A Concurrent-Access File Server for Mach-Based Multiprocessor Machine

• Main Authors: Wu,Shu-Chin, 吳淑琴
• Other Authors: Juang,Jie-Yong;Lee,Share-Young
• Format: Others
• Language: en_US
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/31910301498502170728

碩士 === 國立臺灣大學 === 資訊工程研究所 === 82 === Multimedia is popular and file sizes grow larger and larger. Although micro processor is getting powerful today, the overall performance od systems with massive data will be influenced by I/O speed. Current file systems, such as MS-DOS, and UNIX, are not suitable for large files because they only access one block of data at a time. Suppose the size of a file is 16K bytes and it is stored on a disk continuously. When a user wants to get the whole data of this file at once, these systems need to issue 32 disk requests to get data. It wastes time. Thus, we build a file server, called Archer, such that it good performance for large files. Archer adopts contiguous allocation and disk arrays to improve I/O speed. It is established on a multiprocessor machine. Thus, clients can mount more than one Archer servers, which sit on different hosts. When one host crashes, clients can access data through another host as no matter happens. Furthermore, current file systems can communicate with Archer through NFS interface.