Static Load Balancing Scheme Research Articles

A Competitive Analysis of Load Balancing Strategiesfor Parallel Ray Tracing

This paper examines the effectiveness of load balancing strategies for ray tracing on large parallel computer systems and cluster computers. Popular static load balancing strategies are shown to be...

Parallelization and load balancing of a comprehensive atmospheric chemistry transport model

Chemistry transport models are generally claimed to be well suited for massively parallel processing on distributed memory architectures since the arithmetic-to-communication ratio is usually high. However, this observation proves insufficient to account for an efficient parallel performance with increasing complexity of the model. The modeling of the local state of the atmosphere ensues very different branches of the modules' code and greater differences in the computational work load and, consequently, runtime of individual processors occur to a much larger extent during a time step than reported for meteorological models. Variable emissions, changes in actinic fluxes, and all processes associated with cloud modeling are highly variable in time and space and are identified to induce large load imbalances which severely affect the parallel efficiency. This is more so, when the model domain encompasses more heterogeneous meteorological or regional regimes, which impinge dissimilarly on simulations of atmospheric chemistry processes. These conditions hold for the EURAD model applied in this study, which covers the European continental scale as integration domain. Based on a master-worker configuration with a horizontal grid partitioning approach, a method is proposed where the integration domain of the individual processors is locally adjusted to accommodate for load imbalances. This ensures a minimal communication volume and data exchange only with the next neighbors. The interior boundary adjustments of the processors are combined with routine boundary exchange which is required each time step anyway. Two dynamic load balancing schemes were implemented and compared against a conventional equal area partition and a static load balancing scheme. The methods are devised for massively parallel distributed memory computers of both, Single and Multiple Instruction stream Multiple Data stream (SIMD, MIMD) types. A midsummer episode of highly elevated ozone concentrations over parts of Europe was taken as test case. The dynamic load balancing approaches were found to perform significantly better and reduce idle times of the processors considerably. The efficiency was raised from to 62% for a 128 processor configuration.

Comparison of dynamic and static load-balancing strategies in heterogeneous distributed systems

Although dynamic load-balancing strategies have the potential of performing better than static strategies, they are inevitably more complex. Their complexity and the overheads involved may negate their benefits. A heterogeneous distributed system, with computers of different processing capability but the same functionality, has been examined for two dynamic and two static policies. The results show that both the dynamic and the static policies provide dramatic performance improvements. However, they show that, contrary to common belief, the performance provided by the static policies is not much inferior to that provided by the dynamic policies. Furthermore, if the overheads in load balancing are not negligibly small, static policies are more stable and can offer better performance than dynamic policies.

Integrating Parallelization Strategies for Linkage Analysis

We describe a second parallel implementation of the ILINK program from the LINKAGE package that improves on our previous implementation. To improve running time we integrated the strategy of parallel estimation of the gradient at a candidate recombination fraction vector with a previously implemented strategy for evaluating in parallel the likelihood at one vector. We also integrated an adaptive loadbalancing strategy in conjunction with our previous static loadbalancing strategy. We implemented a strategy for partitioning input pedigrees, but this slowed down the program; we give some evidence for what the problems are. To best exploit parallelism at all levels of the program and to take advantage of both coarse-grain and fine-grain parallelism it is necessary to combine multiple algorithmic strategies.

A SIMD approach to parallel heuristic search

Serial search algorithms often exhibit exponential run times and may require an exponential amount of storage as well. Thus, the design of parallel search algorithms with limited memory is of obvious interest. This paper presents an efficient SIMD parallel algorithm, called IDPS (for iterative-deepening parallel search). At a broad level IDPS is a parallel version of IDA∗. While generically we have called our algorithm an IDPS, performance of four variants of it has been studied through experiments conducted on the well-known test-bed problem for search algorithms, namely the Fifteen Puzzle. During the experiments, data were gathered under two different static load balancing schemes. Under the first scheme, an unnormalized average efficiency of approximately 3 4 was obtained for 4K, 8K, and 16K processors. Under the second scheme, unnormalized average efficiencies of 0.92 and 0.76, and normalized average efficiencies of 0.70 and 0.63 were obtained for 8K and 16K processors, respectively. We show (as shown previously only for MIMD machines) that for admissible search, high average speedup can be obtained for problems of significant size. We believe that this research will enhance AI problem solving using parallel heuristic search algorithms.

An approximation to the response time for shortest queue routing

In this paper we derive an approximation for the mean response time of a multiple queue system in which shortest queue routing is used. We assume there are Κ identical queues with infinite capacity and service times that are exponentially distributed. Arrivals of jobs to this system are Poisson and are routed to a queue of minimal length. We develop an approximation which is based on both theoretical and experimental considerations and, for Κ ≤ 8, has an relative error of less than one half of one percent when compared to simulation. For Κ = 16, the relative error is still acceptable, being less than 2 percent. An application to a model of parallel processing and a comparison of static and dynamic load balancing schemes are presented.

Load balancing strategies for a parallel ray-tracing system based on constant subdivision

Static and dynamic load balancing strategies for a multiprocessor system for a ray tracing algorithm based on constant subdivision are presented. An object space is divided into regular cubes (subspaces), whose boundary planes are perpendicular to the coordinate axes, and these are allocated to the processors in the system. Here, load balancing among the processors is the most important problem. Firstly, in a category of static load balancing, strategies for mapping the subspaces into the processors are evaluated by simulation. Moreover, we propose a hierarchical multiprocessor system in order to realize dynamic load balancing with the static one. Its architecture can overcome the limitation of the static load balancing in a large scale multiprocessor system.

Design of partially replicated distributed database systems: an integrated methodology

The objective of this research is to develop and integrate tools for the design of partially replicated distributed database systems. Many existing tools are inappropriate for designing large-scale distributed databases due to their large computational requirements. Our goal is to develop tools that solve the design problems reasonably quickly, typically by using heuristic algorithms that provide approximate or near-optimal solutions. In developing this design methodology, we assume that information regarding the types of user requests and their rates of arrival into the system is known a priori . The methodology assumes a general model for transaction execution. In this paper we discuss three aspects of the design methodology: the data allocation problem, the use of a static load-balancing scheme in coordination with the allocation scheme, and the design evaluation and review step. Our methodology employs iterative design techniques using performance evaluation as a means to iterate.