Distributed File Systems - Essay Example

Introduction: Distributed file systems allow users to share the resources of a network including storage. The design of the network protocol determines the level of access that clients have to files and thus it is possible for the server to restrict access to storage and network files. Ainsworth (2009, p. 22) points out that the design and implementation of distributed file systems faces many challenges such as caching techniques, naming service and transparency, file replication, access control and scalability considerations, and support for concurrency and availability. These challenges will be discussed in this section and also there will be a description of the main architectural features of the Network File System version 4 in relation to its previous versions. The main challenges in designing and implementing distributed file systems: Adrian, Teodora and Naiana (2011, p. 801) say that in distributed file systems, caching is an approach that is used to reduce the traffic within the network. Moreover, Tanenbaum and Van Steen (2007, p. 276) emphasize that through catching the reliability and speed of access of data from the server is significantly increased. This means that the users of the system are able to access data with a guarantee of no delays in retrieval. Additionally, caching is aimed at preventing repeated access same data or information. It is through catching that access to data within a system is attained by one user without the redundancy which often results from many users requiring to access, retrieve or modify the same data at the same time. Tanenbaum and Van Steen (2007, p. 277) illustrate that unlike server access, which costs thousands of instructions, cache access is faster as it only costs the client terminal hundreds of instructions. Ainsworth (2009, p. 22) argues that the problem of consistency can be overcome by implementing a write-through technique within the cache policy. The reliability of a write-through cache is relatively high. A delayed write approach may also be used because it saves remote writes, but has relatively poor reliability, especially in situations of a system crash. Caching techniques are designed so that the cached copies of files are presented to the user from the server in an effective manner. Therefore, users have access to the cached copy of files instead of contacting the server directly which mightlead to an overload of the network. According to Dolev and Kat (2005, p. 135), caching presents a challenge of consistency within the processes of file sharing and transfer within a distributed file system. This problem results from the difficulties associated with keeping consistency between the cached copies of files and the master file. Consistency between the server and files within the cache is needed because lack of this consistency means that the users will not be able to access files which are up to date. Dolev and Kat (2005, p. 138) show that the naming service of a distributed file system involves mapping the objects of the network. Naming includes both physical and logical objects within the network. This is necessary because logical data such as file names require mapping to physical locations such as disks. The naming of files ensures that the allocated name is relevant to the attributes of the file such as location and contents of the file. According to Kline (1999, p. 39), the challenges that distributed file systems design face include achievement of transparency within the system without compromising on file naming. Location transparency is another challenge within distributed file system. Even though users require the location of files within a system to be as transparent as possible, security concerns for system databases may warrant the need for limited disclosure of file location. Therefore, it is necessary that user are allowed access to the content of files without necessarily knowing the location of file storage. This means that the file location and names are given a degree of independence as a way of overcoming this challenge without denying users access to files. File replication can be used a technique to improve availability and scalability, but on the other hand, the naming scheme of files may lead to mapping files into a specific replica. This means that the location of two or more different files within the memory may be mapped into a single location. The fundamental challenge of replication within a distributed file system is to keep consistency and maintain it among all replicas Dolev and Kat (2005, p. 148). Since the value of each of the logical items in a replication system are stored as copies as one or more data items in a physical location, inconsistent replication within a distributed system is likely to cause lack of file availability to the users Coulouris, Dollimore and Kindberg (2005).Replicas within distributed file system include location replicas and it is necessary for replication control to be implemented so that this challenge can be overcome. The file replication challenge in implementing distributed file systems is determined by the degree of replication and consistency required.The read and write operations for example may be required to be mapped to the corresponding file operations within the physical copies. In this case, the consistency of the system depends on how and when the mappings are carried out Coulouris, Dollimore and Kindberg (2005). Access control is necessary within distributed file systems because the collaboration and data sharing which characterize these systems would lead to security threats by attackers. Liu (1992, p. 459) says that the challenges that face the design and implementation of distributed file systems are the provision of security and scalability through appropriate access control mechanisms. Distributed systems are characterized by remote access of files which would lead to unauthorized access to the system by unauthentic clients. Access control becomes a challenge to distributed systems implementation basically because their goal is to provide file access, sharing and communication among the users of the system. Traditional designs of distributed file system often employ access control mechanisms which depend on the reference monitors to control the access of users to the resources of the network. Reference monitors lead to the problem of scalability. Dolev and Kat (2005, p. 147) argue that the problem of access control can be overcome by the implementation of cryptographic techniques in the design of distributed file systems. The use of cryptography will thus enhance the integrity and confidentiality of data access, use and management within the system. This would be achieved by ensuring that information sources are restricted to the authentic clients of the distributed file system. According to Kline (1999, p. 38), concurrency is another problem within distributed file system design and implementation, which results from situations where more than a single user is accessing or updating a single file. The design challenge within these systems arises due to the problem of ensuring that the update of a file by a single client does not interfere with the access of the same file by a different user even if the file is being accessed or updated simultaneously. Nightingale, Chen and Flinn (2006, p. 361) say that the challenges of distributed file system design that are related to concurrency can be overcome through the implementation of concurrency locking or control mechanisms. However, this would cause a reduction in the scalability of the distributed file system. Affording availability within a distributed file system is another challenge for the design and implementation of these systems. The design of a distributed file system should ensure that during its implementation, users have unlimited access to the system as illustrated by Triantafillou and Neilson (1997, p. 35). This means that the system must be made available to the clients at all times. Availability means that the efficiency of access and sharing of data or information within the system must be enhanced. The challenge which high availability causes on a distributed file system is the exposure of the system into security attacks and thus unauthorized access to information and data within their system. Moreover, the databases within a distributed file system must be accessible and usable by the users. It is through design and implementation of security protocols that the accessibility of a distributed file system would be guaranteed. A secured system will be protected from attackers who would sabotage it and so prevent users from accessing it. The network File system’s Version 4 Main Architectural Features Network File system (NFS) is used within distributed file environments as a file sharing protocol. Coulouris, Dollimore and Kindberg (2005) point out that NFS is designed to enable users to access data within the distributed system with the same efficiency and convenience like that afforded by the direct access of files from the local disks of the client computer. The NFS is defined as a standard protocol and as a result it can be accessed and used by anyone within the distributed file system. The design of the NFS architecture is advantageous because the client computer and server storage lie within the same architectural level of file access and as a result increases accessibility of files within the system. According to Peters(2005, p. 86), version 4 of the NFS is an improved version of the file sharing protocol and it provides superior qualities to a distributed file system as compared to NFSv2 and NFSv3.This is due to the architectural design of the NFSv4 which is in the form of parallel file systems Peters (2005, p. 88). As a result of this architecture, there is more throughput within the NFSv4 as compared to the older versions Chen and Flinn (2006, p. 361). The performance of a distributed file system that employs the NFSv4 is much higher in terms of efficiency, speed and scalability than the older versions of the protocol architecture Peters (2005, p. 89). However, it is argued that the challenge of implementing NFSv4 results from the need to ensure that performance is high without any significant sacrifice on the consistency of the system Kline (1999, p. 38). Because of the architectural design of NFSv4 which allows users to access files at the same level as the server, the users can read a relatively high capacity of data from the server at once. A file capacity of more than 2GB can be read by the user from the server storage at once when the NFSv4 architecture is implemented. This performance is mush high as compared to the NFSv2 architecture in which users can only read a maximum of 2GB file size at once. The high performance of the NFSv4 is attributed to the higher throughput which is achieved by the parallel file system architecture. However, it is important to note that there is lack of a seamless integration within parallel system data transfers. Miltchev, et al. (2008, p. 10) explain that the high performance of NFSv4 is attributed to the fact that the access and retrieval of files by the user from the server into the client computer is faster and more effective than previous versions of the NFS. Parallel file system architecture which is implemented by the NFSv4 system has a security challenge which is related to multiple access which users can achieve to files within the system. However, the architecture provides high performance which demonstrates a desirable tradeoff since access, retrieval and use of files in the NFSv4 architecture is much efficient as compared to the early version. NFSv4 architecture is designed to provide a more secure environment of file sharing as compared to previous versions of the system. Coulouris, Dollimore and Kindberg (2005) explain that high level security protocols are provided within the NFSv4 architecture with the aim of ensuring that the access, retrieval and use, distribution and sharing of files within a distributed file system is secured from both internal and external attacks to the system. The security protocols within the NFSv4 architecture include authentication and cryptography which secures the system more than the previous versions of NFS. Glass (1993, p. 136) demonstrates that the characteristics of NFSv2 and NFSv3 include presentation of the protocol in a stateless manner. This means that the application of the older versions of NFS and their implementation does not represent the core protocol of a system. This argument illustrates that the NFSv4 architecture has superior qualities in its design and so more appropriate and advantageous when applied within distributed file systems. References: ADRIAN, G., TEODORA, V., and NAIANA, T. (2011) Models and Solutions for the Implementation of Distributed Systems.Annals ofthe University Of Oradea. Economic Science Series, Vol. 20, no. 1, pp. 801-806.[online]Available at: http://anale.steconomiceuoradea.ro/volume/2011/n1/106.pdf [Accessed 31st March 2012] AINSWORTH, D., (2009) IT in action: Distributed file systems. Third Sector, (596), pp. 22-22. [online] Available at: http://search.proquest.com/business/docview/231714225/135E133401A73A30559/1?accountid=45049 [Accessed 31st March 2012] COULOURIS, G., DOLLIMORE, J. and KINDBERG T. (2005). Distributed Systems: Concepts and Design, 4th Edition. Addison Wesley. DOLEV, S., and KAT, R. (2005). Self-stabilizing distributed file system. Journal of High Speed Networks, Vol. 14, no. 2, pp. 135-153. [online] Available at: http://www.cs.bgu.ac.il/~kat/dfs/resources/draft/beta-sync.pdf [Accessed 31st March 2012] GLASS, B. (1993). Network File System: Essential but flawed. InfoWorld, Vol. 15, no. 46, pp. 136. [online] Available at:http://search.proquest.com/business/docview/194270507/135DDFB504D597DA650/1?accountid=45049 [Accessed 2nd April 2012] KLINE, B. (1999). Distributed file systems for storage area networks. Computer Technology Review, no. 02789647, pp. 34-43 [online] Available at: http://search.proquest.com/business/docview/220619055/135E13503FE4F7A380C/1?accountid=45049 [Accessed 1st April 2012] LIU S. O. (1992). Analysis of Optimal File Migration Policies in Distributed Computer Systems.Management Science, Vol. 38, no. 4, pp. 459-482. [online] Available at: http://search.proquest.com/business/docview/213230831/135DE004F761A4593E1/1?accountid=45049 [Accessed 31st March 2012] MILTCHEV, S., et al. (2008).Decentralized Access Control in Distributed File Systems.ACM Computing Surveys, Vol. 40, no. 3, p. 10. [online] Available at: http://www.cs.columbia.edu/~angelos/Papers/2008/dfs_survey.pdf [Accessed 2nd April 2012] NIGHTINGALE, E., CHEN, P., and FLINN, J. (2006). Speculative Execution in a Distributed File System.ACM Transactions on Computer Systems, Vol. 24, no. 4, pp. 361-392. [online] Available at: http://www.cs.cmu.edu/~dga/15-849/papers/speculator-sosp2005.pdf [Accessed 31st March 2012] TANENBAUM, A. S. and VAN STEEN, M. (2007). Distributed Systems: Principles and Paradigms, 2nd Edition, Prentice Hall. TRIANTAFILLOU, P. and NEILSON, C (1997). Achieving strong consistency in a distributed file system. IEEE Transactions on Software Engineering, Vol. 23, no. 1, pp. 35-55. [online] Available at: http://search.proquest.com/business/docview/195565858/135DE05A97F773E7D3F/1?accountid=45049 [Accessed 31st March 2012] PETERS, O. (2005). AppleXsan in the real world.Videography, Vol. 30, no. 3, pp. 86-90 [online] Available at: http://search.proquest.com/business/docview/199903313/135E12CB468605F6656/12?accountid=45049 [Accessed 3rd April 2012] Read More