Provisioning and Performance Evaluation of Parallel Systems with Output Synchronization

In this study we propose a method using multi layer perceptron (MLP) neural networks to evaluate and predict the performance of parallel systems and report our findings. Artificial neural networks may provide a good alternative to conventional methods in terms of identifying the contribution of individual system and application parameters to performance. Neural network models presented here are used to predict the computational and communication performance of parallel applications running on different platforms. Two applications are considered: the first one is a 2-Dimensional Fast Fourier Transform (FFT) application that requires intensive data exchange between processors, which is valuable for communication performance tests and the second one is a Monte Carlo application which can be classified as a typical floating-point application. There are two types of data used to train, validate and test the neural network models. A large portion of the input data composed from real measurements taken on SunSparc workstations. To enhance the available data, results obtained by modeling some unavailable systems into PACE (the Performance Analysis and Characterization Environment) have been also included.

Efficient performance evaluation of parallel systems

In a Fork-Join (FJ) queueing system an upstream fork station splits incoming jobs into N tasks to be further processed by N parallel servers, each with its own queue; the response time of one job is determined, at a downstream join station, by the maximum of the corresponding tasks' response times. This queueing system is useful to the modelling of multi-service systems subject to synchronization constraints, such as MapReduce clusters or multipath routing. Despite their apparent simplicity, FJ systems are hard to analyze. This paper provides the first computable stochastic bounds on the waiting and response time distributions in FJ systems. We consider four practical scenarios by combining 1a) renewal and 1b) non-renewal arrivals, and 2a) non-blocking and 2b) blocking servers. In the case of non blocking servers we prove that delays scale as O(logN), a law which is known for first moments under renewal input only. In the case of blocking servers, we prove that the same factor of log N dictates the stability region of the system. Simulation results indicate that our bounds are tight, especially at high utilizations, in all four scenarios. A remarkable insight gained from our results is that, at moderate to high utilizations, multipath routing 'makes sense' from a queueing perspective for two paths only, i.e., response times drop the most when N = 2; the technical explanation is that the resequencing (delay) price starts to quickly dominate the tempting gain due to multipath transmissions.

ES is a tool for estimating the execution times of parallel algorithms on MIMD parallel systems. ES allows the user to model arbitrary task execution times, explicit task precedence and synchronization constraints, resource contention among tasks, and a variety of scheduling policies for shared resources. Given a model of a parallel algorithm and a parallel system, ES constructs a sequencing tree that represents some or all of the possible sequences of events that may occur during the execution of the algorithm on the system, and uses it to estimate the mean and standard deviation of the execution time of the parallel algorithm. The authors compare estimates generated by ES to measurements made of a parallel mergesort executing on an Intel iPSC/860 hypercube. >

Performance evaluation of parallel systems

Parallel computer systems are among the most complex of man's creations, making satisfactory performance characterization difficult. Despite this complexity, there are strong, indeed, almost irresistible, incentives to quantify parallel system performance using a single metric. The fallacy lies in succumbing to such temptations. A complete performance characterization requires not only an analysis of the system's constituent levels, it also requires both static and dynamic characterizations. Static or average behavior analysis may mask transients that dramatically alter system performance. Although the human visual system is remarkedly adept at interpreting and identifying anomalies in false color data, the importance of dynamic, visual scientific data presentation has only recently been recognized Large, complex parallel system pose equally vexing performance interpretation problems. Data from hardware and software performance monitors must be presented in ways that emphasize important events while eluding irrelevant details. Design approaches and tools for performance visualization are the subject of this paper.

Modern network devices need to perform deep packet inspection at high speed for security and application- specific services. For this purpose, regular expressions are used, due to their high expressive power, and Deterministic Finite Automata (DFAs) are adopted to match them. Many works have been proposed to improve DFAs, especially in terms of memory consumption and speed. Instead, we address another issue: the scalability of DFAs to parallel systems and their buffer requirements. To our knowledge, a single attempt to parallelize DFA walk on regular multicore systems (which ex- ploits speculation with limited efficiency) has been proposed in literature. We propose a solution in which a number of processing units are committed to walk in parallel a DFA for the same packet; at this aim, sampling techniques on both text and regular expressions are adopted. This scheme is the first in literature that proposes effective parallelization of DFA walk, hence allowing for packet processing time reduction and less memory for reordering buffers. The result is that speed scales as the number of processing units.

Summary form only given. Interconnection topology plays a major role in achieving high performance in parallel systems. In the past, many interconnection topologies have been proposed - some most popular being hypercube, torus, mesh, De Bruijn networks, etc. Recently, some efficient interconnection topologies called Gaussian networks and Eisenstein-Jacobi networks have been introduced. The topologies of these networks are based on the concepts of Gaussian and EJ integers. In this talk, first a brief overview of these number theory concepts including how to obtain modulo of a given Gaussian or EJ number is given. Then, the interconnection topology, the topological properties, Hamiltonian decomposition of these networks, routing and broadcasting algorithms, resource placement algorithms, etc. are presented.

Performance of combined solar-heat pump systems

This paper proposes a reinforcement learning approach to the output synchronization problem for heterogeneous leader-follower multi-agent systems, where the system dynamics of all agents are completely unknown. First, to solve the challenge caused by unknown dynamics of the leader, we develop an experience-replay learning method to estimate the leader's dynamics, which only uses the leader's past state and output information as training data. Second, based on the newly estimated leader's dynamics, we design an event-triggered observer for each follower to estimate the leader's state and output. Furthermore, the experience-replay learning method and the event-triggered leader observer are co-designed, which ensures the convergence and Zeno behavior exclusion. Subsequently, to free the followers from reliance on system dynamics, a data-driven adaptive dynamic programming (ADP) method is presented to iteratively derive the optimal control gains, based on which we design a policy iteration (PI) algorithm for output synchronization. Finally, the proposed algorithm's performance is validated through a simulation.

Input/Output (I/O) operations can represent a significant proportion of the run-time of parallel scientific computing applications. Although there have been several advances in file format libraries, file system design and I/O hardware, a growing divergence exists between the performance of parallel file systems and the compute clusters that they support. In this paper, we document the design and application of the RIOT I/O toolkit (RIOT) being developed at the University of Warwick with our industrial partners at the Atomic Weapons Establishment and Sandia National Laboratories. We use the toolkit to assess the performance of three industry-standard I/O benchmarks on three contrasting supercomputers, ranging from a mid-sized commodity cluster to a large-scale proprietary IBM BlueGene/P system. RIOT provides a powerful framework in which to analyse I/O and parallel file system behaviour—we demonstrate, for example, the large file locking overhead of IBM's General Parallel File System, which can consume nearly 30% of the total write time in the FLASH-IO benchmark. Through I/O trace analysis, we also assess the performance of HDF-5 in its default configuration, identifying a bottleneck created by the use of suboptimal Message Passing Interface hints. Furthermore, we investigate the performance gains attributed to the Parallel Log-structured File System (PLFS) being developed by EMC Corporation and the Los Alamos National Laboratory. Our evaluation of PLFS involves two high-performance computing systems with contrasting I/O backplanes and illustrates the varied improvements to I/O that result from the deployment of PLFS (ranging from up to 25× speed-up in I/O performance on a large I/O installation to 2× speed-up on the much smaller installation at the University of Warwick).

The author presents the architecture of a new family of parallel operating systems based on an alternative design approach when compared to multiprogramming, namely, generic construction of parallel virtual machines. This construction is made according to needs of models for parallel programming which shall be supported as subsystems by the parallel operating system kernel. The author presents the PAROS architecture and its generic kernel ParX which has been designed for message passing parallel computers. Several parallel virtual machines are also described, such as for instance a diffusion virtual machine, and a message passing-shared data virtual machine. >

In this paper we propose a high performance parallel file system over iSCSI (iPVFS) for cluster computing. iPVFS provides a cost-effective solution for heterogeneous cluster environment by dividing a set of I/O servers into two groups, one group with higher performance servers as I/O nodes, while another group with relatively lower performance machines serves as storage target nodes. This combination provides a higher aggregate performance because of the cooperative cache among different target nodes. We have developed a model to analyze iPVFS. Our simulation results show that using same number of total nodes, iPVFS outperforms PVFS for both small requests and large requests under different workloads

The authors report on the use of formal methods for the development of parallel operating systems for two experimental declarative systems over a five-year period. A common specification approach has evolved as part of the development of these two very different systems: one being for a parallel graph reduction machine and written in a functional language enhanced with state-based objects, the other was written in C/sup ++/. A brief overview of each system is given before concentrating on the use of formal methods. A description is given of how both a technique for formally specifying sequential systems (VDM) and a technique for specifying concurrent systems (temporal logic) have been used together. In both cases, the issue of verification is addressed. >

This paper describes a basic analysis for mission performance of parallel hybrid turbofan systems. A series of equations is derived to illustrate the concept of parallel hybrid performance showing that there is an optimum range for a given battery size that corresponds to the exact usage of the battery capacity stored on the aircraft. The fundamental concepts were then demonstrated on a more detailed problem using a performance model of the Electrically Variable Engine developed by Rolls-Royce on a high aspect ratio vehicle. Trends from this study again demonstrate the concepts derived in the preliminary theoretical analysis. Finally, a standard method for presenting hybrid data is proposed called the “money chart” which is a means for showing all of the required performance parameters of a parallel hybrid system to compute other metrics of interest as desired.