Distributed File Systems   Term Paper ID:31293 Click to get this paper

Essay Subject: Considers various aspects of distributed file systems, including implemention and scalability.... More...

14 Pages / 3150 Words 4 sources, 12 Citations, MLA Format $56.00 Return to List of Papers

Paper Introduction: Distributed File Systems Introduction Today\'s computer systems are highly dependent on distributed filesystems Although today\'s computer systems have storage capacity fargreater than at any time in the past and although that storage is lessexpensive than at any time previous there is so much data to be stored referenced and accessed that without distributed file systems it would benearly impossible for today\'s systems to be as widespread as they are Distributed file systems offer greater flexibility to companies andorganizations by making it possible Text of the Paper: The entire text of the paper is shown below. However, the text is somewhat scrambled. We want to give you as much information as we possibly can about our papers and essays, but we cannot give them away for free. In the text below you will find that while disordered, many of the phrases are essentially intact. From this text you will be able to get a solid sense of the writing style, the concepts addressed, and the sources used in the research paper. Conclusions Distributed file systems offer an attractive and important tool forinformation professionals to offer reliable data to multiple users at thesame time. 8-9.Biggar, H. Generally, this istraced to the 196 s when many large organizations were moving tocomputerization and users began to have the need to share information withone another. In this way, systems that are distributed are able toexchange information-that is, send and receive information-to filesregardless of their geographic placement within the distributed system(Milojicic et al, 2 ). There is the concernthat files will be viewed for malicious purposes, or simply by those whoshould not have access to them. K. A processappears to run on a single workstation throughout its entire processingcycle even when it was migrated to several workstations. A good implementation plan iscritical to the successful deployment of a distributed file system. One of the benefits of Sprite was its transparency in terms oflocation, a key factor for distributed file system success. This is only possible in distributed filesystems where the various processors have access to the data that is neededwhen it is needed, and that can update and replace the modified file sothat the next processor can locate the file (Garg, 2 2). Clustered filesystems are likely to be the next direction for this type of storage asorganizations seek to manage and control the many files that they generate. Migration wasintended to take advantage of idle workstations, but Sprite did not providefull load balancing. If a personal computer or other standalonesystem crashes, it typically becomes unavailable to all users for allfunctions. Generally, security issues aretaken into account during the design and deployment of a distributed filesystem as well as throughout its operation (Garg, 2 2). Key to the MOSIX design is the division of theUnix kernel into a lower kernel and an upper kernel. However, passwords are also considered a rudimentary andeasily-overridden security protocol. New file-sharing options address user demands. Implementation Implementation of any system is a critical component in the process.Ideally, an implementation plan is developed that includes all phases ofthe implementation, from initial research through eventual deployment.Most large enterprises will not be implementing a distributed file systemin a "Greenfield" environment, but instead will be upgrading an existingsystem or replacing an existing system. Security issues,both physical and logical, should also be included in the implementationplan, along with maintenance and upgrade paths. The problem becomes more complex when networksare involved, particularly when there is Internet access present, sincethat makes the entire network more accessible to unauthorized attacks.Here, software such as firewalls can provide some security to the pointthat organizations can have greater confidence in their distributed files. An example of this type of application arises when acustomer's computer contacts a vendor's computer to place an order becausethe customer's stock level has reached a predetermined level.Sophisticated systems will inquire about the stock on hand at the vendor,and adjust to a different product if the requested item is out of stock.To do this, the customer's computer must be able to read the data on thevendor's computer, and have the security clearances necessary to do so. Whenstorage space permits, many organizations require that areas on the networkare set aside for backup data and redundant copies of key applications inorder to ensure that service is not interrupted in case of a problem in akey area. At this point, it was common for a single organization tohave multiple mainframe systems running in different geographic locations,but there was no easy way to share information among these systems evenwithin the same organization. Thisis because Linux is a no-cost operating system that is designed for openplatforms and that has proved to be scalable and flexible enough to run onmany different types of hardware platforms. Distributed systems are generally understood to be collections ofmultiple independent processes, processors or computers. Sprite's goals were to provide workstation autonomy so that the usersof the local workstation had priority over their workstation's resources.Processor migration was used primarily as a way to remove other users'processes from a particular workstation, for example, which can cause someproblems in the implementation. As SFS from Hewlett-Packard demonstrates, thereis likely to be increased blurring of the distinctions between shared anddistributed file systems in the future, and scalability will likely beaddressed by these new hybrid systems (Biggar, 2 4b). This is due in part to the perceived communication problemsthat some have noted with distributed file systems, although proponents ofthese systems maintain that the issue is a strategic rather than tacticalone (Biggar, 2 4a). These are typically solvedthrough a combination of hardware and software solutions. SFS uses a file server configured so that expansion takes placeacross grids. Instead, by"striping" more cells across data, SFS users are able to maintainperformance that meets the performance capabilities of the network as awhole as bottlenecks are eliminated (Biggar, 2 4a). Shortcomings When considering implementing a distributed file system, organizationsmust take into account the goal that they are trying to accomplish. InfoStor, 8, pp. If part of a network crashes, or if some files become corruptin a distributed file system, the problem is typically isolated. Reliability isperhaps the most critical of these factors since it does users little goodto have access to data if they do not have confidence in the reliability ofthat data. One of the key features of a distributed file system is the issue ofreliability through redundancy. This redundancy helps maintain a high level of quality controland provide a high level of service quality (Garg, 2 2). This can requiresignificant programming in the form of conversion programs to ensure thathomogeneity is maintained across various hardware platforms and softwareapplications. The earliest solution to this situation was the development ofARPANET, a wide area network that proved to be the forerunner of theInternet (Milojicic et al, 2 ). (2 2). S., Douglis, F., Paindaveine, Y., Wheeler, R., & Songnian, Z. In the most extreme cases, unauthorizedusers will access and modify files over which they have no authority.Passwords are a common and simple way to provide access to files regardlessof whether they exist on a standalone workstation or within a distributedfile system. One of the key goals ofMOSIX is to provide each process with a uniform view of the systemregardless of where the process is located. Today, MOSIX, Sprite, Mach and LSF are typical examples ofsystems that have been effectively implemented in a variety of applicationenvironments (Biggar, 2 4a). ReferencesBiggar, H. It should makeno difference where the data is located on the network, nor should itmatter when the data is accessed; the goal is to have the data availableand reliable when the user requests it (Biggar, 2 4a). This operating system performs load balancing automaticallybased on process migration, and offers transparency of data as well asdecentralized control and dynamic configuration. Users on acomputer at one workstation can access files-according to permissions andsome other variables-physically located on other computers that may belocated across the room or across the world. The implementation plan starts with the definition of what the filesystem is designed to accomplish; that is, the organizational challengesthat it resolves. Training is also a critical part of the implementation plan, with theinformation technology department-or its equivalent-adequately trained toaddress most problems that arise with the system. Within a single facility,for example, there may be considerable homogeneity in the hardware andsoftware that is used to access the files. This requires that theIT professionals understand not only how the file system is set up, butalso how the system handles failure in one part of its operation as well aswhich processes are part of the organization's critical path. Service quality is also enhanced by the ability of users to takeadvantage of multiple processors in most distributed environments. Even home computers are often networked in this way. Requirements Analysis There are four key areas that are taken into account when developingrequirements or standards for distributed file systems: reliability,communication, homogeneity and security. However,such an approach affects the hardware and software decisions that are madein regard to network architecture, as well. Certainly this applies to peripherals such as printersand scanners, which can be networked and accessed by any computer on thenetwork, but access to storage devices offers even greater benefit.Personal computers offer good ratios of price to performance, but loadingthe same software on hundreds or thousands of computers in an organization,and keeping that software updated, can pose significant maintenance issues. They are largely hardware independent, so that users canaccess data from an IBM computer, a Mac computer, or some other computerthat meets the organization's budget requirements. Implementation also requires a migration plan for files and thecreation of metadata information if this is a new distributed system thatis being implemented. New York: John Wiley and Sons.Milojicic, D. Distributed File Systems Introduction Today's computer systems are highly dependent on distributed filesystems. Under Sprite, individual workstations interacted as though theywere time shared in terms of accessing network resources, but theperformance was enhanced so that users did not notice a degradation inservice levels. Thisclustering also eliminates the problems of scalability that distributedfile systems encounter when additional systems are added-creatinghorizontal scaling problems-or when additional storage is added (creatingvertical scaling problems. The need for distributed file systems originally arose from the needto share information among users of disparate systems. 241. Whiledistributed file systems certainly have advantages-and these can besignificant-there are also advantages associated with shared file systems.Technical computing environments, for example, might well favor distributedfile systems while shared file systems may be found in enterpriseenvironments. The challenge became even greater whenindividuals from different organizations, which might use differenthardware, discovered that they had a need to share information with oneanother. Information about files is stored in metadata smartcells; these cells are used to access specific files across the system as awhole. Thisimplementation plan should include the goals of the system as well asdescriptions of users and how they will use the system. More secure environmentsmight include biometric access limitation. This can help ensure that only authorized users haveaccess to specific data. Security is a key area of any distributed file system. Sprite provides a shared network file system with both asingle-system image and a consistent cache designed to ensure that allusers had access to the most recent data. The lower kernel isresponsible for "lower" functions, such as device drivers and contextswitching, and is independent of all other nodes. In this way, informationexchange remains a key reason for distributed file systems. When files arelocated on a single standalone computer with a single workstation attached-such as a personal computer that is not connected to the Internet or anyother type of network-security can be a relatively straightforward matter.Access can be controlled physically-through locking an office-as well aswith simple software tools such as passwords. This research offers an overviewof distributed file systems, including security, weaknesses and typicalapplications for the use of distributed file systems. This met the requirement fordata reliability that is necessary in distributed file environments(Milojicic et al, 2 ). (2 4b, April). In addition, thisis not an area that has reached its technological limits; as new operatingsystems such as Linux evolve and as new advances in communications andstorage devices are introduced, this field will continue to expand andprove interesting. This does not necessarily mean that the file system providesinformation on the accuracy of the data, only that the system is providingthe data that is being requested in its most current state. Depending on the size of the organization and theanticipated size of the file system, the organization may choose to issuean RFP (request for proposal) from outside vendors that will serve assystem integrators, or it may choose to integrate the system on its own.In some cases, organizations may choose to develop their own distributedfile system, but given the complexity of the systems and the many variablesinvolved-as well as the criticality of the system to the organization'slong-term success-most organizations are advised against this option. They offer greatersecurity by making use of redundant disk arrays and geographicallydispersed data. While there is some debate over what constitutes a shared filesystem and what constitutes a distributed file system, some analystsmaintain that the difference is in the use of clustering. As nodes areadded to traditional file systems, bottlenecks prevent transmission andreception at the same linear rate as nodes are implemented. Thisallows other functions to continue, including the rescue functions. Lustre was developed by Hewlett-Packard, theDepartment of Energy, and Cluster File Systems specifically for the Linuxoperating environment (Biggar, 2 4a). The time necessary to send andreceive data generally outweighs the time required to actually manipulatethe data in wide area networks; in local area networks, communication timetends to be far faster. Internet applicationssuch as Napster make use of rudimentary distributed file systems that arenot accessible to all users and which offer only limited access terms ofthe types of files that can be accessed; more complex distributed filesystems contain metadata about the files themselves and information aboutthe specific data stored on the system. Sprite relies on the idea of a "home" machine, withany process not running on its "home" machine deemed a "foreign" process.Child processes would also be allocated the same "home" machine as theparent. New organizations will beselecting a file system that will serve their needs for the near future andthat will offer easy upgradeability and scalability as the needs of theorganization change. One of the key advantages to SFS is that it overcomes the problemthat traditional file servers have of diminishing returns. From a user perspective, each process appeared to run on its homemachine, however (Milojicic et al, 2 ). Thus a network designer may focus on doing as muchcommunication as possible at the local level and using the wide areanetwork, such as the Internet, for occasional transfers of data. HP to ship shared file system. This topic was interesting to research because it is dynamic and becauseit impacts so many different aspects of an organization. However, once the computer isconnected to a network, it becomes vulnerable to malicious or negligentactions. ACM Computing Surveys, 32, p. Another reason that distributed file systems are commonplace isresource sharing. In keeping with the open architecture that Linux provides,SFS uses the Lustre protocol. These cells may contain entire files, when files are small, orparts of larger files. Users will need to be consulted about the variousways in which files are used, and who has access to the files, and theimplementation plan will also need to take security concerns into account. The larger the distributed file system, the greater the role thathomogeneity plays in its design and execution. Communication time forms another consideration when developingrequirements for distributed file systems. Epilogue This subject is important because the amount of data that is beinggenerated in most organizations continues to expand. Because these systems were often proprietary, it is notsurprising that few standards in terms of distributed operating systems oragents emerged. Quality Control Security is a key concern when files are shared. Process migration. Design As mentioned above, distributed operating systems offer considerableflexibility and the best fit for distributed file system environments.MOSIX is a distributed operating system developed in Jerusalem in the early198 s with regular releases occurring since that time. 8-9.Garg, V. These areautonomous systems, meaning that they are able to be controlled independentof one another. Manyorganizations supply their employees with a discrete computer-either adesktop or laptop computer-at their workstation; these are connected via anetwork. Today,distributed file systems run on most types of operating systems, includingUnix, Windows and the MacOS, although distributed operating systems offerthe greatest flexibility and best fit for distributed file systems.Initially, developments in this area came from academia and governmentagencies as well as researchers at IBM who sought to provide their largestusers with standards that would enable them to access data on distributedfile systems. The grid expansion technique allows nodes to be added in "smartcells". When individuals from different organizationsseek to share information or access files, the challenge can be significantto ensure that all users have access to the same data. As Linux has gained market share, there has been increased interest indistributed file systems that make use of the Linux operating system. As users become moregeographically dispersed, the control over how they access various filesbecomes less homogeneous. Today's computing environment is much more complicated. MOSIX is looselybased on Unix. At the same time, although it can be cost prohibitive to add sufficientdisk space to each personal computer, or even to predict accurately howmuch each user will require, disk space can be added centrally for accessby all users in a straightforward manner. Although today's computer systems have storage capacity fargreater than at any time in the past, and although that storage is lessexpensive than at any time previous, there is so much data to be stored,referenced and accessed that without distributed file systems, it would benearly impossible for today's systems to be as widespread as they are.Distributed file systems offer greater flexibility to companies andorganizations by making it possible for more employees to have access tomore information than could possibly be stored locally. In addition, Linux hasdemonstrated a high level of reliability that enhances its attractivenessto IT professionals considering a new or upgraded distributed file systemsolution (Biggar, 2 4b). According tothese analysts, true clustered file systems use a single pool of storagefor data and metadata; this enhances the scalability of the system. Inthis way, homogeneity can be a critical success factor for both companiesif this time-saving device is actually to save both companies time andmoney (Biggar 2 4b). Increasingly, organizations are turning to shared file systems toovercome some of the problems that have been identified in distributed filesystems. Requirements were developed by independent teams working in largelyisolated environments, although several standards have emerged. Reliability isdependent on the integrity of the system as well as the integrity of thecommunication between various components within the system. This willhelp the IT department respond quickly and efficiently when problems arise,and will help ensure that the system is able to meet the needs of thevarious stakeholders. In thisway, those processes which are particularly processor-intensive can be madeto share work across multiple processors so that the work is done moreeffectively and efficiently. (2 , September). This "single system image"helps maintain data integrity and is critical to the reliability andflexibility of the system. However, without the assurance that localusers would have autonomy over their processes, many users would bereluctant to have foreign processes run on their workstations and thebenefits of distributed file systems would be lost in this environment. This greatly facilitatesperformance, maintenance and access control. Reliability is a key issue when considering distributed filesystems, as is scalability and transparency. Sprite is a network operating system developed at the University ofBerkeley that is loosely based on traditional time-share computingconcepts. Increasingly, companies are usingnetwork-specific security measures, such as determining specificpermissions for specific workstations, or biometric identification ratherthan passwords. Elements of Distributed Computing. Hewlett-Packard is offering its Scalable File Share (SFS) solution tohigh-performance computing environments that use the Linux operatingsystem. (2 4a, August). Firewalls form a passive form of security; honey pots areexamples of aggressive methods of security. InfoStor, 8, pp. The upper kernelprovides the traditional Unix interface, runs on each node, and handlesonly universal objects (Milojicic et al, 2 ). Today's systems are evolvingto the point where it will be difficult to distinguish between distributedand shared file systems in the future, and the systems are growingincreasingly large as they seek to accommodate the ever-larger amounts ofdata generated by companies and governments. If this paper is not what you are looking for, you can search again: Search for: or Click here to request an essay written just for you.

Many of our Papers can be Downloaded From This Site!

PLEASE READ THIS, IT IS IMPORTANT!

Office hours are Monday through Friday, from 9 am to 5 pm (PST). You may place orders for custom research over the phone during office hours. E-mail requests can be made to our graduate and undergraduate department any time, and will be reviewed during office hours. You may also contact customer service any time through e-mail, and we will review your message during business hours. A great many papers can be downloaded right from this site, but not all of them. If you would like to know if a particular paper is downloadable, just look in the description for: "Available for Internet Download: Y" or "Available for Internet Download: N" If you wish to purchase a paper which is NOT available for immediate download, you will need to make other shipping arrangements. Also, please be aware that these orders are processed Monday through Friday from 9 am to 5 pm (PST). If you place your order after 4:45pm on Friday, it will not be processed until the following Monday morning. We charge $8 per page for all of our pre-written reports, plus shipping (and tax for California residents). However, the highest cost of any ONE report is $136, or 17 pages. Please, take a moment. Make sure you have chosen the report you want or need BEFORE you complete your order. If you are not sure, allow us to help you. We do not offer refunds or exchanges, so it is important for you to let us answer your questions during office hours. Reports which are e-mailed or downloaded are in Microsoft Word format. We are making more reports available for e-mail delivery faster than we can update our listings. Please call to check on the status of particular reports. There are many other shipping options which are listed on the Checkout page.