Search

As leadership computing facilities grow their storage capacity into the multi- petabyte range, the number of files and directories leap into the scale of billions. A complete profiling of such a parallel file system in a production environment presents a unique challenge. On one hand, the time, resources, and negative performance impact on production users can make regular profiling difficult. On the other hand, the result of such profiling can yield much needed understanding of the file system's general characteristics, as well as provide insight to how users write and access their data on a grand scale. This paper presents a lightweight and scalable profiling solution that can efficiently walk, analyze, and profile multi-petabyte parallel file systems. This tool has been deployed and is in regular use on very large-scale production parallel file systems at both Oak Ridge National Lab's Oak Ridge Leadership Facility (OLCF) and Lawrence Livermore National Lab's Livermore Computing (LC) facilities. We present the results of our initial analysis on the data collected from these two large-scale production systems, organized into three use cases: (1) file system snapshot and composition, (2) striping pattern analysis for Lustre, and (3) simulated storage capacity utilization in preparation for future file systems. Our analysis shows that on the OLCF file system, over 96% of user files exhibit the default stripe width, potentially limiting performance on large files by underutilizing storage servers and disks. Our simulated block analysis quantitatively shows the space overhead when doing a forklift system migration. It also reveals that due to the difference in system compositions (OLCF vs. LC), we can achieve better performance and space trade-offs by employing different native file system block sizes.

Key-Value stores provide scalable metadata service for distributed file systems. However, the metadata's organization itself, which is organized using a directory tree structure, does not fit the key-value access pattern, thereby limiting the performance. To address this issue, we propose a distributed file system with a loosely-coupled metadata service, LocoFS, to bridge the performance gap between file system metadata and key-value stores. LocoFS is designed to decouple the dependencies between different kinds of metadata with two techniques. First, LocoFS decouples the directory content and structure, which organizes file and directory index nodes in a flat space while reversely indexing the directory entries. Second, it decouples the file metadata to further improve the key-value access performance. Evaluations show that LocoFS with eight nodes boosts the metadata throughput by 5 times, which approaches 93% throughput of a single-node key-value store, compared to 18% in the state-of-the-art IndexFS.

Data services such as search, discovery, and management in scalable distributed environments have traditionally been decoupled from the underlying file systems, and are often deployed using external databases and indexing services. However, modern data production rates, looming data movement costs, and the lack of metadata, entail revisiting the decoupled file system-data services design philosophy.In this paper, we present TagIt, a scalable data management service framework aimed at scientific datasets, which is tightly integrated into a shared-nothing distributed file system. A key feature of TagIt is a scalable, distributed metadata indexing framework, using which we implement a flexible tagging capability to support data discovery. The tags can also be associated with an active operator, for pre-processing, filtering, or automatic metadata extraction, which we seamlessly offload to file servers in a load-aware fashion. Our evaluation shows that TagIt can expedite data search by up to 10X over the extant decoupled approach.

A file can be called an atomic information element that is found all over in computer world just like dust particles in real world. Files are universally used, ranging from personal data on desktop computers, to organizational data shared over local networks and the data that exist on the web. That's why files caught the attention of ours and lead us to study the life of a file from Cradle to Grave or Eternity. In our study we investigate: the birth of a file; sources of file's birth; the tsunami of files; the good, the bad and the ugly files; file aging; file states transition; file lost and the quest of finding a file; a journey in past and future; the death of a file; file resurrection and the gods of clouds.

Navigating through a file hierarchy is one of the most common methods for accessing files, yet it can be slow and repetitive. New algorithms that predict upcoming file accesses have the potential to improve navigation-based file retrieval, but it is unknown how best to present their predictions to users. We present three design goals aiming to improve navigation-based file retrieval interfaces: minimise the time spent at each hierarchical level en route to the target file; reduce the number of levels traversed by providing shortcuts; and promote rehearsal of the retrieval mechanics to facilitate expertise. We introduce three interfaces that augment standard file browsers based on each of these goals: Icon Highlights give greater prominence to predicted items in the current folder; Hover Menus provide shortcuts to predicted folder content; and Search Directed Navigation uses predictive highlighting to guide users through the hierarchy in response to query terms. Results from a user evaluation show that all three interfaces improve file retrieval times, with Icon Highlights and Hover Menus best suited for frequently accessed items and Search Directed Navigation best suited for infrequent ones. We also show that the benefits are larger when folder content is spatially unstable. Finally, we discuss how the interfaces could be combined and deployed in existing file browsers.

Cloud computing systems fundamentally provide access to large amounts of data and computational resources through a variety of interfaces. These resources are provided on demand through file servers available across cloud service providers. Different file replication mechanisms are used to address the key issues such as scalability, reliability and fault tolerance. Though replication potentially improves fault tolerance, it leads to the problem of ensuring consistency of replicas and the nodes are not accessing out-of-date (or stale) file. However, fewer replicas also decrease concurrency and the level of file availability. A correct synchronization between replication mechanism and consistency not only ensures highly reliable and fault tolerant system but also improves system performance significantly. This paper presents a threshold based file replication model that replicates the file on other servers based on the number of file accesses. The simulation results indicate that proposed model reduces the number of messages exchanged for file replication by 33-55% and 25-50% as compared to request reply and request reply acknowledgement protocol, thus improving the system performance significantly.

Data used in high-performance computing (HPC) applications is often sensitive, necessitating protection against both physical compromise of the storage media and "rogue" computation nodes. Existing approaches to security may require trusting storage nodes and are vulnerable to a single computation node gathering keys that can unlock all of the data used in the entire computation. Our approach, Horus, encrypts petabyte-scale files using a keyed hash tree to generate different keys for each region of the file, supporting much finer-grained security. A client can only access a file region for which it has a key, and the tree structure allows keys to be generated for large and small regions as needed. Horus can be integrated into a file system or layered between applications and existing file systems, simplifying deployment. Keys can be distributed in several ways, including the use of a small stateless key cluster that strongly limits the size of the system that must be secured against attack. The system poses no added demand on the metadata cluster or the storage devices, and little added demand on the clients beyond the unavoidable need to encrypt and decrypt data, making it highly suitable for protecting data in HPC systems.

While the amount of data we can process and store grows, our ability to find data remains dependent upon our own memories more often than not. Manual metadata management is common among scientific users, consuming their time while not making use of the computing resources at hand. Our system design proposes to empower users with more powerful data finding tools, such as unified search spaces, provenance, and ranked file system search. By returning the responsibility of file management to the file system, we enable scientists to focus on their science without the need for a customized file organization scheme for their work.

Log files are a necessary record of events on any system. However, as systems scale, so does the volume of data captured. To complicate matters, this data can be distributed across all nodes within the system. This creates challenges in ways to obtain these files as well as archiving them in a consistent manner. It has become commonplace to develop a custom written utility for each system that is tailored specifically to that system. For computer centers that contain multiple systems, each system would have their own respective utility for gathering and archiving log files. Each time a new log file is produced, a modification to the utility is necessary. With each modification, risk of errors could be introduced as well as spending time to introduce that change. This is precisely the purpose of logjam. Once installed, the code only requires modification when new features are required. A configuration file is used to identify each log file as well as where to harvest it and how to archive it. Adding a new log file is as simple as defining it in a configuration file and testing can be performed in the production environment.

Due to their damages to Internet security, malware (such as virus, worms, trojans, spyware, backdoors, and rootkits) detection has caught the attention not only of anti-malware industry but also of researchers for decades. Resting on the analysis of file contents extracted from the file samples, like Application Programming Interface (API) calls, instruction sequences, and binary strings, data mining methods such as Naive Bayes and Support Vector Machines have been used for malware detection. However, besides file contents, relations among file samples, such as a "Downloader" is always associated with many Trojans, can provide invaluable information about the properties of file samples. In this paper, we study how file relations can be used to improve malware detection results and develop a file verdict system (named "Valkyrie") building on a semi-parametric classifier model to combine file content and file relations together for malware detection. To the best of our knowledge, this is the first work of using both file content and file relations for malware detection. A comprehensive experimental study on a large collection of PE files obtained from the clients of anti-malware products of Comodo Security Solutions Incorporation is performed to compare various malware detection approaches. Promising experimental results demonstrate that the accuracy and efficiency of our Valkyrie system outperform other popular anti-malware software tools such as Kaspersky AntiVirus and McAfee VirusScan, as well as other alternative data mining based detection systems.

Along with the rapid development of information technology, individual users of computers are faced with many problems in dealing with massive files, such as enormous amount, serious redundancy and complex version evolution relation. These lead that the file system becomes more and more rough-and-tumble, to a certain extent putting users to trouble in daily work. This paper systematically analyzed the problems in file system management, put forward user cognitive model when faced with massive files, and then introduced the FDTVAS system which is a visual interface based on hyperbolic tree view for exploring the massive files. This system used specific algorithm to do the intelligent analysis of file relationship, took hyperbolic tree as the main visual form, and provided a series of interaction tasks according with users' cognitive model. The built system can help users get a clear insight of the file system and do the analysis and management efficiently.

With the everyday increasing importance of privacy, security, and wise use of computational resources, the corresponding technologies are increasingly being faced with the problem of file type detection. In digital forensic, there are numerous file formats in use. Criminals have started using either non-standard file formats or changing extensions of files while storing or transmitting them over a network. This makes recovering data out of these files difficult. An extension to the file name with the file type is stored in the disk directory, but when a file is deleted, the entry for the file in the directory may be overwritten and hence quite difficult to identify its type which is serious issue in computer forensics. But if the fragment of file has its header information containing type identifying information the mentioned problem may be solved. But it is difficult to identify the type of fragment from the middle or if the header information is deleted or unavailable the identification becomes more complex. This paper focuses on identifying the file types addressing the various scenarios of file type being changed by the malicious user to send some confidential or sensitive information by changing the file type (say.exe banned by Gmail can be converted to any acceptable format and sent across). E-mail spam has become an epidemic problem that can negatively affect the usability of electronic mail as a communication means. Besides wasting users' time and effort to scan and delete the massive amount of junk e-mails received, it consumes network bandwidth and storage space, slows down email servers, and provides a medium to distribute harmful and/or offensive content. Inspired by the success of fuzzy similarity in text classification and document retrieval, the approach investigates its effectiveness in filtering spam based on the textual content of e-mail messages.

Parallel applications running across thousands of processors must protect themselves from inevitable system failures. Many applications insulate themselves from failures by checkpointing. For many applications, checkpointing into a shared single file is most convenient. With such an approach, the size of writes are often small and not aligned with file system boundaries. Unfortunately for these applications, this preferred data layout results in pathologically poor performance from the underlying file system which is optimized for large, aligned writes to non-shared files. To address this fundamental mismatch, we have developed a virtual parallel log structured file system, PLFS. PLFS remaps an application’s preferred data layout into one which is optimized for the underlying file system. Through testing on PanFS, Lustre, and GPFS, we have seen that this layer of indirection and reorganization can reduce checkpoint time by an order of magnitude for several important benchmarks and real applications without any application modification.

Parallel applications often store data in multiple task-local files, for example, to remember checkpoints, to circumvent memory limitations, or to record performance data. When operating at very large processor configurations, such applications often experience scalability limitations when the simultaneous creation of thousands of files causes metadataserver contention or simply when large file counts complicate file management or operations on those files even destabilize the file system. SIONlib is a parallel I/O library that addresses this problem by transparently mapping a large number of task-local files onto a small number of physical files via internal metadata handling and block alignment to ensure high performance. While requiring only minimal source code changes, SIONlib significantly reduces file creation overhead and simplifies file handling without penalizing read and write performance. We evaluate SIONlib's efficiency with up to 288 K tasks and report significant performance improvements in two application scenarios.

Nowadays more and more applications require file systems to efficiently maintain million or more files. How to provide high access performance with such a huge number of files and such large directories is a big challenge for cluster file systems. Limited by static directory structures, existing file systems will be prohibitively inefficient for this use. To address this problem, we present a scalable and adaptive metadata management system which aims to maintain a trillion files efficiently. Firstly, our system exploits an adaptive two-level directory partitioning based on extendible hashing to manage very large directories. Secondly, our system utilizes fine-grained parallel processing within a directory and greatly improves performance of file creation or deletion. Thirdly, our system uses multiple-layered metadata cache management which improves memory utilization on the servers. And finally, our system uses a dynamic loadbalance mechanism based on consistent hashing which enables our system to scale up and down easily.Our performance results on 32 metadata servers show that our user-level prototype implementation can create more than 74 thousand files per second and can get more than 270 thousand files' attributes per second in a single directory with 100 million files. Moreover, it delivers a peak throughput of more than 60 thousand file creates/second in a single directory with 1 billion files.

A cryptographic network file system has to guarantee confidentiality and integrity of its files, and also it has to support random access. For this purpose, existing designs mainly rely on(often ad-hoc) combination of Merkle hash tree with a block cipher mode of encryption. In this paper, we propose a new design based on a MAC tree construction which uses a universal-hash based stateful MAC. This new design enables standard model security proof and also better performance compared with Merkle hash tree. We formally define the security notions for file encryption and prove that our scheme provides both confidentiality and integrity. We implement our scheme in coreFS, a user-level network file system, and evaluate the performance in comparison with the standard design. Experimental results confirm that our construction provides integrity protection at a smaller cost.

Hierarchical folders have been widely used for managing digital files. Previous research has revealed problems with them. Users frequently have to turn to desktop search to re-find files, even if they thought they had saved them in a memorable place. Tagging may have the potential to improve information navigation and organization. This research in progress aims at exploring the possibility of incorporating tagging into the hierarchical folder structure for file management, especially for the process of file organization and file re-finding. This work will provide design implications for future file management tools.

Content pollution is pervasive in the current peer-to-peer file sharing systems. Many previous reputation models have been proposed to address this problem, however, such models strongly rely on the participants' feedback.In this paper, we bring forward a new holistic mechanism which integrates the reputation model, inherent file-source-based information and the statistical data reflecting the diffusion state to defend against pollution attack. First, we deploy a redundancy mechanism to assure that the file requester receives the correct indices that accord with the information published by the file provider. Second, we complement the reputation information with the diffusion data to help the file requester select the authentic file for downloading. Finally, we introduce a block-oriented probabilistic verification protocol to help the file requester discern the polluted files during the downloading with a low cost.We perform a simulation which shows that our holistic mechanism can perform very well and converge to a high accuracy rapidly, even in a highly malicious environment.

File correlation, which refers to a relationship among related files that can manifest in the form of their common access locality (temporal and/or spatial), has become an increasingly important consideration for performance enhancement in peta-scale storage systems. Previous studies on file correlations mainly concern with two aspects of files: file access sequence and semantic attribute. Based on mining with regard to these two aspects of file systems, various strategies have been proposed to optimize the overall system performance. Unfortunately, all of these studies consider either file access sequences or semantic attribute information separately and in isolation, thus unable to accurately and effectively mine file correlations, especially in large-scale distributed storage systems.This paper introduces a novel File Access coRrelation Mining and Evaluation Reference model (FARMER) for optimizing petascale file system performance that judiciously considers both file access sequences and semantic attributes simultaneously to evaluate the degree of file correlations by leveraging the Vector Space Model (VSM) technique adopted from the Information Retrieval field. We extract the file correlation knowledge from some typical file system traces using FARMER, and incorporate FARMER into a real large-scale object-based storage system as a case study to dynamically infer file correlations and evaluate the benefits and costs of a FARMER-enabled prefetching algorithm for the metadata servers under real file system workloads. Experimental results show that FARMER can mine and evaluate file correlations more accurately and effectively. More significantly, the FARMER-enabled prefetching algorithm is shown to reduce the metadata operations latency by approximately 24-35% when compared to a state-of-the-art metadata prefetching algorithm and a commonly used replacement policy.

In most file systems, if a file is deleted, only the metadata of the file is deleted or modified and the file's data is still stored on the physical media. Some users require that deleted files no longer be accessible. This requirement is more important in embedded systems that employ flash memory as a storage medium. In this paper, we have designed a NAND flash file system that has a secure deletion functionality. We modified YAFFS to support secure deletion. Our method uses encryption to delete files and forces all keys of a specific file to be stored in the same block. Therefore, only one erase operation is required to securely delete a file. The proposed method securely deletes not only keys but also all of the metadata of that file. Our simulation results show that the number of block erases due to file creation and file modification is very low and the amortized number of block erases is lower than the simple encryption method. Even though we applied our method only to the YAFFS, our method can be easily applied to other NAND flash file systems.

XML is the predominant format for representing structured information inside documents, but it stops at the level of files. This makes it hard to use XML-oriented tools to process information which is scattered over multiple documents within a file system. File System XML (FSX) and its content integration provides a unified view of file system structure and content. FSX's adaptors map file contents to XML, which means that any file format can be integrated with an XML view in the integrated view of the file system.

Content-based indexing [MCM01] is a technique of proven effectiveness for efficient transference of file contents over low bandwidth network links. Departing from this context, the natural step of extending the application of this technique to local file storage has been proposed by a number of storage solutions [CN02, QD02, BF04]. To some extent, all these solutions share a core storage model. File contents are divided into disjoint chunks of data, each of which is individually stored, along with a unique hash of its contents, in a repository of chunks. The actual files are then stored as sequences of possibly shared references to chunks in the repository.

The paradigm of Ubiquitous computing seeks to build a computing environment that responds to user context. An ideal file system for the Ubiquitous environment is one that can successfully recognize the present context and automate file management. The intelligence in the Ubiquitous file management is achieved by applying a heuristics based clustering approach to the system. The applied heuristics are those that are used on file attributes by users to manually manage files in a traditional file system. File attributes can be used to relate files to the most appropriate work-context and also draw inter-file relationships. We discuss methods to harness the given file information from the file-system to form a context-relation wrapper over disparate files. This enables management of files as context related working sets rather than as individual files. A survey was conducted among regular computer users and the compiled results supported our context based file clustering approach. The experiments also showed promising results that confirm our model on the file heuristics, thus finding semantic relations between those files.

We consider an architecture for a serverless distributed file system that does not assume mutual trust among the client computers. The system provides security, availability, and reliability by distributing multiple encrypted replicas of each file among the client machines. To assess the feasibility of deploying this system on an existing desktop infrastructure, we measure and analyze a large set of client machines in a commercial environment. In particular, we measure and report results on disk usage and content; file activity; and machine uptimes, lifetimes, and loads. We conclude that the measured desktop infrastructure would passably support our proposed system, providing availability on the order of one unfilled file request per user per thousand days.

The self-organizing file cabinet is an information retrieval system associated with a user's physical file cabinet. It enhances a physical file cabinet with electronic information about the papers in it. It can remember, organize, update, and help the user find documents contained in the physical file cabinet. The system consists of a module for extracting electronic information about the papers stored in the file cabinet, a module for representing and storing this information in multiple views, and a module that allows a user to interact with this information. The focus of this paper is on the design and evaluation of the self-organized file cabinet.

Among the most reliable and fault tolerant components in a distributed system are storage systems. Obviously, reliability of storage systems belongs to the most researched issues in distributed computing. Every distributed file system project is based on different assumptions about size, load, amount of sharing, and desirable semantics, making it hard to compare research results fairly. The current Amoeba file server is the Bullet File Server [van Renesse, Tanenbaum, and Wilschut, 1989] which provides immutable files, is optimized for whole-file transfer and does caching at the file server. It has excellent performance for reading cached files (1.5 + 1.5 n ms for n kilobytes) and for sustained file I/O (680 kilobytes per second, both on read and write). Although performance is excellent, there is room for improvement, especially in the area of fault tolerance, sharing semantics and caching. I am currently doing the back-of-the-envelope design for a new file server that will form the basis of both our normal file system and of a complex-object server which is being designed by the database group at CWI. In addition to those desirable properties of fault tolerance, persistency, consistency, and availability, I am anxious to achieve even better performance than the Bullet server by extensive use of client and server caching.This position paper presents some of our design ideas. Note that this is work in progress; that the design is only just starting and that an implementation of a design will not start for another six months or so.