JPH056222B2 - - Google Patents

Description

[Detailed Description of the Invention] [Table of Contents] Overview Industrial Application Fields Conventional Technology Problems to be Solved by the Invention Means for Solving the Problems (Fig. 1) Working Example (a) One Implementation Explanation of an example (Fig. 2) (b) Explanation of another embodiment Effect of the invention [Summary] In an input/output subsystem in which a resource management module controls the operation of an input/output control module in response to input/output processing requests from a management table. , the resource management module and management table are duplicated, and one is operated,
The other side is set to standby, and when the active management table is updated, the standby management table is also updated.
When an active resource management module fails, a standby resource management module can perform input/output processing in place of the faulty module using a standby management table.

[Industrial application field]

The present invention relates to a control method for an input/output subsystem that performs input/output operations in response to requests from a host computer, and in particular a duplex configuration that allows input/output processing to continue in the event of a failure of a resource management module that manages system resources. Concerning the control method of the input/output subsystem.

The input/output subsystem is connected to the host computer and performs input/output processing in response to input/output requests from the host computer.

The input/output subsystem is composed of input/output devices such as magnetic storage devices, optical storage devices, and cache memories, and an input/output control device.

In recent years, such input/output subsystems have been required to have more sophisticated functions as input/output devices have become more sophisticated, host computers have become busier in processing, and computer systems have been required to have more distributed processing.

In line with these demands, the input/output control device is equipped with multiple modules each having a processor.
A link for data exchange has been proposed, which makes it relatively easy to exchange functions and add enhanced functions by exchanging modules. In such input/output subsystems, there is a requirement to avoid malfunctions or failures due to failures as much as possible.

[Conventional technology]

FIG. 3 is an explanatory diagram of the prior art.

The input/output subsystem is composed of an input/output control device 1 (hereinafter referred to as an I/O controller) and input/output devices (hereinafter referred to as I/O devices) such as a magnetic tape device 4a and a magnetic disk device 4b.
An I/O controller 1 is connected to a central processing unit CPU2 via a channel 3.

In the I/O controller 1, channel adapters 10a to 10 serve as an interface with the host.
Device adapters 11a to 11n, which serve as interfaces with I/O devices, are provided as input/output control modules, and a resource management module 12 is provided to manage resources within the I/O devices and controllers. It exists as a module composed of processors.

In addition, the work area of the resource management module and input/output processing requests from the host are handled by the channel adapter (hereinafter referred to as
A management table (hereinafter referred to as TS) 1 that functions as an area for storing information packets (TCB) shared by CA (hereinafter referred to as DA) 10a to 10n and device adapters (hereinafter referred to as DA) 11a to 11n.
3 and these modules 10a to 10n, 11
A connecting link 14 used for exchanging messages between a to 11n and 12 is provided.

By having such a module configuration, it is possible to further transfer modules between I/O devices,
Value-added modules such as a cache memory and a cache module for a cache control section can be easily provided as required.

In such an input/output subsystem, a resource management module (hereinafter referred to as RM) 12 controls operations such as exclusive control of each input/output module 10a to 10n, 11a to 11n, and handles input/output processing requests from the host. , queued by TS13, RM
12, each module independently accesses the TS 13 to proceed with input/output processing.

For example, RM2 queues an input/output request received by CA10a to 10n from the host as TCB to TS13, and then, based on instructions from RM12, CA10
a~10n, DA11a~11n are in TS13
Performs input/output processing by referring to the TCB.

[Problem that the invention seeks to solve]

In such an input/output subsystem,
Since the input/output module is controlled by the queued TCB in TS13 under the instructions of RM12, if a failure occurs in RM12 or TS13, the problem will occur if the subsystem itself goes down. Another problem is that queued input/output processing requests are lost, and the host must reissue the input/output processing requests.

Therefore, the present invention provides an input/output subsystem that can prevent the subsystem from going down and continue input/output processing without losing queued input/output processing requests even if a failure occurs in a management module or management table. The purpose is to provide a control method for

[Means for solving problems]

FIG. 1 is a diagram explaining the principle of the present invention.

In the figure, the same components as those shown in FIG. 3 are indicated by the same symbols, 12a and 12b are resource management modules, one is in operation and the other is on standby, and 13a and 13b are resource management modules. 15 is a switching mechanism, which is connected to both resource management modules 12a and 12b, and is used to manage resources while on standby in the event of a failure in the operating resource management module. This is to switch to a module.

Moreover, each resource management module 12a, 12b is
The configuration is such that when updating the contents of one's own management table, the contents of the other party's management table can also be updated via the copy bus 16.

[Effect]

In the present invention, the resource management module and management table are so-called duplicated, and when an active module or table fails, the switching mechanism 15 switches to the standby module and table, thereby preventing the subsystem from going down. That's what I do.

Therefore, even if a failure occurs not only in an operating module but also in a table, the subsystem can be prevented from going down.

Furthermore, the operating modules are listed in both management tables 1
3a and 13b have been updated to have the same content, so even in highly complex systems where input/output processing requests from the host are queued to a management table, when a running module or table fails, the waiting Using the management table, it becomes possible to execute queued input/output processing requests without re-requesting from the host.

In addition, since both resource management modules and the management table are connected by a copy bus, the contents of both management tables can be updated without using the main link 14, and the control procedure of the link 14 is used to update the contents. The copy operation does not take much time, and there is no delay in controlling the resource management module or the input/output module.

〔Example〕

(a) Description of an Embodiment FIG. 2 is a block diagram of an embodiment of the present invention, showing main parts such as a resource management module and a management table.

In the figure, the same parts as shown in FIGS. 1 and 3 are indicated by the same symbols, and 16 is a copy bus, which includes an address bus, a data bus, and necessary control lines, and a resource management module. (RMa, RMb) 12a, 12b and management tables (TSa, TSb) 13a, 13b are connected to exchange addresses, data, and control signals;
Reference numeral 17 denotes a maintenance module, which is composed of a processor, is connected to RMa 12a and RMb 12b, detects failures in these, performs switching operations, and diagnoses a resource management module in which a failure has occurred; 120 and 122, respectively; Active flags, which are provided in the RMa 12a and RMb 12b, respectively, and are set by the maintenance module (hereinafter referred to as SA) 17 to set the RMa 12a and RMb 12b in operation or on standby;
23 are copy-on flags, and RMa
12a and RMb 12b are provided in 12b to indicate whether copying is possible, and 160 and 161 are AND gates, respectively, and each outputs the TS WRITF signal of the TS write (WRITE) line of the copy bus 16, the active flag 120 or 122, and the copy on flag. 12
1 or 123 and controls a bus gate on the copy bus 16 to be described later; 162;
164 are normally open bus gates, which are normally open and connect the copy bus 16;
A gate that is closed by the gate signal of the AND gate 160 or 161 and disconnects the copy bus 16;
Reference numerals 163 and 165 each designate normally closed bus gates, which are normally closed to disconnect the copy bus 16, and opened by a gate signal from the AND gate 160 or 161 to connect the copy bus.

Therefore, while the gate-on signals are not issued from the AND gates 160 and 161, the copy bus 16
are disconnected by bus gates 163 and 165, and RMa12a is connected to TSa13a and RMb12b.
is connected to the TSb 13b by a copy bus 16. On the other hand, when the gate-on signal is issued from the AND gate 160, the bus gate 163 is opened and the bus gate 162 is closed.
a is connected to TSa13a and TSb13b,
RMb 13b is disconnected from copy bus 16.

Conversely, when the gate-on signal is issued from the AND gate 161, the bus gate 165 opens.
Since bus gate 164 is closed, RMb
13b is connected to TSa13a and TSb13b,
RMa 13a is disconnected from copy bus 16.

Next, the operation of the configuration shown in FIG. 2 will be explained.

First, if the RMa12a is in the operating state and the RMb12b is in the standby state, the SA17 causes the RMa12 to
active flag 120 of a is on, RMb12
The active flag 122 of b is turned off. As a result, the processor of the RMa 12a is instructed to operate, and the processor of the RMb 12b is recognized as having been instructed to stand by.

The RMa 12a that has been instructed to operate turns on its own copy-on flag 121.

Input/output processing requests from host 2 are sent to CA10a~
10n from the link 14 to the RMa 12a. In RMa12a, this is TCB (Task
Control Block), i.e. commands,
Write to the device in pointer format.

That is, the RMa 12a is connected to the copy bus 16.
The TSWRITE line is turned on, thereby generating a gate-on signal from the AND gate 160, opening the bus gate 163, closing the bus gate 162, and sending the TCB onto the copy bus 16. This allows bus gate 16 to be connected to TSa13a.
TCB is written from the copy bus 16 via 4,
to TSb13b via bus gates 164 and 163,
The same TCB is written from the copy bus 16.

In this way, each time the RMa 12a receives an input/output processing request, the TSa 13a,
Write to 13b and queue.

On the other hand, the queued TCB is TSa13
a is read to RMa12, and if the destination I/O device in the TCB is free, link 14
The DA 11a to 11n of the device destination and the CA 10a to 10n that made the request are notified via executed, while notified
CA10a to 10n correspond by referring to TSa TCB
It recognizes the DA, connects to it via link 14, receives read data, and transfers it to channel 3.

In this way, when the processing of CA and DA is completed, they notify the RMa 12a of completion via the link 14, and the RMa 12a is notified that the processing of this TCB is completed.
2a turns on the TS WRITE signal, writes to TSa 13a and TSb 13b as described above, and updates the queue.

On the other hand, if a failure occurs in RMa12a or TSa13a, for example due to parity error detection, RMa1
2a stops. The RMa 12a notifies the SA 17 of this via a hard interrupt.

SA17 thereby turns on the active flag 122 of RMb12b, puts the processor of RMb12b into an operating state, and transfers data from RMa12a to RMb1.
Switch to 2b.

At this time, the ongoing I/O processing will be interrupted and the host will request a retry, but TSb13
The input/output processing request queued in b is
Continued by RMb12b.

In this case, the copy flag 123 of RMb12b
is not turned on, and when updating TSb13b, TSa1
3a is not updated.

The reason why the TSa 13a is not updated is because the SA 17 diagnoses the RMa 12a during this time. That is, the SA 17 diagnoses the RMa 12a while maintaining the contents of the TSa 13a at the time of failure.

As a result of the diagnosis, if the RMa12a becomes usable again, or if the module on the RMa12a side has been replaced and becomes usable, the SA17 will switch to the RMb12b.
of this.

As a result, RMb12b has copy flag 1
23 is turned on, all addresses of the TSb 13b are read, and the read contents are written. Since the TS WRITE signal is turned on during writing, the bus gate 165 opens and the entire contents of TSb13b are written.
Copied to TSa13a.

Then, SA17 turns off the active flag 122 of RMb12b (or
2b may remain active. )RMb12b
enters the standby state again, and the RMa 12a switches to the operating state and executes the same processing as described above.

In this way, the input/output control module
You can proceed with subsequent processing without being aware of the switching between RM and TS.

(b) Description of other embodiments In the above embodiments, updating of TSa and TSb was explained in terms of input/output processing requests. may also be added, or other status information may be updated.

Furthermore, recovery processing can be performed in the same manner for failures in other parts such as ports that are interfaces of inter-module coupling links as failures in RMa.

Although the present invention has been described above with reference to examples, the present invention can be modified in various ways according to the gist of the present invention.
These are not excluded from the present invention.

〔Effect of the invention〕

As explained above, according to the present invention, the following effects are achieved.

In a highly complex input/output subsystem that queues input/output processing requests from the host, the resource management module and management table are duplicated and the contents of the management tables are made to match, so there is no possibility of failure in the resource management module or management table. Even if a problem occurs, the queued input/output processing request can be executed without stopping the system and without re-requesting the host.

Even in this case, since both resource management modules and the management table are connected by a copy bus, the content of both management modules can be updated without using the main link. It can be executed directly without using a procedure, the copy operation does not take much time, the control of the resource management module and the control of the input/output module are not delayed, and high-speed input/output processing is possible.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of the principle of the present invention, FIG. 2 is a configuration diagram of an embodiment of the present invention, and FIG. 3 is an explanatory diagram of the prior art. In the figure, 10a to 10n...channel adapter (input/output control module), 11a to 11n...device adapter (input/output control module), 12,
12a, 12b...resource management module, 13,
13a, 13b...management table, 15...switching mechanism.

Claims (1)

[Claims] 1. A plurality of input/output control modules 10a to 10
n, 11a to 11n and resource management module 12
The resource management module 12 is linked to the management table 1.
3, the input/output control modules 10a to 10n, 11a to 11n
In the input/output subsystem that controls the operation of the resource management module 12 and executes requested input/output processing, a plurality of resource management modules 12 and corresponding management tables 13 are provided, and the plurality of resource management modules 12 and the management table 13 are provided. A copy bus 16 is provided to connect the table 13.
When updating the contents of the management table 13a corresponding to the resource management module 12a in the operating state, the contents of the management table 13b corresponding to the resource management module 12b in the standby state can also be updated; A switching mechanism 15 is provided to switch to the resource management module 12b in a standby state when a failure occurs in the resource management module 12a, and when switching from the resource management module 12a in the operating state to the resource management module 12b in the standby state, 12b
A control method for an input/output subsystem characterized in that the input/output subsystem continues input/output processing requested by the contents of a corresponding management table 13b.