JP3572641B2 - File server selection system - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a technology for selecting a file system of a computer system, particularly a network file system accessed via a network, in which computer, that is, which computer is used as a file server. The present invention relates to a system for selecting a more appropriate file server according to the usage status of each file.
[0002]
[Prior art]
As network file systems for workstations, for example, NFS from Sun Micro Systems, or AFS from Transarc, etc., have become widespread. The NFS stores a file entity (data itself constituting a file) in hardware called a file server, that is, a large-capacity magnetic disk device of a computer, etc., and a plurality of client workstations each store a necessary data portion. The system is designed to share file data by making copies accessible via the LAN.
[0003]
In such a method, the file entity is always located on the file server, and the file entity does not move to other hardware in the network although the file size may increase or decrease. However, when a file located in a file server is accessed from a workstation via an inter-network connection device called a router or a bridge even within the same premises, or remotely via a high area network (WAN) such as AFS. When accessing from the ground, the access time may be significantly increased and may not be able to be accommodated within the time limit requested on the same LAN.
[0004]
[Problems to be solved by the invention]
Conventionally, especially in AFS, when necessary, all or part of a file is transferred in advance to a hard disk (local disk) serving as a storage medium of a client workstation, and the local disk functions as a cache memory to transfer data. The method used is adopted. According to this method, caching a file on the local disk itself takes a certain amount of time, so that the user is forced to wait and some kind of resignation has actually occurred.
In any case, however, file access via a network has a disadvantage in that the file remains on the file server where the file was originally generated, considering the time and economic costs of accessing the network. Big.
[0005]
The present invention has been made in view of such circumstances, and an object of the present invention is to provide a file server selection system that can optimally select a file server in terms of access time from each computer, cost, and the like.
[0006]
[Means for Solving the Problems]
A file server selection system according to the present invention is a system for selecting any one of a plurality of computers connected via a network as a file server in which a file entity is to be located. Is a file server, for each file, an access frequency measurement for specifying a partner computer that has accessed each of the files and measuring a logical distance and the number of accesses to the specified partner computer Multiplying the logical distance between itself and the number of accesses for each of the other computers based on the information measured by the access frequency measuring means or the information measured by the access frequency measuring means of another computer. Access frequency, and add And access frequency recording means for obtaining a Seth frequency index, by comparing accordance themselves and access frequency recording predetermined algorithm access frequency index obtained by means of the other computers, the comparison result of a plurality of computers are small It has towards a file entity position determining means for determining as a file server object file should position, and a file transferring means for transferring the file entities is determined as a new file server computer.
[0007]
[Action]
In the file server selection system of the present invention, the computer serving as the file server measures the access frequency to the file in which the file entity is located by its own access frequency measuring means, and calculates the access frequency index by its own access frequency recording means. The computer that is not the file server receives the information measured by the access frequency measuring unit of the computer that is the file server, and multiplies the logical distance between itself and the number of accesses by its own access frequency recording unit. look, child compared according further determine the access frequency index obtained by adding them with own access frequency recording means, a predetermined algorithm access frequency index obtained by the access frequency recording means of each computer itself and another computer The comparison result of a plurality of computers to determine a computer to serve as a new file server by file entity position determining means of each computer as a file server to be positioned for the files towards a small file server until it Transfer the file entity to the computer newly determined as the file server by its own file transfer means.
[0008]
【Example】
Hereinafter, the present invention will be described in detail with reference to the drawings showing the embodiments.
FIG. 5 is a schematic diagram showing an example of a hardware environment to which the file server selection system according to the present invention is applied.
In FIG. 5, reference numerals L1 to L5 denote LAN (Local Area Network), FS1 and FS2 denote computers that can be file servers, H1 to H6 denote host computers, R1 to R5 denote routers, and W1 denote routers. 1 shows a WAN (Wide Area Network).
[0009]
The host computers H1, H2, and H3 are connected to the LAN L1, the host computer H5 is connected to the LAN L3, the host computer H4 and a computer FS2 that can be a file server are connected to the LAN L4, and the host computer H6 is connected to the LAN L5. The computer FS2 which can be a file server is connected to the WAN W1.
The LAN L1 is connected to the LAN L2 via the router R1, the LAN L3 is connected to the LAN L2 via the router R2, the LAN L4 is connected to the LAN L2 via the router R3, and the LAN L5 is connected to the router R5. It is connected to the WAN W1 via the network.
[0010]
The file to be operated by the file server selection system according to the present invention is file A, the computer on which the substance of file A is currently located is file server FS2, and the computer on which it can be located is a file Server FS2. The file server FS2 where the files are actually located is also called a home site.
In FIG. 5, it is assumed that the file A is located on the file server FS2. That is, the file server FS2 is the current home site.
[0011]
The computers that need to access the file A and are accessible are called host computers. In FIG. 5, six computers H1 to H6 are arranged as described above. Although some computers do not access the file A, they are not shown in FIG.
[0012]
FIG. 1 is a block diagram showing a configuration of a main part of the file servers FS1 and FS2. The file servers FS1 and FS2 are basically computer systems having a large-scale storage device such as a magnetic disk.
Each of the file servers FS1 and FS2 includes a file system management unit 11, an access frequency management table 12, an access frequency measurement unit 13, a file home site determination unit 14, a file transfer unit 15, a communication unit 16, and the like. Although not shown in FIG. 1, a large-capacity storage device such as a hard disk is provided for its original function as a file server.
[0013]
The file system management unit 11 manages the entire file servers FS1 and FS2.
The access frequency measurement unit 13 functions as an access frequency measurement unit. For each target file, information specifying the host computer H1 to H6 that accessed the target file and a logical distance between the host computer H1 to H6 (the A value obtained by multiplying the determined number of LANs and WANs by a predetermined coefficient) and the access frequency (number of times) within a unit time are measured.
However, the logical distance here is replaced by a packet hop count as used in, for example, a TPC / IP communication protocol.
Note that the packet hop count is a mechanism for adding a count every time a packet passes through a router that is an inter-network relay device, and deleting packets that are unnecessarily passing through the network.
[0014]
When the host computer accesses the file A via a WAN (high area network) having a lower transmission speed than the LAN, the access frequency measurement unit 13 uses the LAN to express the cost. A high coefficient that is different from the logical distance in the case of access via is used.
In this embodiment, the coefficient of the logical distance when access is performed via one LAN is set to “1”, and the coefficient of the logical distance when access is performed via one WAN is “1”. "5" is set.
[0015]
The access frequency management table 12 functions as an access frequency recording unit, and records a result of measurement performed by the access frequency measurement unit 13 in a unit time as described above.
FIG. 2 is a schematic diagram showing an example of the recorded contents of the access frequency management table 12. As shown in FIG. 2, in each of the file servers FS1 and FS2, with respect to all the accessed host computers, information specifying the host computers, a logical distance between the host computers, Are recorded in the respective access frequency management tables 12.
Thereafter, the logical distance of the access frequency management table 12 is multiplied by the access frequency to calculate an index of the access frequency for each host computer, and finally the overall access frequency index is obtained.
[0016]
The file home site determining unit 14 functions as a file entity position determining unit, and determines whether or not to change the home site of the file. The details will be described later.
The file transfer unit 15 functions as a file transfer unit. When the home site is changed to another file server, the file transfer unit 15 transfers the file from the file server provided with the file server to the new home site that has become the new home site. Control the transfer.
The communication unit 16 controls various communications with other file servers.
[0017]
The determination of whether or not to change the home site, which is performed by the file home site determination unit 14, is performed as follows.
When the access frequency index is obtained in the access frequency management table 12 of the file server FS2, which is the current home site, as described above, information for identifying the partner host computer, the logical distance between each host computer, The number of times of access from the host computer is transferred by the communication unit 16 to another file server (the file server FS1 in the example of FIG. 5) via the LAN and the WAN, and the data of the access frequency management table 12 is transferred.
[0018]
The file server FS1, which is not the current home site, receives this data from the file server FS2, which is the current home site, and calculates an index of the access frequency for each host computer and an overall access frequency index. Then, the access frequency index is transmitted between the two file servers FS1 and FS2 by the communication unit 16 and the file home site determination unit 14 compares the two, and determines the one with the smallest access cost as the next home site.
[0019]
When the home site change is determined by the file home site determination unit 14 as described above, the file transfer unit 15 transfers the file entity.
During the transfer of the file entity, the file service is continued from the current home site FS2 to each of the host computers H1 to H6 so as not to be interrupted.
After the transfer of the file entity is completed, the host computers H1 to H6 are notified of the change from the file server FS2 of the home site of the file A to the file server FS1. When an access occurs, the internal information is changed so as to access the file server FS1, which is a new home site.
[0020]
It should be noted that, in order not to be affected by the change of the physical position of the file, all of them are described in advance by the logical position algorithm name or the like, and the correspondence between the physical and the logical is rewritten by moving the file entity, so that it is easily realized. It is possible.
[0021]
Further, as a premise of file use, it is assumed that if the application to be used is the same, it is possible to predict the use frequency in the near future from the access frequency.
[0022]
As described above, by changing the home site of a file, the time required to access the file can be shortened and a pay communication network (such as a high-area network) can be shortened, as compared with the case where the host computer keeps accessing the previous home site. If so, it will be possible to reduce the communication fee.
[0023]
Hereinafter, the actual procedure will be specifically described.
In FIG. 5, it is assumed that the file A is located on the file server FS2. That is, the file server FS2 is the current home site.
The number of accesses from each of the host computers H1 to H6 to the file A located on the file server FS2 was measured, for example, over one hour, and the number was recorded in the file server FS2. However, when accessing the file A from each of the host computers H1 to H6, the coefficient “1” of the logical distance is counted every time the file A passes through any one of the LANs L1 to L5. Therefore, the coefficient "5" of the logical distance is counted each time the signal passes once.
[0024]
The numerical value of the coefficient “1” for the LAN and the coefficient “5” for the WAN can be appropriately set by weighting based on the home site determination strategy of the file.
[0025]
Here, it is assumed that the measurement result observed in the storage mechanism of the file server FS2 is a result as shown in FIG.
That is, since the host computer H1 passes through the LANs L1 and L2 and the WAN W1, the logical distance between the two LANs L1 and L2 is "2" and the logical distance is "5" for the WAN W1. Assuming that the access frequency is “7” and the access frequency is “10”, the index is “70”.
[0026]
The logical distance is "7" from the host computer H2 via the LANs L1, L2 and WAN W1, and the index is "140" since the access frequency is "20".
The logical distance is "7" from the host computer H3 through the LANs L1, L2 and WAN W1, and the index is "70" because the access frequency is "10".
The logical distance is "7" from the host computer H4 via the LANs L4, L2 and WAN W1, and the index is "210" because the access frequency is "30".
The logical distance is "7" from the host computer H5 via the LANs L3, L2 and WAN W1, and the index is "210" because the access frequency is "30".
The logical distance from the host computer H6 is "3" because it passes through the LAN L5 and WAN W1, and the index is "30" because the access frequency is "10".
From the above, the final access frequency index is calculated as “730”.
[0027]
Next, the file server FS2 sends the contents of the access frequency management table to the computer that can be a home site for the file A, that is, the file server FS1 in the example shown in FIG. Access frequency information).
The file server FS1 obtains an access frequency index by the same processing as that performed by the file FS2 based on the information sent from the file server FS2.
Specifically, it is as follows.
[0028]
The logical distance is "3" from the host computer H1 via the LANs L1, L2 and L4, and the index is "30" because the access frequency is "10".
The logical distance is "3" from the host computer H2 via the LANs L1, L2 and L4, and the index is "60" because the access frequency is "20".
The logical distance is "3" from the host computer H3 via the LANs L1, L2 and L4, and the index is "30" because the access frequency is "10".
The logical distance is "1" from the host computer H4 via the LAN L4, and the index is "30" because the access frequency is "30".
The logical distance is "3" from the host computer H5 via the LANs L3, L2 and L4, and the index is "90" because the access frequency is "30".
Since the host computer H6 passes through LAN L6, WAN W1, LAN L2 and LAN L4, the logical distance is "8", and the access frequency is "10", so the index is "80".
From the above, the final access frequency index when the file A is located in the file server FS1 is calculated as “320”.
[0029]
The two file servers FS1 and FS2 bring the index and compare them, and determine the file server FS1 having the lower cost as the new home site of the file A. In this case, only the access frequency index obtained as described above may be used as a guide, or weighting for the logical distance between the file servers FS1 and FS2 and the host computers H1 to H6, or comparison logic, for example, when the index difference is " Since there are various determination methods such as only in the case of 500 "or more, the user may appropriately select.
[0030]
After determining the new home site of the file A as the file server FS1 as described above, the file server FS2 transfers the file A to the file server FS1. During the transfer of the file A, each of the host computers H1 to H6 continues to access the file server FS2.
[0031]
When the transfer of the file A from the file server FS2 to the file server FS1 is completed, the file server FS1 notifies all the host computers H1 to H6 that the file A is the home site of the file A. In response, each of the host computers H1 to H6 changes the access of the file A in its own system to the file server FS1.
[0032]
In the above embodiment, an example in which two file servers are connected to the network has been described. However, it goes without saying that any number of file servers may be used as long as the number is two or more.
[0033]
【The invention's effect】
As described in detail above, the conventional method actively evaluates and relocates unnecessary access costs that are not corrected if access costs from each host to the file server are not used without being evaluated. Can be omitted.
In recent years, the network configuration has become complicated, and optimization can be achieved by consciously measuring access costs that are difficult to grasp intuitively.
In the LAN, the real-time property (simultaneity) of file access is impaired every time a router is passed, and in the case of a WAN, not only that, but a charged communication fee is charged. The effectiveness is clear.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a main part of a file server.
FIG. 2 is a schematic diagram showing an example of recorded contents of an access frequency management table provided in a file server.
FIG. 3 is a schematic diagram showing a specific example of recorded contents of an access frequency management table provided in a file server.
FIG. 4 is a schematic diagram showing a specific example of recorded contents of an access frequency management table provided in a file server.
FIG. 5 is a schematic diagram showing an example of a hardware environment to which a file server selection system according to the present invention is applied.
[Explanation of symbols]
12 access frequency management table 13 access frequency measurement unit 14 file home site determination unit 15 file transfer unit

Claims (1)

In a file server selection system for selecting any of a plurality of computers connected via a network as a file server where a file entity should be located,
Each of the plurality of computers,
When the file server itself is used, for each file, an access frequency that specifies a partner computer that has accessed each file, and measures a logical distance to the specified partner computer and the number of accesses. Measuring means;
Based on the information measured by the access frequency measurement means or the information measured by the access frequency measurement means of another computer, an access is performed by multiplying the logical distance between itself and the number of access times for each of the other computers. Access frequency recording means for determining the frequency, and further obtaining an access frequency index obtained by adding them;
By comparing the access frequency index obtained by the access frequency recording means of the own computer and another computer according to a predetermined algorithm , a file server in which the comparison result among the plurality of computers is smaller should be located. File entity position determining means for determining as
File transfer means for transferring a file entity to a computer determined as a new file server,
The computer serving as the file server measures the access frequency to the file in which the file entity is located by its own access frequency measuring means, and obtains the access frequency index by its own access frequency recording means,
The computer which is not the file server receives the information measured by the access frequency measuring means of the computer which is the file server, and obtains the access frequency obtained by multiplying the logical distance between itself and the number of accesses by its own access frequency recording means. , And an access frequency index obtained by adding them by its own access frequency recording means,
The computer to be the new file server is determined by the file entity position determining means of each computer,
A file server selection system, wherein a computer which has been a file server until then transfers a file entity to a computer newly determined as a file server by its own file transfer means.

Images