Load Distribution Strategy Research Articles

A new load distribution strategy for linear network with communication delays

In this paper, we propose a new load distribution strategy called ‘send-and-receive’ for scheduling divisible loads, in a linear network of processors with communication delay. This strategy is designed to optimally utilize the network resources and thereby minimizes the processing time of entire processing load. A closed-form expression for optimal size of load fractions and processing time are derived when the processing load originates at processor located in boundary and interior of the network. A condition on processor and link speed is also derived to ensure that the processors are continuously engaged in load distributions. This paper also presents a parallel implementation of ‘digital watermarking problem’ on a personal computer-based Pentium Linear Network (PLN) topology. Experiments are carried out to study the performance of the proposed strategy and results are compared with other strategies found in literature.

Comments on “Design and performance evaluation of load distribution strategies for multiple loads on heterogeneous linear daisy chain networks”

Min, Veeravalli, and Barlas have proposed strategies to minimize the overall execution time of one or several divisible loads on a heterogeneous linear network, using one or more installments [Han Min Wong, Bharadwaj Veeravalli, Scheduling divisible loads on heterogeneous linear daisy chain networks with arbitrary processor release times, IEEE Trans. Parallel Distrib. Syst. 15 (3) (2004) 273–288; Han Min Wong, Bharadwaj Veeravalli, Gerassimos Barlas, Design and performance evaluation of load distribution strategies for multiple divisible loads on heterogeneous linear daisy chain networks, J. Parallel Distrib. Comput. 65 (12) (2005) 1558–1577]. We show using a very simple example that their approach does not always produce a solution and that, when it does, the solution is often suboptimal. We also show how to find an optimal scheduling for any instance, once the number of installments per load is given. Finally, we formally prove that under a linear cost model, as in both the above-mentioned references, an optimal schedule has an infinite number of installments. Therefore such a cost model should not be used to design practical multi-installment algorithms.

Adaptive Load Distribution Strategies for Divisible Load Processing on Resource Unaware Multilevel Tree Networks

In this paper, we propose load distribution strategies for divisible loads for networked computing environments where computation and communication resource characteristics are unknown and/or vary with time. The principle on which our strategies are formulated is based on using probing loads to estimate the network characteristics and using them to determine the best possible load distribution. This work extends the adaptive strategies proposed in an earlier work to multilevel general networks. These networks, while being more challenging, also offer several opportunities that can be exploited to make the probing phase more efficient. We propose two strategies, one static, which caters to the presence of unknown parameters, and the other dynamic, which caters to both unknown as well as time-varying network parameters. The proposed strategies are robust, resilient, and easily adaptable to network fluctuations. The algorithms are also shown to have a tracking ability, a property that is important in dynamic environments. Examples are presented to illustrate the salient features of these strategies.

Optimal load distribution in nondedicated heterogeneous cluster and grid computing environments

We consider optimal load distribution in a nondedicated cluster or grid computing system with heterogeneous servers processing both generic and dedicated applications. The goal of load balancing is to find an optimal load distribution strategy for generic tasks on heterogeneous servers preloaded by different amount of dedicated tasks such that the overall average response time of generic applications is minimized. The optimization problem is solved for three different queueing disciplines, namely, dedicated applications without priorities, prioritized dedicated applications without preemption, and prioritized dedicated applications with preemption. For each case, we derive equations that permit us to find optimal load distribution of generic tasks such that their average response time is minimized.

Design and implementation of parallel video encoding strategies using divisible load analysis

The processing time needed for motion estimation usually accounts for a significant part of the overall processing time of the video encoder. To improve the video encoding speed, reducing the execution time for motion estimation process is essential. Parallel implementation of video encoding systems using either the software or the hardware approach has attracted much attention in the area of real time video coding. In this paper, we attempt to implement a video encoder on a bus network. Usually, for such a parallel system, the key concern is associated with partitioning and balancing of the computational load among the processors such that the overall processing time of the video encoder is minimized. With the use of the divisible load theory (DLT) paradigm, a strip-wise load partitioning/balancing scheme, a load distribution strategy, two implementation strategies are developed to exploit the data parallelism inherent in the video encoding process. The striking feature of our design is that,both the granularity of the load partitions and all the associated overheads caused during parallel video encoding process can be explicitly considered. This significantly contributes to the minimization of the overall processing time of the video encoder. Extensive experimental studies are carried out to test the effectiveness of the proposed strategies. The performance of the parallel video encoder is quantified using the metrics speedup and performance gain, respectively. The experimental results show that our strategies are effective for exploiting the available parallelism inherent in the video encoding process and provide a theoretical insight on how to analytically quantify and minimize the overall processing time of a parallel system. The proposed strategies can be easily extended and applied to improve other existing parallel systems.

Analysis of internal loading at multiple robotic systems

When multiple robotics systems with several sub-chains grasp a common object, the inherent force redundancy provides a chance of utilizing internal loading. Analysis of grasping space based internal loading is proposed in this work since this method facilitates understanding the physical meaning of internal loadings in some applications, as compared to usual operational space based approach. Investigation of the internal loading for a triple manipulator has been few as compared to a dual manipulator. In this paper, types of the internal loading for dual and triple manipulator systems are investigated by using the reduced row echelon method to analyze the null space of those systems. No internal loading condition is derived and several load distribution schemes are compared through simulation. Furthermore, it is shown that the proposed scheme based on grasping space is applicable to analysis of special cases such as three-fingered and three-legged robots having a point contact with the grasped object or ground.

Design and performance evaluation of load distribution strategies for multiple divisible loads on heterogeneous linear daisy chain networks

In this paper, we consider the problem of scheduling multiple divisible loads on heterogeneous linear daisy chain networks. Our objective is to design a load distribution strategy such that the total processing time of a set of loads is minimized. We assume that the set of loads are resident in one of the farthest end processors, which has a scheduler that will distribute the load to the other processors in the network. When distributing a load from the set, the distribution pattern of the previous load has to be taken into consideration to ensure that no processors are left idle and there are no collisions in the communication links. We design single and multi-installments strategies to achieve the above objective. We derive certain important conditions to determine whether an optimum solution exists. We propose two heuristic strategies when an optimum solution is unattainable. Using all the above strategies, we conduct four different simulation experiments to track the performance of strategies under several real-life situations. We conducted four different simulation experiments based on the two heuristic strategies to identify the best combination suitable for our multiple-loads distribution strategy. We also run simulations for a homogeneous system to quantify the performance under 3 different policies, that is, when the loads are (a) unsorted, (b) sorted with smallest load first (SLF) and (c) sorted with largest load first (LLF). A detailed analysis of the simulation results is presented and based on these, recommendations are made for the choice of strategies. Finally, we compare the performance of a single-load distribution strategy against the multiple-loads distribution strategy designed in this paper to quantify the exact performance gain that can be achieved. Illustrative examples are also provided for ease of understanding.

A least flow-time first load sharing approach for distributed server farm

The most critical property exhibited by a heavy-tailed workload distribution (found in many WWW workloads) is that a very small fraction of tasks make up a large fraction of the workload, making the load very difficult to distribute in a distributed system. Load balancing and load sharing are the two predominant load distribution strategies used in such systems. Load sharing generally has better response time than load balancing because the latter can exhibit excessive overheads in selecting servers and partitioning tasks. We therefore further explored the least-loaded-first (LLF) load sharing approach and found two important limitations: (a) LLF does not consider the order of processing, and (b) when it assigns a task, LLF does not consider the processing capacity of servers. The high task size variation that exists in heavy-tailed workloads often causes smaller tasks to be severely delayed by large tasks. This paper proposes a size-based approach, called the least flow-time first (LFF-SIZE), which reduces the delay caused by size variation while maintaining a balanced load in the system. LFF-SIZE takes the relative processing time of a task into account and dynamically assigns a task to the fittest server with a lighter load and higher processing capacity. LFF-SIZE also uses a multi-section queue to separate larger tasks from smaller ones. This arrangement effectively reduces the delay of smaller tasks by larger ones as small tasks are given a higher priority to be processed. The performance results performed on the LFF-SIZE implementation shows a substantial improvement over existing load sharing and static size-based approaches under realistic heavy-tailed workloads.

네트워크 프로세서를 이용한 기가비트 패킷 헤데 수집기

본 논문에서는 기가비트 트래픽에서도 높은 패킷 헤더 수집률(packet header collection ratio)을 보이는 멀티프로세서(multi-processor), 멀티쓰레드(multi-thread)를 채용한 네트워크 프로세서 기반의 패킷 헤더 수집기를 제안한다. 제안 패킷 수집기는 기가비트 트래픽 패킷 헤더를 분리하여 여러 대의 100Mbps MAC 포트로 분산하여 전달할 수 있는 구조를 가지고 있다. 제안된 구조는 고속 트래픽 처리를 위해 독창적인 버퍼관리 기법과 프로세서간 부하 분산 기법을 사용하고 있으며, 풍부한 실험을 퐁해 그 성능을 검증하였다. This paper proposes a packet header collector, based on a network processor with multi-processor and multi-threads, that shows a high throughput on gigabit network. The proposed collector has an architecture to separate packets coming from gigabit network into headers and payloads, and distribute them to multiple 100Mbit MAC ports. The architecture hiring a unique buffer management method and load distribution strategy among multiple processors is evaluated empirically in depth.

Computing BLAS level-2 operations on workstation clusters using the divisible load paradigm

The problem of executing large BLAS (basic linear algebra subprograms) Level-2 operations, such as matrix-vector products, in a network-based distributed computing environment composed of a bus-oriented workstation cluster is considered. Unlike previous contributions, we take into account the fact that workstations, as against mainframe computers, are not equipped with communication coprocessors or front-ends, precluding any possibility of communication off-loading. Communication delays, which are significant in workstation clusters due to limited bandwidth availability, are specifically accounted for. This aspect is generally ignored in most performance analysis of parallel computing systems. The important contribution of this study is to show that the optimal load partitioning, and the subsequent performance of the network, depends critically on network bandwidth, computing capacity, and load characteristics. We design load distribution strategies for three cases (no communication, broadcast communication, and multicast communication) based on closed-form solutions of the optimal load partitioning problem and also present extensive and complete asymptotic analysis with respect to several parameters of the load and the system. Necessary and sufficient conditions for feasible and optimal load sharing are also derived. A trade-off study between the optimal number of workstations and the bandwidth of the bus is also presented.

Quantized load distribution for tree and bus-connected processors

Divisible load analysis is a valuable tool for generating solutions to data-partitioning and distribution/scheduling problems for data-parallel applications. This paper addresses an essential step required for applying these solutions to real-life problems where computing loads are multiples of some fundamental problem-specific non-divisible load unit. The algorithms that are proposed to this end, are suitable for both single and multi-installment strategies. The worst-case performances of the algorithms are derived for two cases: single installment on a single-level tree and multiple installments on a bus network. Finally, an estimation on the expected performance of the algorithms is obtained from rigorous simulation tests. The extensive analysis that accompanies these tests, depicts many aspects of the parallel computation associated with parallel machine architectures and the load distribution strategies (single- or multi-installment) used.

Scheduling divisible loads on heterogeneous linear daisy chain networks with arbitrary processor release times

The problem of distributing and processing a divisible load in a heterogeneous linear network of processors with arbitrary processors release times is considered. A divisible load is very large in size and has computationally intensive CPU requirements. Further, it has the property that the load can be partitioned arbitrarily into any number of portions and can be scheduled onto processors independently for computation. The load is assumed to arrive at one of the farthest end processors, referred to as boundary processors, for processing. The processors in the network are assumed to have nonzero release times, i.e., the time instants from which the processors are available for processing the divisible load. Our objective is to design a load distribution strategy by taking into account the release times of the processors in such a way that the entire processing time of the load is a minimum. We consider two generic cases in which all processors have identical release times and when all processors have arbitrary release times. We adopt both the single and multiinstallment strategies proposed in the divisible load scheduling literature in our design of load distribution strategies, wherever necessary, to achieve a minimum processing time. Finally, when optimal strategies cannot be realized, we propose two heuristic strategies, one for the identical case, and the other for nonidentical release times case, respectively. Several conditions are derived to determine whether or not optimal load distribution exists and illustrative examples are provided for the ease of understanding.

Theoretical and experimental study on large size image processing applications using divisible load paradigm on distributed bus networks

In this paper, we present a theoretical and an experimental study on the problem of scheduling the processing of a very large size image on a cluster of processors interconnected via a bus network. We use the divisible load paradigm, referred to as divisible load theory (DLT), to schedule the entire processing of an image onto the processors. The objective is to minimize the total processing time of the entire image that is submitted to the bus network system. Here, we consider edge detection as an example of an image processing application, which qualifies to use a divide-and-process-like strategy, supported by DLT model. We first present the analysis of the fundamental load scheduling problem for a bus network consisting of a heterogeneous set of processors. We then implement the load distribution strategy on a homogeneous personal computer-based networking environment consisting of Pentium II machines configured in a bus topology, to verify these theoretical findings. The load (image) distribution strategy is obtained from the DLT analysis and the entire image is distributed among the available machines on the network. We also developed a software support system that could be as flexible as possible in terms of scalability of the network and the load (image) size. We present a detailed comparative study of the theoretical and experimental results. Also, we present a discussion on our experience and provide directions for possible future extensions of this work.

The kinematics for redundantly actuated omnidirectional mobile robots

AbstractOmnidirectional mobile robots have been popularly employed in several application areas. However, the kinematics and singularity analysis for these systems have not been clearly identified, especially for the redundantly actuated case, which is common in current omnidirectional mobile robots. In light of this fact, this article introduces two different kinematic approaches for a typical omnidirectional mobile robot having three caster wheels, and examines singularity configurations of such systems. Then, a singularity‐free load‐distribution scheme for a redundantly actuated three‐wheeled omnidirectional mobile robot is proposed. Through simulation, several advantages of the redundantly actuated mobile robot (singularity avoidance, input‐load saving, and exploiting several subtasks) are presented. © 2002 Wiley Periodicals, Inc.

Incremental Transmission Loss Allocation under Pool Dispatch

Incremental transmission loss analysis has been used for decades, but recent interest in its application to loss allocation calls for new in-depth results. This paper demonstrates that, for incremental methods to be applied correctly in loss allocation, it is first necessary to specify the load distribution and loss supply strategies. Incremental loss allocation among bus power injections is shown to be arbitrary and, therefore, open to challenge as discriminatory. Loss allocation is possible among incremental loads and/or generators, but the proportion of the total losses assigned to either one is arbitrary. Unique, nonarbitrary incremental loss allocations are, however, possible among the "equivalent" incremental bilateral exchanges between generators and loads. From these basic components it is possible then to calculate the allocation among generators or loads in any specified proportion. The main results, although developed initially for small increments, are extended to large variations. Finally, a general incremental loss allocation algorithm is developed and tested.

Incremental transmission loss allocation under pool dispatch

Incremental transmission loss analysis has been used for decades, but recent interest in its application to loss allocation calls for new in-depth results. This paper demonstrates that, for incremental methods to be applied correctly in loss allocation, it is first necessary to specify the load distribution and loss supply strategies. Incremental loss allocation among bus power injections is shown to be arbitrary and, therefore, open to challenge as discriminatory. Loss allocation is possible among incremental loads and/or generators, but the proportion of the total losses assigned to either one is arbitrary. Unique, nonarbitrary incremental loss allocations are however possible among the "equivalent" incremental bilateral exchanges between generators and loads. From these basic components it is possible then to calculate the allocation among generators or loads in any specified proportion. The main results, although developed initially for small increments, are extended to large variations. Finally, a general incremental loss allocation algorithm is developed and tested.

Kinematic Modeling of Omni-Directional Mobile Robots as Parallel Manipulator

Omni-directional mobile robots have been popularly employed in several applicationareas. However, the kinematics and singularity analysis for these systems has not been clearly identified. This paper introduces two different kinematic approaches for omni-directional mobile robots. Then, a singular-free load distribution scheme for redundantly actuated three-wheeled omni-directional mobile robot is proposed. Throughsimulation, several advantages of redundantly actuated mobile robot in aspect of singularity avoiance, input-load saving, and exploiting several subtasks are presented.

Optimizing dense linear algebra algorithms on heterogeneous machines

This paper addresses the execution of inherently sequential linear algebra algorithms namely LU factorization, tridiagonal reduction and the symmetric QR factorization algorithm used for eigenvector computation, which are significant building blocks for applications in our target image processing and analysis domain. These algorithms present additional difficulties to optimize the processing time due to the fact that the computational load for data matrix columns increases with their index, requiring a fine tuned load assignment and distribution. We present an efficient methodology to determine the optimal number of processors to be used in a computation, as well as a new static load distribution strategy that achieves better results than other algorithms developed for the same purpose.

QoS-based Architectures for Geographically Replicated Web Servers

In this paper, we discuss strategies for providing World Wide Web service users with adequate Quality of Service (i>QoS). We argue that QoS can be provided by distributing the service requests processing load among replicated Web servers (i>WSs), that can be geographically distributed across the Internet. In order to support our argument, we compare and contrast several load distribution strategies, and assess their effectivness when deployed within the context of a geographically replicated Web service; the principal figure of merit we use in this assessment is the response time experienced by the users of that service. As a result of this comparison, we propose a specific strategy, named i>QoS-based, that implements load distribution among WS replicas by binding a user to the replica that provides the shortest user response time. We examine several architectures that exploit our QoS-based strategy. Two of these architectures, named, respectively, i>Browser-based and i>Load Distribution-i>based, are described in detail as they are particularly appropriate for implementing our strategy.

Design and analysis of load distribution strategies with start-up costs in scheduling divisible loads on distributed networks

Optimal distribution of divisible loads in bus and linear networks is considered in this paper. The problem of minimizing the processing time is investigated by including all the overhead components that could penalize the performance of the system, in addition to the inherent communication and computation delays. In the case of bus networks, closed-form solutions for the processing time are derived and the influence of overheads on the optimal processing time is analyzed. For the first time in this domain of research, the concept of sequencing is realized for the case of bus networks. Through rigorous analysis, an optimal sequence to distribute the load among the processors is identified, whenever it exists. In the case when such an optimal sequence fails to exist, we present an O(m) greedy algorithm to obtain a suboptimal sequence based on some important properties of the overhead factors. In the case of linear networks, we consider two different scenarios, namely, when the processors are equipped with front-end and when they are not equipped with front-end. As in the case of bus networks, here too, necessary and sufficient conditions for the existence of optimal processing time are derived and performance study is carried out. Numerical examples are presented to illustrate the concepts.