Data processing: database and file management or data structures – Database design – Data structure types
Reexamination Certificate
2000-12-20
2003-11-25
Rones, Charles (Department: 2175)
Data processing: database and file management or data structures
Database design
Data structure types
C707S793000, C709S239000
Reexamination Certificate
active
06654769
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a file system for processing files distributed and managed in a plurality of disk devices, and more particularly to a file system in which when a plurality of I/O paths are provided to access a disk device, it is possible to control switching among the plurality of I/O paths so as to access the disk device through one of the plurality of I/O paths.
In the UNIX file system, which is one of the conventional file systems, a number (a file ID) is defined to uniquely identify each file, and a file server can specify a file on which to perform read/write operation by supplying its file ID. The file server registers and manages a matching relationship between each file ID and an I/O path used to access a disk device storing the file, in a file management table (referred to as an inode in UNIX) stored in a memory. An I/O path is composed of such information as a node number, an I/O interface number, a device number, etc. This management system is described in, for example, a book entitled “The Design of The Unix Operating System” authored by Maurice J. Bach (p60-p72).
Upon receiving a read/write access request with a file ID specified, the file server refers to the above file management table, determines an I/O path used to access the disk device based on the file ID, and accesses the disk device using the I/O path. In addition to the I/O path information, the file management table stores file management information such as each file size, the date of last update of each file, etc., and the file management information is read out from a disk device when a file is opened and it is written back to the disk device regularly or when the file is closed. The file server converts a file name supplied by the user to its file ID.
A known method for handling a plurality of disk devices in a file system is to add a name tree managed by a disk device B to a directory, for example, X within a name tree managed by a disk device A so as to show a plurality of disk devices (that is, a plurality of name trees) as if they existed within a single name tree. In this method, the user can access a file in the disk device B by accessing the directory X. This method is called “mount operation”. At system start-up, the file server carries out one mount operation after another using a specific disk device (a route device) as a starting point so as to show a plurality of disk devices to the user as if they were a single name tree. A mount construction file in the route device stores information on a matching relationship between each disk device to be subjected to mount operation at system start-up and a directory name (a mount point) of a name tree onto which the disk device is mounted. The file server performs mount operation according to the information stored in the mount construction file at system start-up.
The mount construction file includes information on each I/O path used to specify a disk device to access it. The file server reads the matching relationships between I/O paths and mount points registered in the mount construction file into memory as mount construction information at the time of performing mount operation. When the user opens a file by specifying its file name, the file server obtains an I/O path used to access the physical disk device storing the file based on the above mount construction information, and creates a file management table. Accordingly, when the configuration of a system has been changed as a result of, for example, connecting a new disk device to the system, the system administrator must set new construction information in the computer system by rewriting the mount construction file.
On the other hand, to enhance reliability of computer systems, such a patent publication as Japanese Laid-Open Patent Publication No. 10-275090 (1998) describes a method which physically connects two different nodes to a physical disk device so as to be able to access the disk device through two different I/O paths. With this arrangement, one of the I/O paths is used in normal operation, and when a node fault has occurred and as a result, it is no longer possible to use the current I/O path, the other I/O path is used to access the disk device from another node in order to maintain availability of the disk device in case of a fault.
Another well-known method for enhancing reliability of disk devices is to multiplex and store a file in a plurality of disk devices (mirroring). A concept of “a logical volume” is generally used for mirroring. Mirroring is a mechanism which shows a plurality of physical disk devices as a single logical volume to the user. The user creates a logical volume in which information on a plurality of physical disk devices is registered beforehand. When the user has accessed the logical volume for a file operation in the same way as to access a physical disk device, file mirroring operation is performed on the plurality of disk devices. By using a logical volume, it is possible to carry out striping, which distributes and stores a file in a plurality of disk devices.
In order to dynamically switching from a current I/O path to another I/O path to access a physical disk device in the conventional UNIX file system when the current I/O path can no longer be used, it is necessary to search file management tables and mount construction information to rewrite each entry of the unavailable I/O path name with the entry of a new one. The above operation to rewrite an entry in each file management table with a new entry must be carried out for each open file. As a result, in a conventional UNIX file system to which the above technique for switching I/O paths is applied, it takes time to rewrite entries in file management tables, causing a problem that it is not possible to perform I/O operation on the target physical disk device during such rewriting time.
Furthermore, if two I/O paths are simply switched when a fault has occurred in one of the I/O paths, the node which was accessing a physical disk device before occurrence of the fault cannot properly write back the contents of caches which the node currently holds, such as a buffer cache (an area in which data is temporarily stored at the time of reading/writing the data from/to a physical disk device in order to reduce the number of input/output operations on the physical disk device, whose processing speed is slow compared with the memory), file management tables, and a disk cache (a cache memory held by a physical disk device for the same purpose as that of the buffer cache) in the physical disk device, to the physical disk device, raising a problem that important data may disappear. Furthermore, since this compromises integrity of the file system, it is necessary to restore the compromised file system to its proper state based on information on the file system redundantly stored in a physical disk device. This restoring operation requires checking of the entire disk device and therefore takes a long time, making it impossible to perform I/O operation on the physical disk device during the restoring operation.
In addition, since, after switching to the new I/O path, the new I/O path is used to access the disk device, it is necessary for the system administrator to update the mount construction file so that a matching relationship between the new I/O path and the mount point of the disk device is registered in the mount construction file, in order to properly perform the mount operation at the time of restarting the system after switching to the new I/O path. Further, in the case where mirroring of files is employed, the system administrator needs to create a logical volume and carry out a complicated procedure for managing the logical volume.
SUMMARY OF THE INVENTION
A first object of the present invention is to provide a file system capable of reducing time taken to switch I/O paths, and hiding as much of the I/O-path switching operation as possible from the general user. A second object of the present invention is to provide a file system capable of switching I/
Ito Akihiro
Kumazaki Hiroyuki
Sonoda Koji
Utsunomiya Naoki
Abel-Jalil Neveen
Hitachi , Ltd.
Rones Charles
Townsend and Townsend / and Crew LLP
LandOfFree
File system for creating switched logical I/O paths for... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with File system for creating switched logical I/O paths for..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and File system for creating switched logical I/O paths for... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3137072