JP2001184294A - File access method and information processing system using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the efficiency of the data transfer of file between a client and a server. SOLUTION: In a server 2, access to a file and block by access protocols in two levels is made possible, and a mechanism for presenting the correspondence of the file and the block to a client 1 is added. In a client 1, at the time of performing access to the file, a control command (the preparation/deletion/ name change of the file/directory and the attribute reading/writing of the file/ directory) is operated by a file access protocol 52, and data transfer (the reading/writing of file data) is operated by a block access protocol 13. In the latter case, the correspondence of the file and the block is preliminarily requested and preserved to/in the server 2. Thus, the processing load on the client and the service can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information processing apparatus, and more particularly to a method for accessing a storage device of a computer and an information processing system using the same.

[0002]

2. Description of the Related Art A computer has a CPU, a memory and an I / O.
I / O for rapid performance improvement of CPU in recent years
Performance has been more important.

[0003] There are two levels of access methods to storage devices (disks, servers, etc.): block access and file access. Block-level access means input / output by specifying the position of a block, which is the minimum unit of storage device allocation (this unit is also called a "record" or a "sector".
In the present invention, this is called a "block." File-level access means that input / output is performed by specifying a file, which is a related block treated as one unit.

There are communication protocols for each access level of block access and file access. Examples of commonly used protocols for block level access include SCSI and IDE (W. Schwa).
Derer, A .; Wilson Jr. , "Unde
standing I / O Subsystem
s ", Adapttec Press, 1996, ISB
No. 0-96519111-0-9). Normally, in these protocols, data transfer between the main storage (memory) and the storage mechanism is performed by DMA, that is, direct data transfer without passing through the CPU.

An example of a commonly used protocol for file level access is NFS (B. Calllag).
an, B .; Pawlowski, “NFS Vers
ion 3 Protocol Specificat
ion ”, Internet Engineering
Task Force (IETF) RFC1813,
1995) and CIFS (P. Leach, D .;
Naik, “ACommon Internet F
ile System (CIFS / 1.0) Pro
tocol, Preliminary Draft ",
Internet Engineering Task
Force (IETF), draft-leach-
cifs-v1, 1997). Communication of these protocols is usually performed by a network protocol such as RPC or IP (for these protocols, see DE Commercial, "Internet."
rking with TCP / IP ", Vol.
Prentice Hall, 1995, ISBN0
-13-227836-7). Since data transfer using these protocols requires CPU processing, DMA transfer between the memory and the storage mechanism is normally impossible.

[0006] Both file and block access protocols have control command and data transfer portions. As an access performance improvement method, there is an attempt to perform a control command in a low-latency network and to perform data transfer in a high-throughput network (S. Co.).
Leman, R .; Watson, “New Arch
items to Reduce I / O B
ottlenecksin High-Perform
ances Systems ”, Twenty-Six
the Hawaii International C
onence on System Science
ces, 1993, pp. 5-14, hereinafter referred to as Reference Document 1).

[0007]

Problems to be Solved by the Invention Control commands for file access protocol (creation / deletion / rename of file / directory, read / write of file / directory attributes) and data transfer (read / write of file data, etc.) Are not separated, the data transfer passes through protocol layers such as RPC and IP. This processing increases the overhead of the CPU, and thus degrades the storage access performance. And D
The inability of the MA also causes performance degradation. Attempts to perform control commands and data transfer over different networks alleviate this problem but do not solve it.

[0008]

The block access protocol is used as data transfer of the file access protocol.

On the storage device side, in addition to enabling access to both file and block access protocols,
Add a mechanism to provide file / block correspondence to clients.

On the client side, the above correspondence is received,
Based on this correspondence, a file is read or written by the block access protocol.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a storage access method according to the present invention and some embodiments thereof will be described in more detail with reference to the drawings. In the following,
The same reference numbers indicate the same or similar operations. In the second embodiment,
Only the differences from the first embodiment will be mainly described.

<First Embodiment of the Invention> A conventional technique of access using a block access protocol will be described with reference to FIG. First, an example in which the application 10 of the client 1 accesses a file in the disk device 4 will be described. The disk 4 is connected to the host 1 by the adapter 18 and the connection mechanism 3. Examples of the connection mechanism 3 include a SCSI cable and Fiber Channel.
1. Ethernet, ATM network, VIA, internal network of parallel computer, etc.
The connection mechanism 3 is not limited to this example). Since the present invention does not depend on the characteristics of the network, the connection mechanism 3 is not specified in this specification.

When an application 10 accesses a file, it first requests an access operation to a virtual file system layer 11 which is an upper layer of file processing of an operating system (OS) (B. Good).
dhart, J.M. Cox, "The Magic
Garden Explained ”, Prent
ice Hall, 1994, ISBN 0-13
098138-9). This layer 11 checks in which file system the specified file is located. In this example, it is assumed that the file system 12 manages the disk device 4. The file system 12 converts the file access into the block access, that is, the position of the file accessed by the program 10 into the position of the disk block. These block accesses are transferred to the block protocol layer 13 and converted into communication of this protocol. This communication controls the device driver layer 17 controlling the adapter 18 and the adapter 1
8. Arriving at the disk 4 through the connection mechanism 3. Access replies (data and / or operation completion confirmation) pass through the same path in the opposite direction.

As described above, in the case of block access, data transfer between the memory and the adapter is performed without the intervention of the CPU. Next, this operation will be described. The buffer and cache of the local file system 12 are managed by the memory manager 15 of the OS. In FIG. 1, the local file system 12 transmits the memory address of the input / output data to the adapter via the block protocol 13 and the device driver 17, and the adapter writes the read data from the disk 4 directly to this memory area, or writes the data to the disk. Write data is read directly from this memory area. In FIG. 1, this operation is indicated by a dotted line between the adapter 18 and the memory manager 15.

Next, a conventional technique of access using a file access protocol will be described with reference to FIGS.
Here, an example in which the client 1 accesses data existing in the server 2 will be described. The operation is the same as that of the above-described example of the block access protocol, but includes a network file system client 23 and a network protocol 16 layer instead of the local file system 12 and the block protocol in FIG. The network file system client 23 converts an access from the client 10 into a file access protocol. The network protocol 16 puts the communication of the file access protocol on a network protocol such as RPC or IP. The communication of this network protocol is transferred to the server 2.

Next, the operation of the server 2 will be described with reference to FIG. Access received from connection mechanism 3 is adapter 1
8, device driver 17, network protocol 1
6 layers. Then, the network file system server 24 receives this access and converts it into an access to the virtual file system layer 11. Thereafter, the access arrives at the disk 4 through the local file system 12, the block protocol 13, the device driver 17, and the adapter 18, as in the description of FIG. As before, access replies (data and / or operation completion confirmation) pass through the same path in the opposite direction.

Here, the memory manager 15 has the same role as described in the block access protocol. However, in the case of the file access protocol, since the protocol processing of the input / output data to be transferred is necessary, DMA between the adapter and the memory cannot be performed.

The above is the prior art for accessing both blocks and files. Next, an embodiment of the present invention will be described with an example in which the client 1 inputs and outputs data in the server 2 as in the previous description. FIG. 4 shows the client. The client has the same layer 5 as the file system. In this, the protocol selection 51 determines whether to use the block access protocol or the network access protocol for the file access operation. The file system control client 52
In addition to transferring the file access control command to the partner server, a special function obtains a list of block positions of each file corresponding to the file and the block from the server. This list of blocks is stored in the block table 53.

The operation of layer 5 is shown in FIG. When an access is received from the virtual file system 11, it is first checked whether or not the data to be accessed is in the cache (61). If there is, the data is exchanged with the cache (62). If the data is not in the cache, it is determined whether this access is a read or write access (63). In cases other than reading and writing, this access is transferred to the network file system control 52 (64). In the case of reading or writing, the block corresponding to the location of the file to be accessed is the block table 5.
3 (65), and if so, request the block protocol layer 13 to access this block (66). If the corresponding block is not in the list, a request is made to the network file system control 52 to obtain a block table (67).

FIG. 6 shows a block table 53. This table has a host name 71 of the server 2, a file name 72, an identifier 73 of a disk on the server 2, and a pointer 74 to a block information table for each file accessed by the client 1 on the server 2. The block information table has information on the block number 75 of each file and the position 75 on the disk 73 corresponding to this block.

FIG. 7 shows a server according to the first embodiment. This configuration is similar to the conventional server shown in FIG. 3, but the differences will be described below. First, the network file system control server 55 receives an access from the network file system control 52. If this access is a request corresponding to a block, the network file system control server 55 acquires this correspondence from the local file system 12, stores it in the block table 53, and transfers it to the client 1. Control access other than the request corresponding to the block operates in the same manner as in FIG.

FIG. 8 shows the block table 56. This table is
Each client 1 accessing server 2
Table 7 showing the host name 77 of this client, the file 78 accessed by this client, and the block position
It has a pointer 79 to 0. The block position table 70 has the same contents in each file 78 as the block position 76 of the file 78 provided to the client by the server 2 (71 in FIG. 6).

The block transfer server 54 receives a block access from the client 1. This layer refers to the block table 56 in order to avoid a security breach by determining whether the block accessed by the client 1 is within the range corresponding to the previously transferred block. If not, return an error to the client. If it is within the range, the access is requested to the block protocol 13 and a reply of the access is returned to the client 1. In this case, both adapters 18 can perform DMA transfer.

According to the above operation, the first embodiment enables data transfer of the block protocol for file access. As a result, the DMA can be used for data transfer between the client 1 and the server 2, thereby improving the data transfer efficiency. Since the data transfer between the client 1 and the server 2 bypasses the network protocol layer 16, the CPU overhead is reduced. Finally, Embodiment 1 improves the file access performance.

<Embodiment 2> When a client makes block access, the client may be a server instead of the disk 4. In this case, the client 1 communicates with the server 2 on the connection mechanism 3 using a network protocol such as IP, and inputs and outputs data in the server 2. An example of this is IBM's Virtual Sha
red Disks (IBM RISC System)
m / 6000 Scalable POWERpara
llell Systems Administrati
on Guide, Document Number
GC23-3897-01) and Digital / C
Ompaq Petal (E. Lee, C. Th.
ekkath, “Petal: Distribute
d virtual disks ", Proceedi
ngs of ASPLOS, 1996, pp. 8
4-92). As Embodiment 2, application of the present invention to such block access will be described.

First, the prior art of the client 1 will be described with reference to FIG. The operation in this case is the same as the operation described above with reference to FIG. 1, but instead of the block protocol 13, a layer called a network block access client 14 converts the block access of the file system 12 into a block access protocol on the network. Communication of this protocol is transferred to the server 2 through the network protocol layer 16. FIG.
Shows the server 2 of the prior art. Here, the network block access server 19 converts the block access protocol from the client 1 into block access.

The invention will now be described with respect to the prior art described above. FIG. 11 shows the client of the second embodiment.
Shown in This is the same as the client of the first embodiment (FIG. 4), but here, a network block access client 14 is used instead of the block protocol 13. The server is shown in FIG. This is the same as the server of the first embodiment (FIG. 7), but here, a network block transfer server 57 is used instead of the block transfer server 54.

In this embodiment, since the block access protocol uses the network protocol layer 16, DMA transfer between the memory and the adapter 18 is impossible. However, since the processing of the block access protocol is lighter than the processing of the file access protocol,
There is a performance improvement over the prior art.

<Modifications> It is needless to say that the present invention is not limited to the above-described embodiment or its modifications, but can be realized by the following modifications or other modifications. Also, the present invention can be realized by a technique described as a plurality of embodiments or modifications thereof or a combination of the following modifications.

(1) In the above description, an example is shown in which one client, one server, two connection mechanisms, and one disk exist, but there may be a plurality of the present invention. When a plurality of connection mechanisms 3 exist, the method shown in Reference 1 can be used.

(2) When the server 2 manages the data on the disk 4, the disk 4 is directly
The server 2 and the disk 4 can be connected by the connection mechanism 3 as in the other figures. Also in this case, it is possible to connect the disk 4 to the connection mechanism 3 that connects the client 1 and the server 2.

(3) Client 1 and server 2 of the present invention
May have one or more file systems 12 and one or more network file system clients 23 in addition to the layer 5.

(4) In the present invention, the server 2 is shown as a communication partner of the client 1. However, this server is not limited to a computer, and may be a network attach.
edSecure Disks (G. Gibson
et alli. , "A Case for Netw
ork-Attached Secure Disk
s ", Carnegie Mellon University SCS
technical report CMU-CS-9
6-142, September 1996).

(5) In the above embodiment, the server 2 sends the file block position (76, 7) to the client 1.
Send 0). Instead, server 2 uses disk 4
There is also an operation of notifying the position of the block of the corresponding information stored above, and the server 2 and the client 1 reading this information by block access and storing the information in the block tables 53 and 56.

If the computer program for executing the method of the present invention on a computer is stored in a recording medium, the present invention can be executed at any place.

[0036]

As is apparent from the above description, according to the present invention, the processing load on the client and the server can be reduced.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a conventional client using a block access protocol.

FIG. 2 is a configuration diagram of a conventional client using a file access protocol.

FIG. 3 is a configuration diagram of a conventional server using a file access protocol.

FIG. 4 is a configuration diagram of a client according to the first embodiment of the present invention.

FIG. 5 is a flowchart of a file access operation of the client.

FIG. 6 is a diagram showing a block table of the client.

FIG. 7 is a configuration diagram of a server according to the first embodiment of the present invention.

FIG. 8 is a diagram showing a block table of the server.

FIG. 9 is a configuration diagram of a conventional client using a block access protocol on a network.

FIG. 10 is a configuration diagram of a conventional server using a block access protocol on a network.

FIG. 11 is a configuration diagram of a client according to the second embodiment of the present invention.

FIG. 12 is a configuration diagram of a server according to the second embodiment of the present invention.

[Explanation of symbols]

1: client, 2: server, 3: connection mechanism, 4: disk device.

Claims (7)

[Claims] 1. A method for accessing a file stored in each storage mechanism of first and second information processing devices connected by a connection mechanism, comprising: a protocol for designating the file; Using both the protocol for specifying the location of the data stored in the
A file access method for accessing a file stored in a storage mechanism of a second information processing apparatus from the information processing apparatus. 2. The file access method according to claim 1, wherein position information indicating a storage position of the file in the storage mechanism is transmitted from the second information processing device to the first information processing device. 3. A protocol for designating the file or a storage mechanism and a position of data stored therein in accordance with an operation performed on the file using the first information processing apparatus. 3. The file access method according to claim 2, wherein the user selects which protocol to use. 4. The method according to claim 1, wherein the second information processing device is used to
3. The file access method according to claim 2, wherein the position information transmitted from the information processing device to the first information processing device is stored, and it is checked whether or not the access of the first information processing device is within the range of the file. 5. An information processing system for exchanging information between information processing apparatuses by accessing a file stored in each storage mechanism of the first and second information processing apparatuses connected by the connection mechanism. A protocol for specifying a file stored in a storage mechanism of each information processing apparatus, a protocol for specifying the storage mechanism and a position of data stored therein, and the first information using the protocol. An information processing system comprising: means for communicating with a processing device; and means for transferring position information indicating a position where the file is stored in the storage mechanism to the first information processing device. 6. The transfer means includes means for storing the position information, and means for checking whether or not access to the file by the first information processing device is within a range of the file. Item 6. The information processing system according to Item 5. 7. A recording medium on which a computer program for executing the method according to claim 1 on a computer is recorded.