Thomas Duessel, N. Eicker, Florin Isailă et al. · ArXiv.org · 2003
Today's cluster computers suffer from slow I/O, which slows down I/O-intensive applications. We show that fast disk I/O can be achieved by operating a parallel file system over fast networks such as Myrinet or Gigabit Ethernet. In this paper, we demonstrate how the ParaStation3 communication system helps speed-up the performance of parallel I/O on clusters using the open source parallel virtual file system (PVFS) as testbed and production system. We will describe the set-up of PVFS on the Alpha-Linux-Cluster-Engine (ALiCE) located at Wuppertal University, Germany. Benchmarks on ALiCE achieve write-performances of up to 1 GB/s from a 32-processor compute-partition to a 32-processor PVFS I/O-partition, outperforming known benchmark results for PVFS on the same network by more than a factor of 2. Read-performance from buffer-cache reaches up to 2.2 GB/s. Our benchmarks are giant, I/O-intensive eigenmode problems from lattice quantum chromodynamics, demonstrating stability and performance of PVFS over Parastation in large-scale production runs.
Cited by 0Open access
N. Eicker, Florin Isailă, Thomas Lippert et al. · Advances in parallel computing · 2004
Cited by 0
Rajeev Thakur, Rajesh Bordawekar, Alok Choudhary et al. · 2002
We are developing a compiler and runtime support system called PASSION: Parallel and Scalable Software for Input-Output. PASSION provides software support for I/O intensive out-of-core loosely synchronous problems. This paper gives an overview of the PASSION Runtime Library and describes two of the optimizations incorporated in it, namely data prefetching and data sieving. Performance improvements provided by these optimizations on the Intel Touchstone Delta are discussed together with an out-of-core median filtering application.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Cited by 51
David Kotz, Ravi Jain · 1998
One is scientific computing with massive datasets, such as those found in seismic processing, climate modeling, and so forth [dC94]. The second is databases [DG92]. The I/O bottleneck continues to be a serious concern for scientific computing, particularly Grand Challenge problems, where it is now commonly recognized as an obstacle. Many scientific applications generate 1 GB of I/O per run [dC94], and applications performing an order of magnitude more are not uncommon: applications in computational physics and fluid dynamics are projected to require I/O on the order of 1 TB [dC94]. It seems clear that these total I/O requirements will keep increasing as scientists continue to study phenomena at larger space and time scales, and at finer space and time resolutions. Since the response time that humans can tolerate for obtaining computational results--- no matter how comprehensive and detailed--- is always bounded, the I/O rates required will continue to increase also. Thus while curre
Cited by 8
Vo, Johnathan; Brill, Steven Randall · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 2019
Abstract not provided.
Cited by 0Open access
Rui Yang, Marshall L. Ward, Ben Evans · 2019
Abstract. We present an implementation of parallel I/O in the Modular Ocean Model (MOM), a numerical ocean model used for climate forecasting, and determine its optimal performance over a range of tuning parameters. Our implementation uses the parallel API of the netCDF library, and we investigate the potential bottlenecks associated with the model configuration, netCDF implementation, the underpinning MPI-IO library/implementations and Lustre filesystem. We investigate the performance of a global 0.25° resolution model using 240 and 960 CPUs. The best performance is observed when we limit the number of contiguous I/O domains on each compute node and assign one MPI rank to aggregate and write the data from each node, while ensuring that all nodes participate in writing this data to our Lustre filesystem. These best performance configurations are then applied to a higher 0.1° resolution global model using 720 and 1440 CPUs, where we observe even greater performance improvements. In all cases, the tuned parallel I/O implementation achieves much faster write speeds relative to serial single-file I/O, with write speeds up to 60 times faster at higher resolutions. Under the constraints outlined above, we observe that the performance scales as the number of compute nodes and I/O aggregators are increased, ensuring the continued scalability of I/O-intensive MOM5 model runs that will be used in our next generation higher resolution simulations.
Cited by 2Open access
C. Eric Wu · 1995
In recent years, the design and performance evaluation of parallel processors has focused on the processor, memory and communication subsystems. As a result, these subsystems have better performance potential than the I/O subsystem. In fact, the I/O subsystem is the bottleneck in many machines. However, there are a number of studies currently underway to improve the design of parallel I/O subsystems. To develop optimal parallel I/O subsystem designs, one must have a thorough understanding of the workload characteristics of parallel I/O and its exploitation of the associated parallel file system. Presented are the results of a study conducted to analyze the parallel I/O workloads of several applications on a parallel processor using the Vesta parallel file system. Traces of the applications are obtained to collect system events, communication events, and parallel I/O events. The traces are then analyzed to determine workload characteristics. The results show I/O request rates on the order of hundreds of requests per second, a large majority of requests are for small amounts of data (less than 1500 bytes), a few requests are for large amounts of data (on the order of megabytes), significant file sharing among processes within a job, and strong temporal, traditional spatial, and interprocess spatial locality. I.
Cited by 0
Sebastian Lührs · JuSER (Forschungszentrum Jülich) · 2022
Numerical simulations conducted on current high-performance computing (HPC) systems face an ever growing need for scalability. Larger HPC platforms provide opportunities to push the limitations on size and properties of what can be accurately simulated. Therefore, it is needed to process larger data sets, be it reading input data or writing results. Serial approaches on handling I/O in a parallel application will dominate the performance on massively parallel systems, leaving a lot of computing resources idle during those serial application phases. In addition to the need for parallel I/O, input and output data is often processed on different platforms. Heterogeneity of platforms can impose a high level of maintenance, when different data representations are needed. Portable, selfdescribing data formats such as HDF5 and netCDF are examples of already widely used data formats within certain communities. This course will start with an introduction of different parallel I/O strategies. The course will introduce the use of the HDF5, the NetCDF and the SIONlib library interfaces. Optimization potential and best practices are discussed.
Cited by 1Open access
Orran Krieger, Karen Reid, Michael Stumm · 1995
Harnessing the full I/O capabilities of a large-scale multiprocessor is difficult and requires a great deal of cooperation between the application programmer, the compiler and the operating (/file) system. Hence, the parallel I/O interface used by the application to communicate with the system is crucial in achieving good performance. We present a set of properties we believe that a good I/O interface should have and consider current parallel I/O interfaces from the perspective of these properties. We describe the advantages and disadvantages of mapped-file I/O and argue that if properly implemented it can be a good basis for a parallel I/O interface that can fulfill the suggested properties. To demonstrate that such an implementation is feasible, we describe methodology used in our previous work on the Hurricane operating system and in our current work on the Tornado operating system to implement mapped files. 1 Introduction Harnessing the full I/O capabilities of a large-scale share...
Cited by 1
Philip Hatcher, Philip J. Hatcher · 1994
High-level programming languages, including C* and High-Performance Fortran, do not currently have constructs for parallel i/o. Parallel i/o, if done at all, is performed via vendor specific facilities. We present a brief design for extensions to the data-parallel programming model to support parallel i/o. Our planned extensions are based on the notion of a data-parallel file, which encapsulates a set of i/o streams in a single abstraction. Language constructs supporting this extended programming model will be developed and implemented in the UNH C* compiler. 1 Data-Parallel I/O Primitives Our planned extensions are based on the notion of a data-parallel file, which encapsulates a set of i/o streams in a single abstraction. Each virtual processor provided by the underlying data-parallel programming language has its own file pointer. Each input operation reads a set of values and each output operation writes a set of values. A file on a storage device will consist of both data and cont...
Cited by 0
Dror Feitelson Ibm, Dror G. Feitelson · 1994
To be useful, terminal I/O on massively parallel MIMD machines must be able to differentiate between the I/O streams from different tasks. This is done in the Vulcan terminal I/O facility by providing a special control panel, which allows an independent window to be opened for each task. The controls look like LEDs, being color coded to indicate status (e.g. output is available or the task is waiting for input). Additional LEDs are provided as a new form of output, allowing the application to report status visually rather than by using text output. This is useful during program development and debugging. A derivative of the devices reported here has been incorporated in the AIX Parallel Environment on the new IBM SP1 multicomputer. 1
Cited by 0
Chuck Mosher · Proceedings · 2012
JavaSeis is an Open Source project that provides infrastructure for parallel I/O and computation for seismic processing and imaging applications using the Java programming language. JavaSeis is not a seismic processing system, but rather a set of object-oriented classes and methods for building geophysical software applications. To date, the most extensive usage of JavaSeis has been with the SeisSpace seismic processing system from Landmark Graphics Corporation, a subsidiary of Halliburton. The JavaSeis project is hosted by SourceForge, a web-based repository for open source projects. JavaSeis is based on a long history of parallel computing algorithms, dating from the late 1970’s, when parallel computing first began to emerge as an important component of High Performance Computing (HPC). This history provides many of the unique features of JavaSeis, where a significant number of optimized parallel computing algorithms now reside. These features have been selected and tested through repeated application of a wide variety of parallel computing ideas to production processing and imaging problems in the geophysical industry. This history also has allowed the construction of a parallel computing model that performs very well on both shared memory and distributed memory computing hardware, allowing users to construct algorithms without tying their application to a specific hardware platform.
Cited by 0
Ying Bai · 2016
Cited by 0
Suren Byna, Yong Chen, Xian‐He Sun et al. · 2008
Parallel I/O prefetching is considered to be effective in improving I/O performance. However, the effectiveness depends on determining patterns among future I/O accesses swiftly and fetching data in time, which is difficult to achieve in general. In this study, we propose an I/O signature-based prefetching strategy. The idea is to use a predetermined I/O signature of an application to guide prefetching. To put this idea to work, we first derived a classification of patterns and introduced a simple and effective signature notation to represent patterns. We then developed a toolkit to trace and generate I/O signatures automatically. Finally, we designed and implemented a thread-based client-side collective prefetching cache layer for MPI-IO library to support prefetching. A prefetching thread reads I/O signatures of an application and adjusts them by observing I/O accesses at runtime. Experimental results show that the proposed prefetching method improves I/O performance significantly for applications with complex patterns.
Cited by 100
Jing Fu, Ning Liu, Onkar Sahni et al. · 2010
With the development of high-performance computing, I/O issues have become the bottleneck for many massively parallel applications. This paper investigates scalable parallel I/O alternatives for massively parallel partitioned solver systems. Typically such systems have synchronized ¿loops¿ and will write data in a well defined block I/O format consisting of a header and data portion. Our target use for such an parallel I/O subsystem is checkpoint-restart where writing is by far the most common operation and reading typically only happens during either initialization or during a restart operation because of a system failure. We compare four parallel I/O strategies: 1 POSIX File Per Processor (1PFPP), a synchronized parallel IO library (syncIO), ¿Poor-Man's¿ Parallel I/O (PMPIO) and a new ¿reduced blocking¿ strategy (rbIO). Performance tests using real CFD solver data from PHASTA (an unstructured grid finite element Navier-Stokes solver) show that the syncIO strategy can achieve a read bandwidth of 6.6GB/Sec on Blue Gene/L using 16K processors which is significantly faster than 1PFPP or PMPIO approaches. The serial ¿token-passing¿ approach of PMPIO yields a 900 MB/sec write bandwidth on 16K processors using 1024 files and 1PFPP achieves 600 MB/sec on 8K processors while the ¿reduced-blocked¿ rbIO strategy achieves an actual writing performance of 2.3GB/sec and perceived/latency hiding writing performance of more than 21,000 GB/sec (i.e., 21TB/sec) on a 32,768 processor Blue Gene/L.
Cited by 18
Zanhua Huang, Kaiyuan Hou, Ankit Agrawal et al. · 2023
Large-scale parallel applications can face significant I/O performance bottlenecks, making efficient I/O crucial. This work presents a comparative study of several parallel I/O implementations in the Weather Research and Forecasting model, including PnetCDF blocking and non-blocking I/O options, netCDF4, HDF5 Log VOL, and ADIOS. For I/O methods creating files in a canonical data layout, PnetCDF's non-blocking option offers up to 2x improvement over its blocking option and up to 4.5x over HDF5 via netCDF4, demonstrating the effectiveness of the write request aggregation technique. The HDF5 Log VOL outperforms ADIOS with a 4x improvement in write performance when creating files in the log layout, although both require non-negligible time to convert the file back to canonical order for post-run analysis. From these results we extract some observations that can guide I/O strategies for modern parallel codes.
Cited by 4
Michael Kluge, Andreas Knüpfer, Matthias Müller et al. · Lecture notes in computer science · 2009
Cited by 5
Christian M. Chilan, Albert Cheng · 2012
The performance of a parallel application is heavily affected by the I/O access pattern, file system type, and MPI communication modes. A determination of how these elements interact offers an explanation for the performance behavior, and provides hints for improving the execution of a particular pattern.
Cited by 0
Santiago Domínguez-Domínguez, Jorge Buenabad-Chávez · 2005
Local area networks are now widely used to run parallel applications. They are particularly suitable to run I/O intensive applications, because each node usually includes disk space. However, programming these applications is rather difficult. The programmer must partition data into disk nudes and, at run time, transfer data from disk space into the memory of each node that uses the data and vice versa. Also, such partitioning gives data a fixed location in disk space, and is usually not adequate for performance because processors do not access mostly data in their local memory or disk, but on disk space in remote nodes. This paper presents a distributed parallel file system that both eases the programming and improves the performance of parallel I/O intensive applications. Our file system eases programming through mapping into memory files of size up to hundreds of Giga bytes. It improves performance through automatically diffusing, or migrating and replicating, data in files to the local memory or local disk of the processors that use the data. Data diffusion occurs under a multiple-readers-single-writer protocol. On applications tested the performance gain can be up to 20 % compared to versions using the MPI file system.
Cited by 0
Virginia Lo, Jens Mache · 1999
A serious bottleneck to high-performance parallel computing is the high cost of data transfer. As communication and I/O traffic on the links in the interconnection network increases, network contention becomes a critical problem, drastically reducing effective data throughput. Minimizing network contention is crucial in order to achieve fast job response times and high system throughput.
One factor that affects network contention is the resource management issue of processor allocation, the assignment of a set of processors to each scheduled job. Previous processor allocation strategies have essentially ignored the I/O and communication demands of parallel applications and the resulting network contention. Our analysis shows that the spatial layout of the compute nodes and the I/O nodes in relation to each other within the interconnection network topology is a key factor that affects network contention. Based on the results of this analysis, we design and test new processor allocation strategies that minimize network contention by being sensitive to spatial layout and its effect on communication and parallel I/O.
Our analysis is based on analytic modeling and on simulations driven by synthetic workloads and realistic workload traces captured at supercomputing sites. We analyze and minimize network contention in three different situations. First, we concentrate on communication intensive jobs. We analyze inter-job link contention due to communication among compute nodes, and we design a new strategy that allocates each job as compactly as possible. Second, we concentrate on parallel I/O intensive jobs. We analyze traffic hotspots due to data transfer between compute nodes and I/O nodes, and we design new strategies that optimize the shape and location of jobs relative to the I/O nodes. Strategies that are optimal for parallel I/O often conflict with strategies that are optimal for communication. Thus as a final step, we design an integrated allocation strategy that accommodates workloads that are both communication intensive and parallel I/O intensive. Our new strategies are successful in improving both average job response times and system throughput, and thus make a contribution towards efficient resource management of teraflops-scale computing systems.
Cited by 2