Large-scale Parallel Machines Research Articles

A K-means-Teaching Learning based optimization algorithm for parallel machine scheduling problem

With the continuous increase of workshop production scale, traditional heuristic algorithms in solving the scheduling problem have the defects of unsatisfactory computing time and insufficient stability of the solution result. However, data mining has a good performance in solving large-scale scheduling problems. A data mining method was proposed for industrial big data to solve the problem of large-scale parallel machines scheduling. This methodology can obtain an effective initial solution for single-operation parallel machine scheduling problem by exploring the effective information in the historical scheduling data. Based on historical customer orders, the offline learning was used to continuously generate simulated data for learning, which makes up for the shortcomings of insufficient data. A TLBO framework (teaching-learning-based optimization) hybrid K-means algorithm was redesigned to enhance the accuracy of offline learning and the efficiency of data searching. In the online operation part, according to the optimal solutions for high-similarity manufacturing orders are the approximate solutions, the new customer order will be quickly matched with the most similar manufacturing order through similarity calculation, and then and then local search is performed. Finally, the globally optimal solution is obtained after screening. Experimental results show that the hybrid teaching–learning methodology can solve the large-scale parallel machines scheduling problem with a better learning performance and computational efficiency.

Transitive Reduction Approach to Large-Scale Parallel Machine Rescheduling Problem With Controllable Processing Times, Precedence Constraints and Random Machine Breakdown

This paper studied a novel parallel machine rescheduling problem with controllable processing times under machine breakdown and precedence constraints. This problem is strongly <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {NP}$ </tex-math></inline-formula> -hard, and we modeled it as a mixed-integer problem (MIP). The primal problem is decomposed into the discrete subproblem and the continuous one. We treat the discrete with the dispatching rule and analyze the continuous in the terms of mathematical programming. Pros and cons from the analysis lead us to implement the commercial solver. The proposed method is capable of efficiency and nonzero initial state. We introduce transitive reduction to cull the redundant constraints out of its directed acyclic graph (DAG) representation. Transitive reduction extends the efficiency from the dispatching rule to the solver. The proposed method can do the predictive scheduling and pick up the partial solution left by the machine breakdown in the reactive session. As a result, the complete method solves the rescheduling problem with efficiency, and allows for large cases. At last, the computational results showed that this technique significantly brought down the time and RAM consumption in using the solver. This technique allows the scheduler to solve big instances with computational economy and efficiency.

Effective heuristic for large-scale unrelated parallel machines scheduling problems

This paper is concerned with non-preemptive scheduling of large-scale unrelated parallel machines (UPM) with the objective of minimizing total weighted completion times (TWCT). We propose a sequential improvement local search algorithm using multiple-jump strategy embedded within Tabu search (TS) components for TWCT, and use a highly efficient data structure to provide a necessary and sufficient condition for local optimality of a solution. We will generate a set of large-scale test problems to evaluate the performance of proposed algorithm in term of scalability, solution quality and efficiency. The non-parametric tests of algorithm components will be used to validate the consistent performance across problem types and problem sizes in the proposed algorithm.

Decomposition method for fast computation of gigapixel-sized Fresnel holograms on a graphics processing unit cluster.

A parallel computation method for large-size Fresnel computer-generated hologram (CGH) is reported. The method was introduced by us in an earlier report as a technique for calculating Fourier CGH from 2D object data. In this paper we extend the method to compute Fresnel CGH from 3D object data. The scale of the computation problem is also expanded to 2gigapixels, making it closer to real application requirements. The significant feature of the reported method is its ability to avoid communication overhead and thereby fully utilize the computing power of parallel devices. The method exhibits three layers of parallelism that favor small to large scale parallel computing machines. Simulation and optical experiments were conducted to demonstrate the workability and to evaluate the efficiency of the proposed technique. A two-times improvement in computation speed has been achieved compared to the conventional method, on a 16-node cluster (one GPU per node) utilizing only one layer of parallelism. A 20-times improvement in computation speed has been estimated utilizing two layers of parallelism on a very large-scale parallel machine with 16 nodes, where each node has 16 GPUs.

Anatomy of machine learning algorithm implementations in MPI, Spark, and Flink

With the ever-increasing need to analyze large amounts of data to get useful insights, it is essential to develop complex parallel machine learning algorithms that can scale with data and number of parallel processes. These algorithms need to run on large data sets as well as they need to be executed with minimal time in order to extract useful information in a time-constrained environment. Message passing interface (MPI) is a widely used model for developing such algorithms in high-performance computing paradigm, while Apache Spark and Apache Flink are emerging as big data platforms for large-scale parallel machine learning. Even though these big data frameworks are designed differently, they follow the data flow model for execution and user APIs. Data flow model offers fundamentally different capabilities than the MPI execution model, but the same type of parallelism can be used in applications developed in both models. This article presents three distinct machine learning algorithms implemented in MPI, Spark, and Flink and compares their performance and identifies strengths and weaknesses in each platform.

Non-Galerkin Coarse Grids for Algebraic Multigrid

Algebraic multigrid (AMG) is a popular and effective solver for systems of linear equations that arise from discretized partial differential equations. While AMG has been effectively implemented on large scale parallel machines, challenges remain, especially when moving to exascale. In particular, stencil sizes (the number of nonzeros in a row) tend to increase further down in the coarse grid hierarchy, and this growth leads to more communication. Thus, as problem size increases and the number of levels in the hierarchy grows, the overall efficiency of the parallel AMG method decreases, sometimes dramatically. This growth in stencil size is due to the standard Galerkin coarse grid operator, $P^T A P$, where $P$ is the prolongation (i.e., interpolation) operator. For example, the coarse grid stencil size for a simple three-dimensional (3D) seven-point finite differencing approximation to diffusion can increase into the thousands on present day machines, causing an associated increase in communication costs...

An iterated greedy algorithm for the large-scale unrelated parallel machines scheduling problem

In this work, we tackle the problem of scheduling a set of jobs on a set of unrelated parallel machines with minimising the total weighted completion times as performance criteria. The iterated greedy metaheuristic generates a sequence of solutions by iterating over a constructive heuristic using destruction and construction phases. In the last few years, iterated greedy has been employed to solve a considerable number of problems. This is because it is based on a very simple principle, it is easy to implement, and it often exhibits an excellent performance. Moreover, scalability for high-dimensional problems becomes an essential requirement for modern optimisation algorithms. This paper proposes an iterated greedy model for the above-mentioned scheduling problem to tackle large-size instances. The benefits of our proposal in comparison to existing metaheuristics proposed in the literature are experimentally shown.

Overcoming Communication Latency Barriers in Massively Parallel Scientific Computation

Anton, a massively parallel special-purpose machine that accelerates molecular dynamics simulations by orders of magnitude, uses a combination of specialized hardware mechanisms and restructured software algorithms to reduce and hide communication latency. Anton delivers end-to-end internode latency significantly lower than any other large-scale parallel machine, and its critical-path communication time for molecular dynamics simulations is less than 3 percent that of the next-fastest platform.

A Fast Parallel Algorithm for Selected Inversion of Structured Sparse Matrices with Application to 2D Electronic Structure Calculations

An efficient parallel algorithm is presented for computing selected components of $A^{-1}$ where $A$ is a structured symmetric sparse matrix. Calculations of this type are useful for several applications, including electronic structure analysis of materials in which the diagonal elements of the Green's functions are needed. The algorithm proposed here is a direct method based on a block $LDL^T$ factorization. The selected elements of $A^{-1}$ we compute lie in the nonzero positions of $L+L^T$. We use the elimination tree associated with the block $LDL^T$ factorization to organize the parallel algorithm, and reduce the synchronization overhead by passing the data level by level along this tree using the technique of local buffers and relative indices. We demonstrate the efficiency of our parallel implementation by applying it to a discretized two dimensional Hamiltonian matrix. We analyze the performance of the parallel algorithm by examining its load balance and communication overhead, and show that our parallel implementation exhibits an excellent weak scaling on a large-scale high performance distributed-memory parallel machine.

ArchSim: A System-Level Parallel Simulation Platform for the Architecture Design of High Performance Computer

High performance computer (HPC) is a complex huge system, of which the architecture design meets increasing difficulties and risks. Traditional methods, such as theoretical analysis, component-level simulation and sequential simulation, are not applicable to system-level simulations of HPC systems. Even the parallel simulation using large-scale parallel machines also have many difficulties in scalability, reliability, generality, as well as efficiency. According to the current needs of HPC architecture design, this paper proposes a system-level parallel simulation platform: ArchSim. We first introduce the architecture of ArchSim simulation platform which is composed of a global server (GS), local server agents (LSA) and entities. Secondly, we emphasize some key techniques of ArchSim, including the synchronization protocol, the communication mechanism and the distributed checkpointing/restart mechanism. We then make a synthesized test of some main performance indices of ArchSim with the phold benchmark and analyze the extra overhead generated by ArchSim. Finally, based on ArchSim, we construct a parallel event-driven interconnection network simulator and a system-level simulator for a small scale HPC system with 256 processors. The results of the performance test and HPC system simulations demonstrate that ArchSim can achieve high speedup ratio and high scalability on parallel host machine and support system-level simulations for the arhitecture design of HPC systems.

A genetic algorithm based on double linear-linked-list encoding

Regarding the fault of the Genetic Algorithm based on classical encoding,which is to resolve the large-scale parallel machines batch-scheduling problem,this text proposed a method DLGA,including a Double Linear-linked-list encoding and the relevant genetic operators.This method was testified by simulating to resolve the scheduling problem of vulcanizing missions in certain tyre factory.The simulation result indicates that the indexes of scheduling performance and the feasibility of application are improved by this algorithm.

A Performance Comparison of Continuous and Discontinuous Finite Element Shallow Water Models

We present a comparative study of two finite element shallow water equation (SWE) models: a generalized wave continuity equation based continuous Galerkin (CG) model--an approach used by several existing SWE models--and a recently developed discontinuous Galerkin (DG) model. While DG methods are known to possess a number of favorable properties, such as local mass conservation, one commonly cited disadvantage is the larger number of degrees of freedom associated with the methods, which naturally translates into a greater computational cost compared to CG methods. However, in a series of numerical tests, we demonstrate that the DG SWE model is generally more efficient than the CG model (i) in terms of achieving a specified error level for a given computational cost and (ii) on large-scale parallel machines because of the inherently local structure of the method. Both models are verified on a series of analytic test cases and validated on a field-scale application.

The ASC-Alliance Projects: A Case Study of Large-Scale Parallel Scientific Code Development

Computational scientists face many challenges when developing software that runs on large-scale parallel machines. However, software-engineering researchers haven't studied their software development processes in much detail. To better understand the nature of software development in this context, the authors examined five large-scale computational science software projects operated at the five ASC-Alliance centers.

Scheduling algorithm based on evolutionary computing in identical parallel machine production line

Evolutionary programming is a kind of evolutionary computing method based on stochastic search suitable for solving system optimization. In this paper, evolutionary programming method is applied to the identical parallel machine production line scheduling problem of minimizing the number of tardy jobs, which is a very important optimization problem in the field of research on CIMS and industrial engineering, and researches on problem formulation, expression of feasible solution, methods for the generation of the initial population, the mutation and improvement on the local search ability of evolutionary programming. Computational results of different scales of problems show that the evolutionary programming algorithm proposed in this paper is efficient, and that it is fit for solving large-scale identical parallel machine production line scheduling problems, and that the quality of its solution has advantage over so far the best heuristic procedure.

Efficient Performance Prediction for Large-Scale, Data-Intensive Applications

This paper presents a simulation-based performance prediction framework for large-scale, data-intensive applications on large-scale machines. The framework consists of two components: application emulators and a suite of simulators. Application emulators provide a parameterized model of data access and computation patterns of the applications and enable changing critical application components (input data partitioning, data declustering, processing structure, etc.). The suite of simulators executes quickly on a high performance workstation to allow performance prediction of large-scale parallel machine configurations. The key to efficient simulation of very large configurations is to elide the majority of low-level hardware events while preserving data dependencies and distributions. The authors evaluate their performance prediction tool using a set of three data-intensive applications.

Massively parallel computing: A Sandia perspective

The computing power available to scientists and engineers has increased dramatically in the past decade, due in part to progress in making massively parallel computing practical and available. The expectation for these machines has been great. The reality is that progress has been slower than expected. Nevertheless, massively parallel computing is beginning to realize its potential for enabling significant breakthroughs in science and engineering. This paper provides a perspective on the state of the field, colored by the authors' experiences using large-scale parallel machines at Sandia National Laboratories. We address trends in hardware, system software and algorithms, and we also offer our view of the forces shaping the parallel computing industry.

User-Level Parallel File I/O

&#x0D; &#x0D; &#x0D; &#x0D; Parallel disk I/O subsystems are becoming more important in today’s large-scale parallel machines. Parallel disk systems provide a significant boost in I/O performance reducing the gap between processor and disk speeds. We describe a Unix-like file I/O user interface, implemented in a parallel file I/O subsystem on an MIMD machine, the nCUBE 2. Based on message passing, we develop parallel disk read/write algorithms to achieve higher parallelism when more disk drives are used. We use the closed queuing network model to analyze the effect of some tunable system parameters for our parallel file system. We then performed simulation experiments in order to obtain more realistic performance data, by comparing the original vendor-supplied file system with ours. The results indicate that the speedup in I/O performance is almost equal to the number of disk drives we use. Thus, our user-level parallel file I/O approach will provide scalable I/O performance.&#x0D; &#x0D; &#x0D; &#x0D;

Evaluation of architectural support for global address-based communication in large-scale parallel machines

Large-scale parallel machines are incorporating increasingly sophisticated architectural support for user-level messaging and global memory access. We provide a systematic evaluation of a broad spectrum of current design alternatives based on our implementations of a global address language on the Thinking Machines CM-5, Intel Paragon, Meiko CS-2, Cray T3D, and Berkeley NOW. This evaluation includes a range of compilation strategies that make varying use of the network processor; each is optimized for the target architecture and the particular strategy. We analyze a family of interacting issues that determine the performance trade-offs in each implementation, quantify the resulting latency, overhead, and bandwidth of the global access operations, and demonstrate the effects on application performance.

Evaluation of architectural support for global address-based communication in large-scale parallel machines

Large-scale parallel machines are incorporating increasingly sophisticated architectural support for user-level messaging and global memory access. We provide a systematic evaluation of a broad spectrum of current design alternatives based on our implementations of a global address language on the Thinking Machines CM-5, Intel Paragon, Meiko CS-2, Cray T3D, and Berkeley NOW. This evaluation includes a range of compilation strategies that make varying use of the network processor; each is optimized for the target architecture and the particular strategy. We analyze a family of interacting issues that determine the performance trade-offs in each implementation, quantify the resulting latency, overhead, and bandwidth of the global access operations, and demonstrate the effects on application performance.

Parallel fiber-optic SCI links

Parallel fiber-optic links promise to be a key enabling technology for distributed computing, substantially increasing transmission distance and bandwidth. The drive to include more and more processors and I/O devices in parallel-processor complexes has created a need for box-to-box interconnection links capable of delivering bandwidth comparable to system backplanes. Point-to-point links overcome the speed and cost constraints of shared buses. However, copper-based point-to-point links can reliably carry high-speed data only a limited distance. Parallel fiber-optic links substantially increase transmission distance and surpass copper's bandwidth capability. Our research addresses the integration of parallel fiber-optic data links into system designs that require a high-bandwidth link over moderate distances. To demonstrate the feasibility of using parallel fiber-optic technology as a fundamental building block in large-scale commercial and parallel machines, we constructed a link testbed.