Taming aggressive replication in the Pangaea wide-area file system

Modern file systems associate the deletion of a file with the immediate release of storage, and file writes with the irrevocable change of file contents. We argue that this behavior is a relic of the past, when disk storage was a scarce resource. Today, large cheap disks make it possible for the file system to protect valuable data from accidental delete or overwrite.This paper describes the design, implementation, and performance of the Elephant file system, which automatically retains all important versions of user files. Users name previous file versions by combining a traditional pathname with a time when the desired version of a file or directory existed. Storage in Elephant is managed by the system using file-grain user-specified retention policies. This approach contrasts with checkpointing file systems such as Plan-9, AFS, and WAFL that periodically generate efficient checkpoints of entire file systems and thus restrict retention to be guided by a single policy for all files within that file system.Elephant is implemented as a new Virtual File System in the FreeBSD kernel.

Despite continual improvements in the performance and reliability of large scale file systems, the management of user-defined file system metadata has changed little in the past decade. The mismatch between the size and complexity of large scale data stores and their ability to organize and query their metadata has led to a de facto standard in which raw data is stored in traditional file systems, while related, application-specific metadata is stored in relational databases. This separation of data and semantic metadata requires considerable effort to maintain consistency and can result in complex, slow, and inflexible system operation. To address these problems, we have developed the Quasar File System (QFS), a metadata-rich file system in which files, user-defined attributes, and file relationships are all first class objects. In contrast to hierarchical file systems and relational databases, QFS defines a graph data model composed of files and their relationships. QFS incorporates Quasar, an XPATH-extended query language for searching the file system. Results from our QFS prototype show the effectiveness of this approach. Compared to the de facto standard, the QFS prototype shows superior ingest performance and comparable query performance on user metadata-intensive operations and superior performance on normal file metadata operations.

This paper presents a design of effective cryptographic file system (ECFS) for storage services. ECFS can be used to provide the data security in existing network and p2p file system such as NFS, CIFS, OceanStore and Yahoo! Briefcase. We assume that the storage system isn't trusted entirely and end user must be able to control some security functions. Adding security to storage systems leads to performance degradation. In order to achieve good performance, ECFS uses a key management using symmetric cipher instead of public-key approach. ECFS also uses two-phases revocation to decrease the overhead of key revocation. We evaluate the ECFS security based on the set of criteria for evaluating a storage solution.

Careful simulation‐based evaluation plays an important role in the design of file and disk systems. We describe here a particular approach to such evaluations that combines techniques in workload synthesis, file system modeling, and detailed disk behavior modeling. Together, these make feasible the detailed simulation of I/O hardware and file system software. In particular, using the techniques described here is likely to make comparative file system studies more accurate.In addition to these specific contributions, the paper makes two broader points. First, it argues that detailed models are appropriate and necessary in many cases. Second, it demonstrates that detailed models need not be difficult or time consuming to construct or execute.

As bandwidth constraints on LAN/WAN environments decrease, the demand for distributed services will continue to increase. In particular, the proliferation of user-level applications requiring high-capacity distributed file storage systems will demand that such services be universally available. At the same time, the advent of high-speed networks have made the deployment of application and communication solutions based upon an Intelligent Mobile Agent (IMA) framework practical. Agents have proven to present an ideal development paradigm for the creation of autonomous large-scale distributed systems, and an agent-based communication scheme would facilitate the creation of independently administered distributed file services. This thesis thus outlines an architecture for such a distributed file system based upon an IMA communication framework.

As one of the workload characteristics, the anomaly behaviorsin real workload have been recognized as a critical requirement for file system design. In this paper, a set of traces collected from typically realistic file system have been analyzed. Studying the correlations of I/O request inter-arrival times shows that it is necessary to examine the self-similarity in file system workload. And then the phenomenon of self-similarity which had been initially observed in network and disk I/O workload has also been visually and statistically observed in file system workload used in this paper. Furthermore, this property is further explored and we find that the self-similarity of I/O accesses in those sets of traces studied in this paper is mainly due to reads. This study is extremely important to provide useful insight into the design and analysis of file system for performance gains, as well as to model file system I/O workloads.

The ARINC-653 standard, dictating the use of time and space partitioning in real-time avionics systems, has been widely embraced by the industry. We present the design of real-time file system (RTFS), a file system that complies with the ARINC-653 standard and the emerging standard for a file system for ARINC-653. RTFS provides realtime accesses to data stored on a variety of mass storage device. In addition to an interface complying with emerging standard, RTFS provides an application interface that complies with a subset of the POSIX standard. RTFS is implemented as a normal ARINC-653 partition. Task partitions communicate their file operation requests to RTFS via queuing ports; such ports are also used to deliver the responses from RTFS to the task partitions. The temporal behavior of RTFS is predictable and the response times for file operations are bounded. The design of RTFS handles a mix of hard and soft real-time file access requests. RTFS implements metadata journaling using on-board non-volatile memory devices to provide fast file updates and fast file system recovery on faults. Finally, RTFS includes facilities to support network-centric operations and a set of design and maintenance tools. This paper overviews the design of RTFS and describes the realization of the many unique features of RTFS.

This paper proposes a backup system based on mirroring filesystem GMFS has been developed to mirror data in realtime on the filesystem layer. The GMFS is a stackable filesystem which flexibly mirrors without changing the existing environment by operating as a wrapper of other filesystems.Because the conventional mirroring technology utilizes the mirroring function on the device layer or needs a special filesystem, the allocation of the disk and the specific format of the filesystem are needed, and so the disk design is fixed. Therefore, the conventional mirroring technology cannot adjust when the mirroring function not assumed will be needed later. In this situation, the mechanism that adds the mirroring function without changing the existing disk design is necessary. The GMFS conducts the operation of other filesystems transparently, thereby users need not be aware of the GMFS. The conventional filesystem looks as if it performs mirroring data by itself. GMFS can therefore add the function that mirrors in realtime without destroying the existing environment.GMFS uses NFS which is a typical network file system to communicate with an existing environment. The throughput of reading and writing has been improved by adopting the method to call system function of NFS from the inside of the filesystem.We developed this filesystem, and evaluated the performance from the viewpoint of throughput and system call speed and CPU loads. As a result, it was shown that there was no problem in the viewpoint of the performance compared with the conventional filesystem, and the throughput of the read and write of GMFS was 2.0 times faster than conventional mirroring filesystem.

We present the design, implementation, and evaluation of Direct File System (DFS) for virtualized flash storage. Instead of using traditional layers of abstraction, our layers of abstraction are designed for directly accessing flash memory devices. DFS has two main novel features. First, it lays out its files directly in a very large virtual storage address space provided by FusionIO's virtual flash storage layer. Second, it leverages the virtual flash storage layer to perform block allocations and atomic updates. As a result, DFS performs better and is much simpler than a traditional Unix file system with similar functionalities. Our microbenchmark results show that DFS can deliver 94,000 I/O operations per second (IOPS) for direct reads and 71,000 IOPS for direct writes with the virtualized flash storage layer on FusionIO's ioDrive. For direct access performance, DFS is consistently better than ext3 on the same platform, sometimes by 20%. For buffered access performance, DFS is also consistently better than ext3, and sometimes by over 149%. Our application benchmarks show that DFS outperforms ext3 by 7% to 250% while requiring less CPU power.

Vendors cannot provide all the operating-system services that users demand. So, there has been a persistent desire to make operating systems more flexible and customizable. Naturally, object-oriented technology has come to bear on this area. However the ease of use of many such solutions has been disappointing. Ease of use is a key feature of Frigate, an object-oriented file system. It differs from most file system designs in that it's for ordinary users, not sophisticated operating system gurus. Frigate's modular, extensible framework allows new extensions to be "plugged in" on the fly. It is fully implemented and supports a set of example file system extensions.

A global and parallel file system for grids

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).

We introduce Pastis, a completely decentralized multi-user read-write peer-to-peer file system. In Pastis every file is described by a modifiable inode-like structure which contains the addresses of the immutable blocks in which the file contents are stored. All data are stored using the Past distributed hash table (DHT), which we have modified in order to reduce the number of network messages it generates, thus optimizing replica retrieval. Pastis' design is simple compared to other existing systems, as it does not require complex algorithms like Byzantine-fault tolerant (BFT) replication or a central administrative authority. It is also highly scalable in terms of the number of network nodes and users sharing a given file or portion of the file system. Furthermore, Pastis takes advantage of the fault tolerance and good locality properties of its underlying storage layer, the Past DHT. We have developed a prototype based on the FreePastry open-source implementation of the Past DHT. We have used this prototype to evaluate several characteristics of our file system design. Supporting the close-to-open consistency model, plus a variant of the read-your-writes model, our prototype shows that Pastis is between 1.4 to 1.8 times slower than NFS. In comparison, Ivy and Oceanstore are between two to three times slower than NFS.

Embedded systems such as PVR, set-top box, HDTV put unique demand on I/O subsystem design. Underlying software, particularly file system, needs to be elaborately designed so that it can meet tight constraints of consumer electronics platform: performance, price, reliability, and etc. In this work, we develop state-of-art file system elaborately tailored for A/V workload. There are two design objectives in our file system: performance and support for logical level abstraction. For performance, we develop a number of novel features: extent based allocation, single level file structure with block index augmentation scheme, aggressive free block allocation to minimize disk fragmentation, elaborate file system meta data layout, separation of name space data and file data and etc. HERMES enables the user to view file as a collection of semantic units (frame or audio samples). HERMES file system encompasses most of state-of-the-art file system technologies published in preceding works. Via extensive physical experiment, we verify that HERMES file system successfully addresses the original issues: good scalability, predictable I/O latency (minimizing variability in I/O latency), efficient disk head movement and etc. This is the result of harmonious effort of large I/O size, aggressive free block allocation algorithm, data block placement strategy, file organization, layout of HERMES file system and etc. The result of performance experiments indicate that HERMES file system prototype successfully meets the file system constraints for high volume and high bandwidth multimedia application. HERMES file system exhibits superior performance to EXT2 file system (Linux) and XFS file system (SGI).

High-performance distributed computing appears to be shifting away from tightly-connected supercomputers to composed of heterogeneous systems of networks, computers, storage devices and various other devices that collectively act as a single geographically distributed virtual computer. One of the great challenges for this environment is providing efficient parallel data access to remote distributed data sets. In this paper, we discuss some of the issues associated with parallel I/O and computational grids and describe the design of a flexible parallel file system that allows the application to control the behavior and functionality of virtually all aspects of the file system.