Han, Runzhou, Zhang, Duo, Zheng, Mai · Zenodo (CERN European Organization for Nuclear Research) · 2020
This is the prototype of PCheck, a type-aware fault injection framework to fingerprint PFS checker policies. This repository contains the necessary code to reproduce the experimental results mentioned in our paper "Fingerprinting the Checker Policies of Parallel File Systems" (SC PDSW'20). For more details of the usage of each component in the repo, please refer to the readme.md under the root directory of the repo.
Cited by 0Open access
Sanjay Khanna, David Kotz · 1997
We describe the effects of a new user-level library for the Galley Parallel File System. This library allows some pre-existing sequential programs to make use of the Galley Parallel File System with minimal modification. It permits programs to efficiently use the parallel file system because the user-level library groups accesses together. We examine the performance of our library, and we show how code needs to be modified to use the library. 1 Introduction Many parallel applications are limited by the performance of the I/O system, and the performance of many I/O systems is currently limited by the file system. The Galley Parallel File System [NK97] has demonstrated that it can provide parallel applications with high-throughput access to their data files, if they use new file-system interfaces. Unfortunately, it is sometimes inconvenient for programmers to rewrite their application code to fit the new interface. In this paper, we describe a new user-level library that runs on top of ...
Cited by 0
William M. Jones · 2001
This document provides a survey of literature ranging from parallel file systems to system modeling and simulation. This background research is being conducted in order facilitate an investigation into the plausibility of modeling the Parallel Virtual File System (PVFS) and the applications that run on it. The goal is to determine what kind of work has been done is this area, while discovering potential dissertation-level
Cited by 0
David Kotz, Carla Ellis · Journal of Parallel and Distributed Computing · 1993
Cited by 28
Peter Corbett, Dror G. Feitelson, Jean-Pierre Prost et al. · IBM Systems Journal · 1995
Parallel computer architectures require innovative software solutions to utilize their capabilities. This statement is true for system software no less than for application programs. File system development for the IBM SP product line of computers started with the Vesta research project, which introduced the ideas of parallel access to partitioned files. This technology was then integrated with a conventional Advanced Interactive Executive™ (AIX™) environment to create the IBM AIX Parallel I/O File System product. We describe the design and implementation of Vesta, including user interfaces and enhancements to the control environment needed to run the system. Changes to the basic design that were made as part of the AIX Parallel I/O File System are identified and justified.
Cited by 71
Avery Ching, Alok Choudhary, Wei-keng Liao et al. · 2003
Parallel scientific applications store and retrieve very large, structured datasets. Directly supporting these structured accesses is an important step in providing high-performance I/O solutions for these applications. High-level interfaces such as HDF5 and Parallel netCDF provide convenient APIs for accessing structured datasets, and the MPI-IO interface also supports efficient access to structured data. However, parallel file systems do not traditionally support such access. In this work we present an implementation of structured data access support in the context of the Parallel Virtual File System (PVFS). We call this support “datatype I/O” because of its similarity to MPI datatypes. This support is built by using a reusable datatype-processing component from the MPICH2 MPI implementation. We describe how this component is leveraged to efficiently process structured data representations resulting from MPI-IO operations. We quantitatively assess the solution using three test applications. We also point to further optimizations in the processing path that could be leveraged for even more efficient operation. 1.
Cited by 0
Scherson, Isaac D., Reis, Veronica L. M., Chen, Fan · eScholarship (California Digital Library) · 1996
This paper addresses the problem of providing a parallel virtual memory system with an efficient swap file structure. It thas been observed, experimentally, that the way data is organized in a parallel I/O environment greatly influences the I/O performance [9,6]. Thereforem if virtual memory is to be implemented in parallel machienes, its disk swap space organization will be one of the key factors determining the virtual memory stystem's efficiency.We analyze different parallel file organiations in order to determine which one better suits the needs of a parallel virtual memory system.Simulation results collected thus far indicate that the best strategy is to assign to each I/O node data that will be utilized mostly by neighbor processors. We observed that this approach better load balances I/O requests among I/O nodes and better explores inter-processor locality.
Cited by 0
Isaac Scherson Luis, Luis Miguel Campos, Veronica L. M. Reis et al. · 1997
This paper addresses the problem of providing a parallel virtual memory system with an efficient swap file structure. It has been observed, experimentally, that the way data is organized in a parallel I/O environment greatly influences the I/O performance. Therefore, if virtual memory is to be implemented in parallel machines, its disk swap space organization will be one of the key factors determining the virtual memory system's efficiency. We analyze different parallel file organizations in order to determine which one better suits the needs of a parallel virtual memory system. Simulation results collected thus far indicate that the best strategy is to assign to each I/O node data that will be utilized mostly by neighbor processors. We observed that this approach better load balances I/O requests among I/O nodes and better explores inter-processor locality. 1 Introduction Modern parallel machines, such as the CRAY T3D, do not provide virtual memory: it is the programmer's responsib...
Cited by 0
Kasick, Michael P. · KiltHub Repository · 2015
Parallel file systems target large, high-performance storage systems. Since these storage systems<br>are comprised of a significant number of components (i.e., hundreds of file servers, thousands of<br>disks, etc.), they are expected to (and in practice do) frequently exhibit “problems”, from degraded<br>performance to outright failure of one or more components. The sheer number of components,<br>and thus, potential problems, makes manual diagnosis of these problems difficult. Of particular<br>concern are system-wide performance degradations, which may arise from a single misbehaving<br>component, and thus, pose a challenge for problem localization. Even failure of a redundant component<br>with a less-significant performance impact is worrisome as it may, in absence of explicit<br>checks, go unnoticed for some time and increase risk of system unavailability.<br>As a solution, this thesis defines a novel problem-diagnosis approach, capitalizing upon the<br>parallel-file-system design criterion of balanced performance, that peer-compares the performance<br>of system components to localize problems within storage systems running unmodified, “off-theshelf”<br>parallel file systems. Performed in support of this thesis is a set of laboratory experiments<br>that demonstrate proof-of-concept of the peer-comparison approach by injecting four realistic<br>problems into 12-server, test-bench PVFS and Lustre clusters. This thesis is further validated<br>by taking the diagnosis approach and adapting it to to work on a very-large, production GPFS storage<br>system consisting of 128 file servers, 32 storage controllers, 1152 disk arrays, and 11,520 total<br>disks. Presented in a 15-month case study is the problems observed through analysis of 624GB of<br>instrumentation data, in which a variety of performance-related storage-system problems are localized<br>and diagnosed, in a matter of hours, as compared to days or longer with manual approaches.
Cited by 0Open access
G. Paolillo · Università degli Studi di Napoli Federico II · 2007
Cited by 0Open access
Han, Runzhou, Zhang, Duo, Zheng, Mai · Zenodo (CERN European Organization for Nuclear Research) · 2020
This is the prototype of PCheck, a type-aware fault injection framework to fingerprint PFS checker policies. This repository contains the necessary code to reproduce the experimental results mentioned in our paper "Fingerprinting the Checker Policies of Parallel File Systems" (SC PDSW'20). For more details of the usage of each component in the repo, please refer to the readme.md under the root directory of the repo.
Cited by 0Open access
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
松宮遼, 佐々木慎, 高橋一志 et al. · Medical Entomology and Zoology · 2014
Cited by 0Open access
William Loewe, Richard Hedges, Tyce T. McLarty et al. · University of North Texas Digital Library (University of North Texas) · 2004
Livermore Computing is an early and aggressive adopter of parallel file systems including, for example, GPFS from IBM and Lustre for our present Linux systems. As such, we have acquired more than our share of battle scars from encountering bugs in 'bleeding edge' file systems that we have pressed into production to serve our customers' massive I/O requirements. A major role of the Scalable I/O Project is to detect errors before our end users do. In order to do this, we have developed highly parallel test codes to stress and probe potentially weak areas of file system behavior. This paper describes those test programs and how we make use of them.
Cited by 4Open access
Peng Gu, Jun Wang, Robert Ross · 2008
Parallel I/O plays an increasingly important role in to-day’s data intensive computing applications. While much attention has been paid to parallel read performance, most of this work has focused on the parallel file system, mid-dleware, or application layers, ignoring the potential for improvement through more effective use of local storage. In this paper, we present the design and implementation of Segment-structured On-disk data Grouping and Prefetching (SOGP), a technique that leverages additional local stor-age to boost the local data read performance for parallel file systems, especially for those applications with partially overlapped access patterns. Parallel Virtual File System (PVFS) is chosen as an example. Our experiments show that an SOGP-enhanced PVFS prototype system can outperform a traditional Linux-Ext3-based PVFS for many applications and benchmarks, in some tests by as much as 230 % in terms of I/O bandwidth. 1
Cited by 0
Randal E. Bryant · 2011
Both GFS &amp; HDFS are designed with similar goals: * High throughput (latency less important)- Designed for batch processing jobs, e.g., MapReduce * High capacity (large block size). Typical files&gt; 1GB * High scalability. Handle&gt; 10^7 files * Reliability through replication. Treat failure as normal. Although GFS was developed first, HDFS is much simpler, and so will describe it first. HDFS Based on &quot;write once, read many &quot; model: * Each file has single writer * File fully written and closed before any reader given access Three components: * Clients. * NameNode. Single node containing all metadata about all files * DataNodes. Set of nodes that store actual file contents. * File represented as sequence of blocks of fixed size (64 or 128 MB). (Given byte offset for read, can immediately determine which block to read.) * Each block has unique block ID. * Blocks are distributed across multiple DataNodes to enable parallel access. * Blocks replicated (default 3X) to enable recovery when DataNode fails. * When block created, NameNode decides placements- Default: two within single rack, third on a different rack- Access time / safety tradeoff. NameNode * Metadata: Information about file + plus set of block IDs + location of all replicas of all blocks * Treats each FS operation as transaction- Maintains all information in memory- Logs to EditLog file- (Possibly outdated) backup copy stored on disk.- Can bring this copy up to date by replaying entries in EditLog * Periodic checkpoint:- Apply transactions in editlog to disk image- Delete old parts of editlog- Can do in background while other updates occuring.- Does not store locations of replicas * Tries to satisfy given read request with nearby DataNode
Cited by 0
Ananth Devulapalli, Dennis Dalessandro, Pete Wyckoff et al. · 2007
As storage systems evolve, the block-based design of today's disks is becoming inadequate. As an alternative, object-based storage devices (OSDs) offer a view where the disk manages data layout and keeps track of various attributes about data objects. By moving functionality that is traditionally the responsibility of the host OS to the disk, it is possible to improve overall performance and simplify management of a storage system. The capabilities of OSDs will also permit performance improvements in parallel file systems, such as further decoupling metadata operations and thus reducing metadata server bottlenecks.
Cited by 34
Toni Cortés · 1997
Cited by 4Open access
International Journal of Advance Engineering and Research Development · 2016
We study the problem of key establishment for secure many-to-many communications. The problem is inspired by the proliferation of large-scale distributed file systems supporting parallel access to multiple storage devices. Our work focuses on the current Internet standard for such file systems, i.e. , parallel Network File System (pNFS), which makes use of Kerberos to establish parallel session keys between clients and storage devices. Our review of the existing Kerberos-based protocol shows that it has a number of limitations: (i) a metadata server facilitating key exchange between the clients and the storage devices has heavy workload that restricts the scalability of the protocol; (ii) the protocol does not provide forward secrecy; (iii) the metadata server generates itself all the session keys that are used between the clients and storage devices, and this inherently leads to key escrow. In this paper, we propose a variety of authenticated key exchange protocols that are designed to address the above issues. We show that our protocols are capable of reducing up to approximately 54% of the workload of the metadata server and concurrently supporting forward secrecy and escrow-freeness. All this requires only a small fraction of increased computation overhead at the client.
Cited by 1Open access
Yang Wu · TigerPrints (Clemson University) · 2009
One of the challenges that the design of parallel file system for HPC(High Performance Computing) has to face today is maintaining the scalability to handle the I/O generated by parallel applications that involve accessing directories containing a large number of entries and performing hundreds of thousands of operations per second. Currently, highly concurrent access to large directories is poorly supported in parallel file systems. As a result, it is important to build a scalable directory service for parallel file systems to support efficient concurrent access to larger directories. In this thesis we demonstrate a scalable directory service designed for parallel file systems(specifically for PVFS) that can achieve high throughput and scalability while minimizing bottlenecks and synchronization overheads. We describe important concepts and goals in scalable directory service design and its implementation in the parallel file system simulator--HECIOS. We also explore the simulation model of MPI programs and the PVFS file system in HECIOS, including the method to verify and validate it. Finally, we test our scalable directory service on HECIOS and analyze the performance and scalability based on the results. In summary, we demonstrate that our scalable directory service can effectively handle highly concurrent access to large directories in parallel file systems. We are also able to show that our scalable directory service scales well with the number of I/O nodes in the cluster.
Cited by 3Open access