards Energy Budget Control in HPC

Green Power Constrained Scheduling for Sequential Independent Tasks on Identical Parallel Machines

Assessing the Use of Genetic Algorithms to Schedule Independent Tasks Under Power Constraints

Contact Info

Product

Resources

About