Heterogeneous Distributed System Research Articles

An autonomic parallel strategy for exhaustive search tree algorithms on shared or heterogeneous systems

SummaryBacktracking branch‐and‐prune (BP) algorithms and their variants are exhaustive search tree techniques widely employed to solve optimization problems in many scientific areas. However, they characteristically often demand significant amounts of computing power for problem sizes representative of real‐world scenarios. Given that their search domains can often be partitioned, these algorithms are frequently designed to execute in parallel by harnessing distributed computing systems. However, to achieve efficient parallel execution times, an effective strategy is required to balance the nonuniform partition workloads across the available resources. Furthermore, with the increasing integration of servers with heterogeneous resources and the adoption of resource sharing, balancing workloads is becoming complex. This paper proposes a strategy to execute parallel BP algorithms more efficiently on even shared or heterogeneous distributed systems. The approach integrates a self‐adjusting dynamic partitioning method in the BP algorithm with a dynamic scheduler, provided by an application middleware, which manages the parallel execution while addressing any issues of imbalance. Empirical results indicate better scalability with efficiencies above 90% for instances of an application case study for the discretizable molecular distance geometry problem (DMDGP). Improvements of up to 38% were obtained in execution speed‐ups compared to a more traditional parallel BP implementation for DMDGP.

Computation Offloading Method for Large-Scale Factory Access in Edge-Edge Collaboration Mode

Large-scale manufacturing enterprises have complex business processes in their production workshops, and the edge-edge collaborative business model cannot adapt to the traditional computation offloading methods, which leads to the problem of load imbalance. For this problem, a computation offloading algorithm based on edge-edge collaboration mode for large-scale factory access is proposed, called the edge and edge collaborative computation offloading (EECCO) algorithm. First, the method partitions the directed acyclic graphs (DAGs) on edge server and terminal industrial equipment, then updates the tasks using a synchronization policy based on set theory to improve the accuracy effectively, and finally achieves load balancing through processor allocation. The experimental results show that the method shortens the processing time by improving computational resource utilization and employs a heterogeneous distributed system to achieve high computing performance when processing large-scale task sets.

Modeling the Training Iteration Time for Heterogeneous Distributed Deep Learning Systems

Distributed deep learning systems effectively respond to the increasing demand for large-scale data processing in recent years. However, the significant investment in building distributed learning systems with powerful computing nodes places a huge financial burden on developers and researchers. It will be good to predict the precise benefit, i.e., how many times of speedup it can get compared with training on single machine (or a few), before actually building such big learning systems. To address this problem, this paper presents a novel performance model on training iteration time for heterogeneous distributed deep learning systems based on the characteristics of the parameter server (PS) system with bulk synchronous parallel (BSP) synchronization style. The accuracy of our performance model is demonstrated by comparing real measurement results on TensorFlow when training different neural networks with various kinds of hardware testbeds: the prediction accuracy is higher than 90% in most cases.

Analysis of the components of information reliability of distributed systems for monitoring a critical agricultural object

The article deals with the issues of information reliability of distributed monitoring systems on the example of a critical agricultural facility. The relevance of the chosen topic is determined by the significant role of agricultural development in the country's food security, the success of which depends, in particular, on the effectiveness of monitoring systems as not an integral element of management systems. The main components of a complex heterogeneous distributed monitoring system are considered. The conclusion is summarized that the use of each of the components requires additional clarification for which component this or that reliability property is applied.

Energy-efficient distributed password hash computation on heterogeneous embedded system

This paper presents the improved version of our cool Cracker cluster (cCc), a heterogeneous distributed system for parallel and energy-efficient bcrypt password hash computation. The cluster consists of up to 8 computational units (nodes) with different performances measured in bcrypt hash computations per second [H/s]. In the cluster, nodes are low-power heterogeneous embedded systems with programmable logic containing specialized hash computation accelerators. In the experiments, we used a combination of Xilinx Zynq-series SoC boards and ZTEX 1.15y board which was initially used as a bitcoin miner. Zynq based nodes use the improved version of our custom bcrypt accelerator, which executes the most costly parts of the bcrypt hash computation in programmable logic. The cluster was formed around the famous open-source password cracking software package John the Ripper (abbr. JtR). On the communication layer, we used Message Passing Interface (MPI)library with a standard Ethernet network connecting the nodes. To mitigate the different performances among the cluster nodes and to balance the load, we developed and implemented password candidate distribution scheme based on the passwords' probability distribution, i.e. the order of appearance in the dictionary. We tested individual nodes and the cluster as a whole, trying different combinations of nodes and evaluating our distribution scheme for password candidates. We also compared our cluster with various GPU implementations in terms of performance, energy-efficiency, and price-efficiency. We show that our solution outperforms other platforms such as high-end GPUs, by a factor of at least 3 in terms of energy-efficiency and thus producing less overall cost of password attack than other platforms. In terms of the total operational costs, our cluster pays off after 4500 cracked passwords for a bcrypt hash with cost parameter 12, which makes it more appealing for real-world password-based system attacks. We also demonstrate the scalability of our cCc cluster.

Dynamic Multiworkflow Deadline and Budget Constrained Scheduling in Heterogeneous Distributed Systems

Dynamic scheduling of multiworkflow in a heterogeneous distributed system has become a crucial issue in recent days due to user-defined deadlines and budget constraints. In order to settle down this problem, we present a dynamic multiworkflow deadline and budget constrained heterogeneous scheduling (DMW-HDBS) algorithm. DMW-HDBS owns some unique features: in the task selection phase, an optimized priority selection strategy is applied to determine the priority scheduling tasks; in the stage of resource allocation, the cost and budget ratio is utilized to better balance therelationship between deadline and budget constraint; and the discard judgment mechanism is used to improve the success rate of multiworkflow scheduling. Real-world application experiments are conducted to verify that our algorithm has better scheduling performance than other competing algorithms.

Achieving Fairness-Aware Two-Level Scheduling for Heterogeneous Distributed Systems

In a heterogeneous distributed system composed of various types of computing platforms such as supercomputers, grids, and clouds, a two-level scheduling approach can be used to effectively distribute resources of the platforms to users in the first-level, and map tasks of the users in nodes for each platform in the second-level for executing many-task applications. When scheduling heterogeneous resources, service providers of the system should consider the fairness among multiple users as well as the system efficiency. However, the fairness cannot be achieved by simply distributing an equal amount of resources from each platform to every user. In this paper, we investigate how to address the fairness issue among multiple users in a heterogeneous distributed system. We present three first-level resource allocation policies of a provider affinity first policy, an application affinity first policy, and a platform affinity based round-robin policy, and two second-level task mapping policies of a most affected first policy and a co-runner affinity based round-robin policy. Using trace-based simulations, we evaluate the performance of various combinations of the first and second level scheduling policies. Our extensive simulation results demonstrate that the first-level policy plays a crucial role to achieve relatively good fairness.

Solving data collection tasks in heterogeneous distributed software intensive system using model approach

The article discusses possible approaches to data collection and information collection in multi-level distributed systems. It is suggested to use model based approach. The requirements to these models are analyzed and the system of models is proposed. The system of models describes the structure and behavior of the observed system in terms of graphs. The resource graph, the data flow graph, the control flow graph, and the query flow graph are used. These graphs are automatically constructed on the basis of information received from the observed system in the form of log files. A virtual machine that implements this approach is proposed.

A METHOD FOR IMPROVING THE QUALITY INDICATORS OF A DISTRIBUTED FORENSIC INFORMATION SYSTEM

Subject of research: the process of monitoring the quality indicators of a computer network of a forensic information system using methods for increasing the frequency resolution of acousto-optic spectrum analyzers. The aim of this work is the possibility of increasing the QoS indicators of a computer network by achieving the potential accuracy of estimating the frequencies of two simultaneously arriving radio pulses using the super-Rayleigh resolution method in acousto-optic spectrum analyzers (AOSA). The following tasks are solved in the article: analysis of various information systems of forensic examination and the tasks that they solve; analysis of the use of acousto-optic spectrum analyzers to determine the parameters of the transmitted signal in a distributed forensic system; derivation of an analytical expression for evaluating the potential accuracy of the frequency detuning value of two pulses. The following methods and models are used: models and methods of systems theory (based on the information approach), models and methods for constructing distributed systems and their components in accordance with various hierarchical levels of their organization and operating conditions; models and methods of theories of propagation and interaction of electromagnetic waves with wave structures of various physical nature. The following results were obtained: The analysis has established that forensic information systems are distributed in nature and impose high requirements on the quality of the functioning of telecommunications equipment of the corresponding computer networks (CN). To meet these requirements, it is necessary to improve the systems for monitoring, protecting and transmitting information in the corresponding CN, which is impossible without solving the problem of monitoring signal parameters. Research has been carried out on the use of AOSA to ensure control over communication channels of computer networks. It was found that most of the AOSA frequency resolution methods have limitations associated with the Rayleigh criterion, which significantly limits the potential for frequency separation of pulses. The use of the method of super-Rayleigh resolution of narrow-band pulses in terms of their arrival time is proposed. An analytical expression is obtained to estimate the potential accuracy of determining the value of the detuning in the frequency of two pulses. Conclusions. The forensic information system is a complex heterogeneous distributed system. To improve the efficiency of its functioning, it is necessary to control the parameters of the signal propagating in the network. The use of the super-Rayleigh resolution method in AOSA will increase the resolution of AOSA by increasing the frequency resolution of two non-simultaneous radio pulses of long duration.

Reliability in volunteer computing micro-blogging services

Volunteer computing is a type of large-scale heterogeneous distributed system where resources necessary to run the system are donated by volunteers. A draw back of volunteer computing is the unreliability of the donated resources, so redundancy is required to guarantee the fulfillment of tasks or the availability of data In this work we consider the problem of designing a directory service policy for a Distributed Volunteer Computing Micro-Blogging Service (DVCMBS). In such services, nodes donate storage space (repositories), which is managed by a centralized directory service that decides which nodes will store replicas of blogs, ensuring their online availability when bloggers are offline. Since nodes are under no obligation to remain online, the task of a DVCMBS directory service is to allocate blog replicas to online repositories, such that the rate of availability of all blogs is maximized. At the same time, since donated storage resources are limited and directory service operations use processing resources, minimizing the number of blog replicas generated (i.e., the inefficiency of the directory service) is critical. We present a simulation model of a DVCMBS, which uses a probabilistic sort and select approach to select host repositories for replicas and blogs to replicate. Exhaustive computational experiments analyze the trade-off between blog replica availability and efficiency, and identify the efficient directory service policies, with respect to availability and efficiency maximization.

RETRACTED ARTICLE: Design and implementation of fuzzy priority deadline job scheduling algorithm in heterogeneous grid computing

In heterogeneous distributed computing system, huge amount of attention is attracted by task scheduling process in recent days. Scheduling of task is an important problem, where, overall completion of task is completed within a shortest achievable time by scheduling dissimilar task to target processors. This research work concentrates on designing and implementing a fuzzy priority deadline based task scheduling algorithm (FPDSA) having a fuzzy deadline limitation to competent job execution. This proposed FPDSA algorithm is compared with conventional task scheduling techniques like,Improvised Deadline Scheduling Algorithm (IDSA), Earliest Deadline First (EDF) and Prioritized Based Deadline Scheduling Algorithm with respect to Average Actual Execution (AAE) and amount of Non-Delayedand Delayed Jobs. For 4000 tasks, proposed algorithm achieves 7.45%, 27.94% and 30.84% less AAE than IDSA, EDF, and PDSA and for same number of tasks, computational results by proposed FPDSA for non-delayed tasks are 0.32%, 2.17%, and 1.70% higher than IDSA, EDF, and PDSA. This enhances proposed FPDSA’s performance when compared to present scheduling algorithms and illustratesFPDSA is more appropriate scheduling technique forgrid system.

Resource-aware load balancing model for batch of tasks (BoT) with best fit migration policy on heterogeneous distributed computing systems

PurposeLoad balancing is an important issue for a heterogeneous distributed computing system environment that has been proven to be a nondeterministic polynomial time hard problem. This paper aims to propose a resource-aware load balancing (REAL) model for a batch of independent tasks with a centralized load balancer to make the solution appropriate for a practical heterogeneous distributed environment having a migration cost with the objective of maximizing the level of load balancing considering bandwidth requirements for migration of the tasks.Design/methodology/approachTo achieve the effective schedule, load balancing issues should be addressed and tackled through efficient workload distribution. In this approach, the migration has been carried out in two phases, namely, initial migration and best-fit migration. Using the best-fit policy in migrations helps in the possible performance improvement by minimizing the remaining idle slots on underloaded nodes that remain unentertained during the initial migration.FindingsThe experimental results reveal that the proposed model exhibits a superior performance among the other strategies on considered parameters such as makespan, average utilization and level of load balancing under study for a heterogeneous distributed environment.Originality/valueDesign of the REAL model and a comparative performance evaluation with LBSM and ITSLB have been conducted by using MATLAB 8.5.0.

Job scheduler for streaming applications in heterogeneous distributed processing systems

In this study, we investigated the problem of scheduling streaming applications on a heterogeneous cluster environment and, based on our previous work, developed the maximum throughput scheduler algorithm (MT-Scheduler) for streaming applications. The proposed algorithm uses a dynamic programming technique to efficiently map the application topology onto the heterogeneous distributed system based on computing and data transfer requirements, while also taking into account the capacity of the underlying cluster resources. The proposed approach maximizes the system throughput by identifying and minimizing the time incurred at the computing/transfer bottleneck. The MT-Scheduler supports scheduling applications structured as a directed acyclic graph. We conducted experiments using three Storm microbenchmark topologies in both simulation and real Apache Storm environments. In terms of the performance evaluation, we compared the proposed MT-Scheduler with the simulated round robin and the default Storm scheduler algorithms. The results indicated that the MT-Scheduler outperforms the default round robin approach in terms of both the average system latency and throughput.

Multi-GPU, Multi-Node Algorithms for Acceleration of Image Reconstruction in 3D Electrical Capacitance Tomography in Heterogeneous Distributed System.

Electrical capacitance tomography (ECT) is one of non-invasive visualization techniques which can be used for industrial process monitoring. However, acquiring images trough 3D ECT often requires performing time consuming complex computations on large size matrices. Therefore, a new parallel approach for 3D ECT image reconstruction is proposed, which is based on application of multi-GPU, multi-node algorithms in heterogeneous distributed system. This solution allows to speed up the required data processing. Distributed measurement system with a new framework for parallel computing and a special plugin dedicated to ECT are presented in the paper. Computing system architecture and its main features are described. Both data distribution as well as transmission between the computing nodes are discussed. System performance was measured using LBP and the Landweber’s reconstruction algorithms which were implemented as a part of the ECT plugin. Application of the framework with a new network communication layer reduced data transfer times significantly and improved the overall system efficiency.

A hybrid Method to augment the efficiency of Distributed Computing System by DAG -Using Finest Task Allocation with Dual Mode Processors

Distributed computing system creates or provides a platform having multiple computing nodes linked in a specified manner. On the basis of literature review of last few decades it has been noticed that most of distributed computing researchers have shown their effort to maintain load balancing between processors ,effective task scheduling and optimizing different parameters affecting execution cost and throughput .With these above scenario an additional parameter “Self reconfiguration of CPU” is also a countable parameter to augment the efficiency of distributed computing system .Through this research paper we want to present new approach of adaptive scheduling algorithm which is the mix output of effective task allocation to processor involved in computing and self-reconfiguration of those processors as per need of computing. By this proposed method we will optimize the execution cost, service rate and maximize the throughput as an outcome of organized processors consist in heterogeneous distributed computing system, resulting provide the considerable enhancement in the performance of Distributed computing environment.

A Novel Deep Learning Approach for Semantic Interoperability Between Heteregeneous Multi-Agent Systems

This article focuses on the issue of semantic interoperability in heterogeneous distributed multi-agent systems. Existing middleware technologies offer programming models that strongly combine agents’ learning models and communication models, which can lead to performance weaknesses when the number of agents is very important. Moreover, existing methods in the field of semantic interoperability solve the problem of understanding messages exchanged between distributed agents with heterogeneous ontologies, using several techniques to combine these ontologies. The first category of these methods relies on the fusion principle, others use alignment, and finally, there are those founded on Semantic Web technique. All these methods are limited to abstract concepts and do not deal with concrete concepts such as those represented by images. We propose in this paper a new approach that addresses the problem of semantic interoperability between heterogeneous distributed agents based on two principles: At first, the communication aspect of the agent from the learning aspect is separated. Then, we propose extending semantic interoperability to concrete concepts by combining two techniques: Semantic Web technology, which allows terms representing abstract concepts to be interpreted and deep learning technology, which is introduced as a new method to ensure semantic interoperability in the case of concrete concepts such as images. A detailed description of the proposed approach is provided, showing that it is very useful in solving the disadvantages of existing multi-agent platforms.

Implementation and Comparative Analysis of k-means and Fuzzy c-means Clustering Algorithms for Tasks Allocation in Distributed Real Time System

Distributing tasks to processors in distributed real time systems is an important step for obtaining high performance. Scheduling algorithms play a vital role in achieving better performance and high throughput in heterogeneous distributed real time systems. To make the best use of the computational power available, it is essential to assign the tasks to the processor whose characteristics are most appropriate for the execution of the tasks in a distributed processing system. This study develops two algorithms for clustering the heavily-communicating tasks to reduce the inter-tasks communication costs by using k-means and fuzzy c-means clustering techniques respectively. In order to minimize the system cost and response time, an algorithm is developed for the proper allocation of formed clusters to the most suitable processor. The present algorithms are collated with problems in literature. The proposed algorithms are formulated and applied to numerous numerical examples to demonstrate their effectiveness.

Topology and computational-power aware tile mapping of perfectly nested loops with dependencies on distributed systems

Nested loops are main source of the parallelism in many scientific applications. Partitioning the iteration space of nested loops with data dependencies into tiles and assigning them to processing nodes for parallel execution is essential for achieving high performance. Although most of the previous work focused on tiling on fully connected homogeneous distributed systems, some studies have been devoted to tiling on partially connected distributed systems. In this paper, we address the parallelization of perfectly nested loops with dependencies on partially connected heterogeneous distributed systems and present a topology and computational-power aware tile mapping. This work aims to take into account not only the node’s computational power when tiling iteration space of nested loops but also the exploitation of the network topology when mapping tiles to processing nodes. This approach allows minimizing the parallel execution time by improving the load balancing and minimizing the communication costs. We demonstrate the performance of proposed method by comparing it with the computational-power aware tile mapping and the topology aware tile mapping. The experimental results show that the proposed method improves the parallel execution time by up to 62% and 28% compared with the computational-power aware tile mapping and the topology aware tile mapping, respectively.

Design and topological analysis of probabilistic distributed mutual exclusion algorithm with unbiased refined ordering

The applications of distributed computing systems are pervasive in nature involving multiple shared resources. The distributed mutual exclusion algorithms of various classes are employed to control concurrency of accessing shared resources maintaining data consistency. In general, the distributed mutual exclusion algorithms are designed based on fixed or dynamic graph structures formed by a set of processes, where the distributed mutual exclusion mechanisms are realized depending upon timestamp based ordering of events or by employing token circulation in the graph. On the contrary, in large scale heterogeneous distributed systems, an aggregate set of processes can be generated under special circumstances, where processes in a group are equally eligible to enter into critical section. In order to maintain safety and liveness properties of mutual exclusion in such cases, the probabilistic characterization as well as topological analysis of aggregate set in computing space is necessary. This paper proposes a probabilistic algorithm and its topological characterization for mutual exclusion in aggregate set of processes. The analysis of failure model of strictly ordered distributed inclusion–exclusion designs is constructed in the presence of aggregate set. The unbiased probabilistic algorithm is based on two-phased elastic randomization. The algorithm is evaluated through detailed simulation and, the related probabilistic characterization in topological subspace is evaluated. A detailed comparative analysis of the algorithm with respect to other distributed mutual exclusion algorithms is presented.

Scheduling-Efficient Framework for Neural Network on Heterogeneous Distributed Systems and Mobile Edge Computing Systems

As the volume of machine learning training data sets and the quantity of model parameters continue to grow, the pattern in which machine learning models are trained alone can no longer accommodate large-scale data environments. However, distributed systems and mobile edge computing systems are unpredictable and have heterogeneous nodes, resulting in interruptions in training or low convergence rate. In addition, existing distributed machine learning frameworks cannot guarantee a good convergence rate and speedup ratio in a variety of operating environments. Considering the above shortcomings, this paper proposes an adaptive scheduling framework for machine learning based on a heterogeneous distributed system and mobile edge computing system for machine learning model optimization. The framework detects and analyzes the dynamic changes of resources in the distributed system and mobile edge computing system through the resource detection system; then, the task scheduling system adaptively modifies the environmental parameters and schedules calculations. Relevant experiments conducted with the public data set show that the robustness and scalability of the framework are significantly better than the traditional distributed machine learning framework under the premise of ensuring high convergence rate.