Dealing with Heterogeneity in Load Balancing Algorithms

Beltrán, Marta; Guzmán, Antonio; Bosque, José Luis

doi:10.1109/ispdc.2006.17

Cited by 14 publications

(10 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The idea is to assign to each process a data partition that is proportional to the relative computing power (RCP ) of the processor on which it is running. The relative computing power of processor i (RCP i ) [36,37] is computed in Equation 1 as the computing power of the processor (in F LOP S) divided by the sum of the computing power of all of the p processors on which the MPI program is running. Rather than balance the workload using the number of rows or columns as reference, our balancing algorithm uses the number of nonzero elements.…”

Section: Load Balancingmentioning

confidence: 99%

Enhancing the performance of malleable MPI applications by using performance-aware dynamic reconfiguration

et al. 2015

View full text Add to dashboard Cite

The work in this paper focuses on providing malleability to MPI applications by using a novel performance-aware dynamic reconfiguration technique. This paper describes the design and implementation of Flex-MPI, an MPI library extension which can automatically monitor and predict the performance of applications, balance and redistribute the workload, and reconfigure the application at runtime by changing the number of processes. Unlike existent approaches, our reconfiguring policy is guided by user-defined performance criteria. We focus on iterative SPMD programs, a class of applications with critical mass within the scientific community. Extensive experiments show that Flex-MPI can improve the performance, parallel efficiency, and cost-efficiency of MPI programs with a minimal effort from the programmer.

show abstract

Section: Load Balancingmentioning

confidence: 99%

Enhancing the performance of malleable MPI applications by using performance-aware dynamic reconfiguration

et al. 2015

View full text Add to dashboard Cite

show abstract

“…An efficient approach, originally designed for homogeneous systems, consists in adapting the parallel code by software techniques which balance the load depending on the computational power of each computing unit. But to adapt parallel code to run on a heterogeneous system requires prior knowledge about the characteristics of the architecture [4], which is not always feasible.…”

Section: Related Workmentioning

confidence: 99%

“…The fraction of the workload assigned to process i is computed in Equation 2 depending on the relative computing power (RCP i ) of a process i, which is computed as the M F LOP S i divided by the total M F LOP S for all of the p processes. RCP is used to estimate workload distribution on parallel applications, since it provides a normalized value of the computational power of a process relative to the computational power of the whole system [4,7].…”

Section: Load Balancingmentioning

confidence: 99%

FLEX-MPI: An MPI Extension for Supporting Dynamic Load Balancing on Heterogeneous Non-dedicated Systems

Martín

Marinescu

Singh

et al. 2013

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This paper introduces FLEX-MPI, a novel runtime approach for the dynamic load balancing of MPI-based SPMD applications running on heterogeneous platforms in the presence of dynamic external loads. To effectively balance the workload, FLEX-MPI monitors the actual performance of applications via hardware counters and the MPI profiling interface-with a negligible overhead and minimal code modifications. Our results show that by using this approach the execution time of an application may be significantly reduced.

show abstract

“…The methodology chosen, however, allows for the implementation of balancing algorithms which may be more efficient. It is well known that to exploit a heterogeneous system the workload of every node has to be proportional to its computational power or speed [3,6]. On nondedicated systems, the computational power available to every task is a temporal parameter, α i (t), it has to be periodically measured to dynamically adapt the load attached to every processor.…”

Section: Model Of Computational Power On Nondedicated Systemsmentioning

confidence: 99%

Adaptive load balancing of iterative computation on heterogeneous nondedicated systems

et al. 2011

View full text Add to dashboard Cite

Dynamic load balancing in heterogeneous systems is a fundamental research topic in parallel computing due to the high availability of such systems. The efficient utilization of the heterogeneous resources can significantly enhance the performance of the parallel system. At the same time, adapting parallel codes to state-ofthe-art parallel computers composed of heterogeneous multinode-multicore processors becomes a very hard task because parallel codes are highly dependent on the parallel architectures. That means that applications must be tailored requiring a great deal of programming effort. We have developed the ALBIC (Adaptive Load Balancing of Iterative Computation) system that allows for the dynamic load balancing of iterative codes in heterogeneous dedicated and nondedicated Linux based systems. In order to validate the system several parallel codes have been analyzed in different scenarios. The results show that the ALBIC approach achieves better performance than the other proposal. This lightweighted library eases porting homogeneous parallel codes to heterogeneous platforms, since the code intrusion is low and the programming effort is quite reduced.

show abstract

Dealing with Heterogeneity in Load Balancing Algorithms

Cited by 14 publications

References 10 publications

Enhancing the performance of malleable MPI applications by using performance-aware dynamic reconfiguration

Enhancing the performance of malleable MPI applications by using performance-aware dynamic reconfiguration

FLEX-MPI: An MPI Extension for Supporting Dynamic Load Balancing on Heterogeneous Non-dedicated Systems

Adaptive load balancing of iterative computation on heterogeneous nondedicated systems

Contact Info

Product

Resources

About