JPS60256852A - File recovery system for decentralized processing system - Google Patents

Abstract

PURPOSE:To attain a function-based file copy system suited to a decentralized system by selecting a function module of the copy side equal to a function module among the function modules forming the entire decentralized system. CONSTITUTION:The function modules are classified into three types A, B and C. An A function module of a CPU7 transmits en bloc the copy request annoucement of the A function. This annoucement is received by CPU6, 8 and 10 of other modules of the same function A. Thus the unique CPU numbers allocated to own CPU are transmitted en bloc in reply together with the request reception announcement. In case the CPU8 is selected as the one at the copy side, the start request of the A function module is shunted and preserved until the copy is over. Furthermore, the file contens controlled by a module itself are sent en bloc after conversion into a fixed transmission format. At the side to be copied, the copy file data is received for file recovery. When the copy is over, the A function module executes the start request processing for the shunted A function module and is reset to a steady state.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a file recovery method in a distributed processing system.

[Background of the invention]

One type of file recovery in a multi-system computer system is generally one type of file copying between two computers that implement dual operation.

In Figure 1, 1 and 2 are central processing units (hereinafter referred to as CPU).
That's what I say. ), 3 is a CPU dedicated memory device, 4 is C
This is a memory device dedicated to PU2. 5 is a transmission line for communication between CPUs. Now, copy cput to CPU2
Assuming that the copy destination is the copy target side, when a copy request is issued to CPU 2, the file copy request is output to CPU 1 through the transmission line.

Upon receiving the copy request, the CPUI temporarily pauses all programs that access white files, transfers the file contents and the program's dormant state to the CPU 2 via the transmission line, and releases the program's dormant state after the transfer is completed. The CPU 2 recovers the file contents based on the received data and makes the program suspension state match the CPUI before starting operation. Through the above series of file copy procedures, matching the memory devices 3 and 4 is achieved by
This is a common method.

However, since this method is premised on application to a synchronous dual system, the file configurations of files 3 and 4 must match completely, and it is not compatible with the Load 5hore type, which is a common form of distributed system. In a system configuration that allows a mixture of DUAL types, the file configurations of the memory devices on the copy side and the copy target side are necessarily different, and the above-described one-sided file copy method cannot be adopted.

Furthermore, the above-mentioned copying method has the disadvantage that all programs accessing the memory device are suspended from the start of copying until the end of copying, which can significantly reduce processing efficiency depending on the size of the copy file. There is.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to eliminate the drawbacks of the prior art described above and to provide a single-function file copy system suitable for a distributed processing system.

[Summary of the Invention] The file recovery method according to the present invention provides a file copy function, which will be described later, in each functional module, and when a request for the functional module to become a copy destination occurs,
The present invention is characterized in that a copy-side function module that is the same as the function module is selected from among all the function modules constituting the entire distributed system, thereby making it possible to realize file copy. In addition, regarding other identical functional modules excluding the functional modules on the copy side and the copied side,
It is possible to continue processing normally without stopping the processing, and it is possible to construct a distributed system with excellent processing performance and continuity without deteriorating the processing performance of the functional module as a whole system.

[Embodiments of the invention]

First, the file copy function added to each functional module according to the present invention, its conditions, and method will be described.

File copy 1 for each functional module means that the functional memory 1□ (φ module itself has a means for copying the file area that is essential for the operation of the functional module, and the file area is , so as not to conflict with other function modules in its use.In other words, the function module itself copies the file area it manages, and after the copy is completed, other function modules do not know what state it is in. However, the functional module can normally execute the process.

The outline of the one-sided copy method is to first output a copy request notification from the functional module on the copied side to all other identical functional modules making up the distributed system using a computer-to-computer transmission line. A plurality of copy side functional modules that have received the copy request notification output request acceptance to the copyee side as a response.

The copyee side selects the copy side function module that responded the fastest among the multiple request acceptance notifications (it can be determined that it is the copy side function module with the lightest load), and outputs a copy start request at the same time.

Thereafter, all activation requests for the functional module that occur are saved until the copying is completed.

The only functional module on the copying side that accepts the copy start request outputs the file data to the copied side according to a predetermined format, and at the same time saves the activation request of the functional module until the copying is completed, just like the copied side. . Finally, after outputting a copy completion notification from the copy side functional module, the saved startup request for the functional module is processed and the normal state is returned.

Similarly, after receiving the copy completion notification, the copied function module also processes the saved activation request and returns to the normal state. This completes the series of operations, but other modules with the same function, excluding the copy side and the copied side, can return to a steady state when they receive a copy start request addressed to a source other than their own CPU. Therefore, even during copy processing, other identical functional modules continue processing in a steady state.

As a whole, a single functional module of the distributed system does not stop, and a file recovery method with excellent processing performance and continuity can be provided.

As an embodiment of the present invention, an application example to a distributed system using five CPUs will be taken, and the effects and features of the present invention will be explained in detail.

In Fig. 2, 6, 7.8, 9, and 10 are multi-distributed CPUs, 11, 12 degrees, 13, 14.1
5 is a dedicated memory connected to each of the above-mentioned CPU'Us, and 16 is a Broad Ca5t transmission line for communication between the CPUs. Now, the types of function modules are A, B, and C.
Assuming that the types are the same, the functional module configuration of each CPU is as shown in Figure 2, with functional modules A and B for CPU6, A and C for CPU7, A and c for CPU8, B for PN2, and
Let us consider a case where the C-CPU 1 (N: A, B, and C functional modules are distributed and the CPU '7 returns to the online system from a stopped state.

'Here, functional modules A, B, and C are not related to each other and have completely independent functions.
It receives and processes only the data it needs from among the received data.

Now, if we consider the case where the CPU 7 is brought up from a stopped state to an operating state, the CPU 7 independently determines that a file recovery is necessary at the timing of returning online from the initial state when the functional modules A and C return online. ) and transmission line 16.
To send a copy request notification using Broad Ca5t (
– actual transmission).

For the sake of explanation, here, we will focus on the He function module and explain its operation in detail.

Figures 3.4, 5, and 6.7 show the transition states of a series of file copy procedures.

First, the function module of the CPU 7 broadcasts a copy request notification for the function (■), and the copy request notification is received by all other identical function modules that make up the system (FIG. 3). In other words.

CPU6. CPU8. Each A function module of CPUI O will receive it. Each A function module of CPUs 6, 8 and 10 sends a request acceptance notification to their own CPU as a response.
Bro
ad Ca5t (Fig. 4 ■). Unique C here
The PU number can be interpreted as a logical number roughly assigned to each CPU.

CPU7 (copied side) determines that the CPtJ number of the request acceptance notification received first is the only copying side function module CPU, and sends the copying side C to the copy start request.
Attach a PU number and perform Broadcast (■ in Figure 5), and save and save A function module startup requests that will occur thereafter until the copy is completed.

(Figure 5 shows the case where CPU 8 is selected as the CPU on the copying side.) Function module A on the copying side, which is instructed by the copy start request, requests activation of the functional module on the copying side as well. Save the file until the copy is completed, and then convert the contents of the file you manage into a predetermined transmission format and broadcast it (■). The copy target side receives the copy file data and performs file recovery (FIG. 6).

After the copying is completed, the functional module outputs a copying completion notification to the copying side (2), executes activation request processing for the saved functional module A, and returns to the steady state. Similarly, when the copied side receives the copy completion notification, it executes the activation request process of the saved functional module and enters a steady state (Figure 7). The module looks at the CPU number of the copy start request after the request acceptance response, and
It recognizes that a functional module other than itself is the copying CPU and immediately returns to a steady state. In addition, if there is no request acceptance notification within a certain period of time in response to the copy request notification from the copyee side, the function module will independently
This means that it is in the initial state, so it is determined that copying is not necessary. Furthermore, if the same functional module *@ on the copied side issues a copy request notification and conflicts with the functional module on the copying side, the functional module on the copying side performs first-come-first-served processing for the copy start request. For other copy start requests, NG with the CPU number of the copied side added.
Output a notification. The copy target side performs retry processing in response to the NG notification and outputs a copy start request again. This completes the series of copy operations, but as expected,
However, it goes without saying that files can be recovered using the same procedure for C function modules as well.

Also, while the file recovery of the function module (in this example, function A) is in progress, other function modules can operate completely independently, and since the copy CPU is not physically fixed, the function is not fixed. Expansion and deletion can be easily realized.

〔Effect of the invention〕

The file recovery method according to the present invention can arbitrarily select the same functional module on the copy side as the copy requesting functional module on the side to be copied, and perform file copying.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of a conventional file recovery method, Fig. 2 is a diagram illustrating the present invention in detail, taking a distributed system using five CPUs as an example, and Figs. 3 to 7
The figure is a state transition diagram for explaining one type of file recovery according to the present invention. ) 1.2.6~IO...CPU, 3,4 river memory. Agent Patent Attorney Akio Takahashi y No Non J Ka 3/j;7J 77 June

Claims (1)

[Claims] 1. In a multi-system computer system that has multiple computers constituting the distributed processing system and a transmission path for communication between the computers, the software structure of each computer is divided into functional programs that can be managed in a distributed manner, and the A one-sided functional file recovery method in which a file recovery function is provided for each program, and the functional program that requires recovery requests internal recovery of all functional modules that make up the distributed processing system.6