An Analysis Framework for Investigating the Trade-Offs between System Performance and Energy Consumption in a Heterogeneous Computing Environment

Friese, Ryan; Khemka, Bhavesh; Maciejewski, Anthony A.; Siegel, Howard Jay; Koenig, Gregory A.; Powers, Sarah; Hilton, Marcia; Rambharos, Jendra; Okonski, Gene; Poole, Stephen W

doi:10.1109/ipdpsw.2013.142

Cited by 20 publications

(13 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To generate this matrix a set of five benchmarks executed over nine machine types were used to construct the initial matrices [21]. Then the method found in [9] was used to construct a larger ETC matrix. Nominally there are 1100 tasks composed of 30 task types.…”

Section: Simulation Setupmentioning

confidence: 99%

See 1 more Smart Citation

Scalable linear programming based resource allocation for makespan minimization in heterogeneous computing systems

Tarplee

Friese

Maciejewski

et al. 2015

Journal of Parallel and Distributed Computing

Self Cite

View full text Add to dashboard Cite

h i g h l i g h t s• We present a novel scheduling algorithm for heterogeneous computing environments.• Uses groupings of similar tasks and machines to reduce the computational complexity.• Computes upper and lower bounds on the optimal makespan. • Schedule approaches a lower bound on the makespan as the number of tasks increases. • Scheduling algorithm run time scales linearly with the number of tasks. a b s t r a c tResource management for large-scale high performance computing systems poses difficult challenges to system administrators. The extreme scale of these modern systems require task scheduling algorithms that are capable of handling at least millions of tasks and thousands of machines. Highly scalable algorithms are necessary to efficiently schedule tasks to maintain the highest level of performance from the system. In this study, we design a novel linear programming based resource allocation algorithm for heterogeneous computing systems to efficiently compute high quality solutions for minimizing makespan. The novel algorithm tightly bounds the optimal makespan from below with an infeasible schedule and from above with a fully feasible schedule. The new algorithms are highly scalable in terms of solution quality and computation time as the problem size increases because they leverage similarity in tasks and machines. This novel algorithm is compared to existing algorithms via simulation on a few example systems.

show abstract

Section: Simulation Setupmentioning

confidence: 99%

“…Nominally there are 1100 tasks composed of 30 task types. The number of tasks per task type varies from 11 to 75 and was generated by the method used in [9]. There are nine machine types with four machines of each type for a total of 36 machines.…”

Section: Simulation Setupmentioning

confidence: 99%

Scalable linear programming based resource allocation for makespan minimization in heterogeneous computing systems

Tarplee

Friese

Maciejewski

et al. 2015

Journal of Parallel and Distributed Computing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Similarly let AP C be a T × M matrix where APC ij is the average power consumption for a task of type i on a machine of type j. These matrices are frequently used in scheduling algorithms [1], [5], [6], [12]. ET C and AP C are generally determined empirically based on prior task execution times.…”

Section: B Lower Boundmentioning

confidence: 99%

“…The number of tasks was increased by applying the method found in [5]. For all the simulations, there are nine machine types and 40 machines of each type for a total of 360 machines.…”

Section: A Introductionmentioning

confidence: 99%

Energy-Aware Profit Maximizing Scheduling Algorithm for Heterogeneous Computing Systems

Tarplee

Maciejewski

Siegel

2014

2014 14th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing

Self Cite

View full text Add to dashboard Cite

Abstract-With the advent of energy-aware scheduling algorithms, it is now possible to find solutions that trade-off performance for decreased energy usage. There are now efficient algorithms to find high quality Pareto fronts that can be used to select the desired balance between makespan and energy consumption. One drawback of this approach is that it still requires a system administrator to select the desired operating point. In this paper, a market-oriented technique for scheduling is presented where the high performance computing system administrator is trying to maximize the return on investment. A model is developed where the users pay a given price to have a bag-of-tasks processed. The cost to the system administrator for processing this bag-of-tasks is strongly related to the energy consumption for executing these tasks. A novel algorithm is designed that efficiently finds the maximum profit resource allocation and tightly bounds the optimal solution. In addition, this algorithm has very desirable runtime and solution quality properties as the number of tasks and machines become large.

show abstract

“…We model general-purpose machine types and special-purpose machine types [16]. The special-purpose machine types execute certain special-purpose task types much faster than the generalpurpose machine types, although they may be incapable of executing the other task types.…”

Section: Problem Description a System Modelmentioning

confidence: 99%

Utility Driven Dynamic Resource Management in an Oversubscribed Energy-Constrained Heterogeneous System

Khemka

Friese

Pasricha

et al. 2014

2014 IEEE International Parallel &Amp; Distributed Processing Symposium Workshops

Self Cite

View full text Add to dashboard Cite

Abstract-In this paper, we address the problem of scheduling dynamically-arriving tasks to machines in an oversubscribed heterogeneous computing environment. Each task has a monotonically decreasing utility function associated with it that represents the utility (or value) based on the task's completion time. Our system model is designed based on the environments of interest to the Extreme Scale Systems Center at Oak Ridge National Laboratory. The goal of our scheduler is to maximize the total utility earned from task completions while satisfying an energy constraint. We design an energy-aware heuristic and compare its performance to heuristics from the literature. We also design an energy filtering technique for this environment that is used in conjunction with the heuristics. The filtering technique adapts to the energy remaining in the system and estimates a fair-share of energy that a task's execution can consume. The filtering technique improves the performance of all the heuristics and distributes the consumption of energy throughout the day. Based on our analysis, we recommend the level of filtering to maximize the performance of scheduling techniques in an oversubscribed environment.

show abstract

An Analysis Framework for Investigating the Trade-Offs between System Performance and Energy Consumption in a Heterogeneous Computing Environment

Cited by 20 publications

References 17 publications

Scalable linear programming based resource allocation for makespan minimization in heterogeneous computing systems

Scalable linear programming based resource allocation for makespan minimization in heterogeneous computing systems

Energy-Aware Profit Maximizing Scheduling Algorithm for Heterogeneous Computing Systems

Utility Driven Dynamic Resource Management in an Oversubscribed Energy-Constrained Heterogeneous System

Contact Info

Product

Resources

About