JPH03134756A - Storage area managing system for composite computer system - Google Patents

Abstract

PURPOSE:To prevent the processing of a program from being complicated by constituting the above system so that storage areas of all nodes can be brought to access by the same address format. CONSTITUTION:A memory controller (global storage area managing device) 15 or 25 for dividing an access request into a local storage area managing device 13 or 23 of each node interprets a pair of a well-defined identifier of the node and an address in the node as a global address in a composite computer system, and requests an access to a storage device 12 or 22 with respect to a necessary local storage area managing device 13 or 23. Accordingly, storage areas of all the nodes on the composite computer system can be brought to access by the same address format. In such a way, the processing of a program can be simplified.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a storage area management method for a complex computer system in which a plurality of computer systems are loosely coupled.

(Prior Art) In a compound computer system in which a plurality of computer systems are loosely coupled, each computer system constituting the compound computer system has an accessible storage area limited to the local storage area of the computer system. For information on storage areas belonging to computer systems other than the computer system, it was necessary to perform I10 access using communication means. Therefore, in the past, processing had to be switched between when accessing the local storage area of the own computer system and when accessing the storage area of another computer system.

(Problem to be solved by the invention' The process had to be switched depending on whether the storage area of the computer was accessed or not, leading to complicated processing.

The present invention has been made in view of the above-mentioned circumstances, and is based on the fact that in a program executed on a compound = 1 degree computer system, a memory that is not directly managed by the computer system (hereinafter referred to as a node) on which the program is actually executed is stored. The storage area can also be accessed in the same way as local storage on the node.
The purpose of the present invention is to provide a storage management method for a compound computer system that simplifies program processing and can easily expand storage capacity.

[Structure of the Invention] (Means and Effects for Solving the Problem) The present invention enables each node in a compound computer system to identify the node and exchange information using a unique identifier of each node. A communication device that manages a storage area specific to each node according to an address specified within each node, and a local storage management device that manages a storage area specific to each node according to an address specified within each node, A global storage management device is provided that interprets the address as the address of the node and distributes the access request to the local storage management device of each node. If the corresponding area is accessible within the node to which it belongs, it requests access to the local storage management device of the node to which it belongs, and accesses it within the node to which it belongs. If this is not possible, for the global storage management device of the node indicated by the unique identifier, the extended address consisting of a pair of the requested node's unique identifier and an address within the node is assigned the uniqueness of the node to which it belongs. Equipped with a function that requests access with an identifier and, when receiving the above request from the global storage management device of another node, returns information on the requested address area to the global storage management device of the requesting node. i
With this, in a program executed on a compound computer system, information in any storage area can be accessed using the same access method without being aware of the node where the information exists.

(Example) An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a compound computer system according to an embodiment of the present invention.

In FIG. 1, 10 to 15 are components of computer system A, and 10 is computer system A.
11 is a CPU that controls the computer system A, and 12 is a storage device provided on the local bus 1o. 13 is a memory controller (hereinafter referred to as a local storage management device) that manages and controls the storage device 12 on the local bus 1o; 14 is a local storage management device 1;
3 is a page table that holds information that manages the current storage area that can be accessed directly. Reference numeral 15 denotes a memory controller (hereinafter referred to as global storage management device) that distributes access requests to the local storage management device (13 or 23) of each node. The pair with the address of
3) to request access to the storage device (12 or 22).

20 to 25 are components of computer system B, respectively, and 20 is a local bus of computer system B;
Reference numeral 21 denotes a cPU 122 that controls the computer system B, a storage device provided on the local bus 20, 23 a local storage management device that manages and controls the storage device 22 on the local bus 20, and 24 a local storage management device. 23 is a page table that holds information for managing the current storage area that can be directly accessed.

25 is a global storage management device that distributes access requests to the local storage management device (23 or 13) of each node, and assigns a set of a unique identifier of a node and an address within the node to a global storage management device within the compound computer system. A storage device (22 or 12) for a local storage management device (23 or 13) that interprets and requires an address.
request access to.

Reference numeral 30 denotes a communication path that loosely couples the computer system A and the computer system B as a composite computer system.

FIG. 2 is a flowchart showing the procedure of storage access processing in the multifunction computer system shown in FIG. 1 above. In FIG. 2, S3 is a step for requesting access from another node, and the requesting node's unique identifier is added to the extended address, which is a pair of the node's unique identifier and the address within the node. , requests access to the global storage manager of the node that has the information it needs.

S5 is a step of transferring the page data to the requesting node, and one page worth of data containing data at the requested address is transferred to the requesting node. S6 is a step of executing local paging processing, in which page data corresponding to the address of the own node is transferred to the real memory. S9 is a step of discharging the page data to another node, and if the unique identifier of the node in the extended address of the page to be discharged is another node, the page data is transferred to the original node.

510 is a step of discharging page data to the own node, and if the unique identifier of the node in the extended address of the page to be discharged is the own node, the page data is discharged to its own auxiliary storage device. .

The operation of an embodiment of the present invention will now be described with reference to FIGS. 1 and 2.

Here, a unique identifier of a node and an address within the node are combined as a global address of the compound computer system, and an access request is made using this global address (extended address).

For example, in computer system A, the global storage management device 15 sends a global address that is a pair of the unique identifier of the node and the address within the node to the CPU.
When received from II or the like, it interprets the same address, refers to the page table 14 of its own node, and determines whether the address is currently accessible on its own node (steps S3 and S2 in FIG. 2).

At this time, if access is possible on the own node, the access request is passed as is to the local storage management device 13, and the access is executed under the control of the local storage management device 13 (step S11 in FIG. 2). .

Further, if the n node is currently inaccessible and the unique identifier of the node included in the global address is that of the node itself, normal page fault processing is performed (step S6 in FIG. 2).

Furthermore, if the node is currently inaccessible by its own node and the unique identifier of the node included in the global address belongs to another node, an access request is made to the global storage management device 25 of the node (the Figure 2 Step S
3).

Global storage management device 2 that received this access request
5 requests access to the local storage management device 23 of its own node, and furthermore, the local storage management device 23 that received this access request, if the page for the specified address is accessible, accesses the page. The data is transferred to the global storage management device 15 of the access requesting node (steps S4 and S5 in FIG. 2). At this time, if the page corresponding to the specified address is inaccessible, an access error is determined.

Also, the above page fault processing (step S6 in Figure 2)
, or when the local storage management device 23 performs access (step S5 in FIG. 2), if page ejection is necessary and the page to be ejected is a page of the own node, normal local Execute the page ejection process (steps S7, S8, 5IO in FIG. 2), and
If the page to be ejected is a page of another node, the page data is transferred to the node indicated by the unique identifier in the global address of the page (step S7 in Figure 2).
~S9).

In this way, each storage area within a multi-computer system is accessed using a global address (extended address) that is a pair of a node's unique identifier and an address within the node. The storage areas of all nodes can be accessed using the same address format, allowing programs running on a multicomputer system to access information in any storage area using the same access method without being aware of the node where the information exists. can do.

In the above embodiment, a -identification identifier (for example, a processor ID) is interpreted in the communication path 30 that connects the computer systems constituting the compound computer system, and when the same identifier is an access to another node, The system configuration may be such that the corresponding node is provided with a function of transmitting the global address together with the request source identification identifier to reduce the processing load on the global storage management device. Furthermore, although the above embodiments have exemplified a management method in page units, the present invention is not limited to this, and it is also possible to implement management methods in other data units.

[Effects of the Invention] As detailed above, according to the compound computer system storage management method of the present invention, each computer system constituting the compound computer system can communicate with each other's computer systems using a unique identifier within the compound computer system. A means for uniquely identifying a computer system, a means for accessing the local storage area of the computer system, and a combination of the above-mentioned identifier and the local address used by the above access means = 1 global address within the computer system A configuration in which the computer system can access any storage area in the compound computer system, with a means for accessing storage areas within the compound computer system. By doing this, it is possible to access the storage areas of all nodes on a multi-computer system using the same address format.
In a program executed on a compound computer system, arbitrary storage information can be accessed using the same access method without being aware of the node where the information exists.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention, and FIG. 2 is a flowchart showing a storage area access processing procedure in the multifunction computer system shown in FIG. 1 above. 10, 20...local bus, 11.21...CP
U. 12, 22...Storage device, 13.23...Local storage management device, 14.24...Page table, 1
5.25...Global storage management device, 30...
communication channel.

Claims (1)

[Claims] Each of the computer systems constituting the compound computer system is provided with a means for uniquely identifying each computer system using a unique identifier within the compound computer system, an access means for the local storage area of the computer system, and the unique identifier and the above. and a means for accessing a storage area within the multifunction computer system by using a pair with a local address used by the access means as a global address within the multifunction computer system, so that each computer system within the multifunction computer system can use the above address to access the storage area within the multifunction computer system. A storage management method for a complex computer system characterized by accessing any storage area within the computer system.