Data Caching Based Transfer Optimization in Large Scale Networks

Parallel file systems (PFSes) play an essential role in high performance computing. To ensure the integrity, many PFSes are designed with a checker component, which serves as the last line of defense to bring a corrupted PFS back to a healthy state. Motivated by real-world incidents of PFS corruptions, we perform a fine-grained study on the capability of PFS checkers in this paper. We apply type-aware fault injection to specific PFS structures, and examine the detection and repair policies of PFS checkers meticulously via a well-defined taxonomy. The study results on two representative PFS checkers show that they are able to handle a wide range of corruptions on important data structures. On the other hand, neither of them is perfect: there are multiple cases where the checkers may behave sub-optimally, leading to kernel panics, wrong repairs, etc. Our work has led to a new patch on Lustre. We hope to develop our methodology into a generic framework for analyzing the checkers of diverse PFSes, and enable more elegant designs of PFS checkers for reliable high-performance computing.

The Importance of the Federal Government’s Role in High-Performance Computing

Nowadays, supercomputers play an essential role in high-performance computing. In general, modern supercomuputers are built as a cluster system, which is a system of multiple computers interconnected on a network. In coding a parallel program on such a cluster system, MPI (Message Passing Interface) is utilized. In this paper, we aim to reduce the execution time of MPI Allreduce, a frequently used MPI collective communication in many simulation codes. To this end, we have integrated network programmability by Software Defined Networking into MPI Allreduce so that it effectively uses the bandwidth of the interconnect of the cluster system. An experiment conducted on a cluster system with fat-tree interconnect indicates that our proposed MPI Allreduce is superior to MPI Allreduce in OpenMPI implementations.

Today, based on fast development especially in Network-on-Chip (NoC)-based many-core systems, the task scheduling problem plays a critical role in high-performance computing. It is an NP-hard problem. The complexity increases further when the scheduling problem is applied to heterogeneous platforms. Exploring the whole search space in order to find the optimal solution is not time efficient, thus metaheuristics are mostly used to find a near-optimal solution in a reasonable amount of time. We propose a compound method to select the best near-optimal task schedule in the heterogeneous platform in order to minimize the execution time. For this, we combine a new parallel meta-heuristic method with a greedy scheme. We introduce a novel metaheuristic method for near-optimal scheduling that can provide performance guarantees for multiple applications implemented on a shared platform. Applications are modeled as directed acyclic task graphs (DAG) for execution on a heterogeneous NoC-based many-core platform with given communication costs. We introduce an order-based encoding especially for pipelined operation that improves (decreases) execution time by more than 46%. Moreover, we present a novel multi-population method inspired by both genetic and imperialist competitive algorithms specialized for the scheduling problem, improving the convergence policy and selection pressure. The potential of the approach is demonstrated by experiments using a Sobel filter, SUSAN filter, RASTA-PLP, and JPEG encoder as real-world case studies.

It may not be feasible for sensor networks monitoring nature and inaccessible geographical regions to include powered sinks with Internet connections. We consider the scenario where sinks are not present in large-scale sensor networks, and unreliable sensors have to collectively resort to storing sensed data over time on themselves. At a time of convenience, such cached data from a small subset of live sensors may be collected by a centralized (possibly mobile) collector. In this paper, we propose a decentralized algorithm using fountain codes to guarantee the persistence and reliability of cached data on unreliable sensors. With fountain codes, the collector is able to recover all data as long as a sufficient number of sensors are alive. We use random walks to disseminate data from a sensor to a random subset of sensors in the network. Our algorithms take advantage of the low decoding complexity of fountain codes, as well as the scalability of the dissemination process via random walks. We have proposed two algorithms based on random walks. Our theoretical analysis and simulation-based studies have shown that, the first algorithm maintains the same level of fault tolerance as the original centralized fountain code, while introducing lower overhead than naive random-walk based implementation in the dissemination process. Our second algorithm has lower level of fault tolerance than the original centralized fountain code, but consumes much lower dissemination cost.

Recently, Egypt has recognized the pivotal role of High Performance Computing in advancing science and innovation. Additionally, Egypt realizes the importance of collaboration between different institutions and universities to consolidate their own computational and data resources into a unified platform to serve different disciplines (e.g., scientific, industrial, governmental). Otherwise, additional resources would be needed to be purchased with the associated cost, effort, and time difficulties (e.g., setup, administration, maintenance, etc.). Thus, this paper delves into the architecture and capabilities of the EN-HPCG grid using two different workload management systems: (i) Slurm (Open-Source) and (ii) PBS Pro (Licensed). This paper compares the performance of the grid between Slurm and PBS Pro in specific high-throughput computing (HTC) applications using the NAS Grid parallel benchmark (NGB) to determine which workload manager is more suitable for EN-HPCG. The evaluation includes grid-level performance metrics such as throughput, and the number of tasks completed as a function of time. Also, the presented methodology aims to assist potential partners in their decision-making process to join the EN-HPCG grid, with a focus on the site speed-up metric. Our results showed that, unless an open-source solution without cost and license problems is an obligation (in which case, Slurm is the viable solution), then it is not advisable to integrate a cluster with high-speed hardware with a cluster possessing outdated hardware when using the Slurm scheduler. In contrast, the PBS Pro scheduler takes into account online decision-making in a dynamic environment using a unified grid.

Multiobjective shortest path problem (MSPP) is one of the most critical issues in network optimization, aimed at identifying all efficient paths across conflicting objectives. Nowadays, existing methods face substantial bottlenecks in addressing the diverse preferences of decision makers and high spatiotemporal overhead caused by the calculation process, particularly in cases with large-scale networks. To overcome these obstacles, a generalized MSPP in large-scale networks is investigated with the aim of solving it with diverse preferences of decision makers satisfied and low spatiotemporal overhead. Toward this end, with a novel concept, the generalized dominance relation is introduced, and the generalized multiobjective shortest path algorithm via the generalized dynamic programming approach is developed. Moreover, the H-reducible technique is further employed to accelerate the convergence of the proposed algorithm. Additionally, several rigorous proofs are provided for the conclusions that all efficient paths could be found within a tolerable time by the developed algorithm and the algorithm could be implemented in a distributed manner under mild assumptions. Finally, numerous routing experiments are conducted on large-scale communication networks for demonstrating the effectiveness and competitiveness of our approach.

Multi-GPU machines are being increasingly used in high-performance computing. Each GPU in such a machine has its own memory and does not share the address space either with the host CPU or other GPUs. Hence, applications utilizing multiple GPUs have to manually allocate and manage data on each GPU. Existing works that propose to automate data allocations for GPUs have limitations and inefficiencies in terms of allocation sizes, exploiting reuse, transfer costs, and scalability. We propose a scalable and fully automatic data allocation and buffer management scheme for affine loop nests on multi-GPU machines. We call it the Bounding-Box-based Memory Manager (BBMM). BBMM can perform at runtime , during standard set operations like union, intersection, and difference, finding subset and superset relations on hyperrectangular regions of array data (bounding boxes). It uses these operations along with some compiler assistance to identify, allocate, and manage data required by applications in terms of disjoint bounding boxes. This allows it to (1) allocate exactly or nearly as much data as is required by computations running on each GPU, (2) efficiently track buffer allocations and hence maximize data reuse across tiles and minimize data transfer overhead, and (3) and as a result, maximize utilization of the combined memory on multi-GPU machines. BBMM can work with any choice of parallelizing transformations, computation placement, and scheduling schemes, whether static or dynamic. Experiments run on a four-GPU machine with various scientific programs showed that BBMM reduces data allocations on each GPU by up to 75% compared to current allocation schemes, yields performance of at least 88% of manually written code, and allows excellent weak scaling.

Optimization of road space allocation (RSA) from a network perspective is computationally challenging. Analogues to the Network Design Problem (NDP), RSA can be classified as a NP-hard problem. In large scale networks when the number of alternatives increases exponentially, there is a need for an efficient method to reduce the number of alternatives as well as a computational approach to reduce the computer execution time of the analysis. A heuristic algorithm based on Genetic Algorithm (GA) is proposed to efficiently select Transit Priority Alternatives (TPAs). In order to reduce the execution time, the GA is modified to implement two parallel processing techniques: A High Performance Computing (HPC) technique using Multi-threading (MT) and a High Throughput Computing (HTC) technique. The advantages and limitations of the MT and HTC techniques are discussed. Moreover, the proposed framework allows for a TPA to be analyzed by a commercial package which is a significant provision for large scale networks in practice.KeywordsCentral Processing UnitHigh Performance ComputingNetwork Design ProblemLarge Scale NetworkParallel Genetic AlgorithmThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Latency-optimized high performance Data Vortex optical switching network

Energy efficiency has become the first priority when designing high-performance computing systems, and caches play a key role in improving the performance. However, as caches represent a dominant proportion of the total area in a processor, they dissipate a significant amount of the total energy consumption. A non-unified data cache architecture is suggested to tackle this issue without sacrificing performance. This architecture is based on the principle of providing a smaller cache for a group of frequently used references, which does not require a large cache to provide a high hit rate. Thus, the non-unified design reduces the number of accesses to a relatively large L1 data cache, thereby saving energy and enabling efficient performance. This paper examines the energy consumption of a non-unified data cache design as compared to that of a conventional unified data cache. As a powerful system requires allocation of a significant number of cores on a single die, minimizing the L1 caches in each core helps to incorporate more cores in one chip. Therefore, we conduct experiments for different sizes of L1 data caches, ranging from quite large to small. The experimental results show that compared to a conventional L1 data cache, the non-unified design reduces data cache dynamic energy consumption, on average, by up to 22% for a small data cache (such as a 4KB cache), and by up to 82% for a relatively large (such as a 32KB) cache.

Although optimization algorithms have made significant progress in a wide range of power system applications, many research areas and case studies remain unexplored and require additional investigation. Therefore, this paper tackles the coordination process of overcurrent relays (OCRs) in large-scale and complicated networks comprising hundreds of decision variables and constraints. In this context, savannah bengal tiger optimizer (SBTO) is employed to solve the OCR’s coordination of the 42-bus and 30-bus power grids by optimizing pickup current, time dial, and operating curve. It is worth noting that the 42-bus network comprises 291 decision variables and 713 constraints while the 30-bus incorporates 333 decision variables and 1119 constraints: indicating challenging problem formulation to be optimized. Thus, high-performance computing is necessary to provide the necessary computational efficiency and real-time results since the suggested method requires parallel processing capabilities for large-scale simulations and high-resolution models. The objective function (OF) is adapted to have a minimal total operating time (TOT) along with fulfilling problem constraints. Exceptional results of the utilized SBTO are validated against other benchmark algorithms such as genetic algorithm, particle swarm optimizer, water cycle algorithm, and gray wolf optimizer. Obviously, SBTO effectively meets the coordination criteria by achieving a minimal OF and incurring only a limited number of constraint violations. Furthermore, SBTO attains the minimum possible TOT of 19.66 s among other competitors for the 42-bus study case. This approach highlights SBTO’s capability in effectively addressing highly complex optimization problems with numerous constraints.

Combinatorial Scientific Computing explores the latest research on creating algorithms and software tools to solve key combinatorial problems on large-scale high-performance computing architectures. It includes contributions from international researchers who are pioneers in designing software and applications for high-performance computing systems. The book offers a state-of-the-art overview of the latest research, tool development, and applications. It focuses on load balancing and parallelization on high-performance computers, large-scale optimization, algorithmic differentiation of numerical simulation code, sparse matrix software tools, and combinatorial challenges and applications in large-scale social networks. The authors unify these seemingly disparate areas through a common set of abstractions and algorithms based on combinatorics, graphs, and hypergraphs. Combinatorial algorithms have long played a crucial enabling role in scientific and engineering computations and their importance continues to grow with the demands of new applications and advanced architectures. By addressing current challenges in the field, this volume sets the stage for the accelerated development and deployment of fundamental enabling technologies in high-performance scientific computing.

Harmonizing model with transfer tax on water pollution across regional boundaries in a China’s lake basin

Replication technology is commonly used to improve data availability and reduce data access latency in the cloud storage system by providing users with different replicas of the same service. Most current approaches largely focus on system performance improvement, neglecting management cost in deciding replicas number and their store places, which cause great financial burden for cloud users because the cost for replicas storage and consistency maintenance may lead to high overhead with the number of new replicas increased in a pay-as-you-go paradigm. In this paper, towards achieving the approximate minimum data sets management cost benchmark in a practical manner, we propose a replicas placements strategy from cost-effective view with the premise that system performance meets requirements. Firstly, we design data sets management cost models, including storage cost and transfer cost. Secondly, we use the access frequency and the average response time to decide which data set should be replicated. Then, the method of calculating replicas’ number and their store places with minimum management cost is proposed based on location problem graph. Both the theoretical analysis and simulations have shown that the proposed strategy offers the benefits of lower management cost with fewer replicas.