JPH09269855A - Data processing system capable of on-line exchange of module - Google Patents

Abstract

PROBLEM TO BE SOLVED: To securely execute work at ease without adversely affecting a module and a system bus by securely and electrically detaching a part being an exchange object before the module is physically inserted/extracted. SOLUTION: Bus switches 120A and 120B are connected to bus switch control parts 121A and 121B which on/off-change over the switches and a system bus 100. Connection/detachment timing detection parts 122A and 122B detect timing when the other module connected to the system bus 100 does not malfunction even if it receives noise occurring at the time of the electric connection/ detachment of the bus switches 120A and 120B when the bus connection/ detachment request command 123, the voltage abnormality 125 and the clock abnormality of the module 130 occur. Then, effect is transmitted to the switch control parts 121A and 121B. Furthermore, a function for outputting the extraction permission signal 124 of the module 130 after the bus switches 120A and 120B are turned off is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data processing system in which components are multiplexed, and in particular, the removal and replacement of a module when the module fails, and the maintenance and maintenance of the system can be performed without stopping the online operation of the system. The present invention relates to a data processing system capable of exchanging excellent modules online.

[0002]

2. Description of the Related Art In recent years, the degree of dependence on computers has increased with the progress of the information society, and the failure of computers has caused a major obstacle to business operations, and computers have to be stopped in order to expand the functions of the system. The impact when it disappears is increasing, and high reliability and expandability that does not hinder business are required.

Therefore, even if the online operating state is maintained,
There has been a need for an online replacement method that enables module replacement. As a means for realizing online replacement, there is a method in which the output of the three-state buffer circuit is put into a high impedance state and the module is replaced by utilizing a module insertion lever as disclosed in Japanese Patent Laid-Open No. 63-153899.

[0004]

In the above prior art,
In both cases, a special circuit is added to the side of the module to be replaced so as to reduce noise generated during replacement. However, since noise is not protected on the side receiving the noise, when the noise occurs, there is a possibility that the module operating online connected to the same bus receives the noise and malfunctions.

Further, since there is only one connecting / disconnecting circuit, if the connecting / disconnecting circuit itself fails, it is impossible to perform online replacement without generating noise.

An object of the present invention is to reliably electrically disconnect the replacement target portion before physically inserting / removing the module, so that the module and the system bus are not adversely affected, and the online operation can be performed securely and with peace of mind. It is to realize the exchange.

Further, in a fault tolerant computer (hereinafter referred to as FTC), a defective system module replacement work can be performed without adversely affecting the normal system work, and the function as the FTC is prevented from being lost. It is possible to exchange.

[0008]

In order to achieve the above object, the present invention provides means for connecting / disconnecting a connection between a system bus and a module, or an multiplexed electrical connection / disconnection on each multiplexed system bus. In addition, in addition to the module connection / disconnection request, voltage abnormality, and clock abnormality, a timing for not malfunctioning even when receiving noise generated at the time of connection / disconnection and means for detecting a failure of the connection / disconnection section are provided. Thus, even if one of the systems of the connecting / disconnecting unit fails, the module can be replaced without adversely affecting the online system.

Further, by providing means for detecting a failure of the connecting / disconnecting section in addition to the multiplexed electrical connecting / disconnecting means for each of the above-mentioned multiplexed system buses, one of the connecting / disconnecting sections can be provided. When a failure is detected, the other connecting / disconnecting unit forcibly disconnects, and the connecting / disconnecting unit can be replaced without adversely affecting the normal system bus and module.

Since the connecting / disconnecting unit is multiplexed as described above, online replacement is not impossible even if the connecting / disconnecting unit fails during module replacement.

Further, each connection / disconnection unit monitors the connection state, and when it detects that one connection / disconnection unit is out of order, the other connection / disconnection unit forcibly disconnects. / Failure of the disconnection part does not adversely affect the system bus and modules.

[0012]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a block diagram for realizing the online exchange system of the present invention. System bus side connector 1 for connecting the module 130 to the system bus 100
11, a module side connector 131 is provided, and two bus switches 120A and 120B, which are MOS switches, are provided between the connector 111 and the system bus 100.

The bus switches 120A and 120B are connected in series and are also connected to the bus switch control units 121A and 121B for switching on / off and the system bus 100.

Bus connection / disconnection request command 123 of module 130, voltage abnormality 125, clock abnormality 12
6 does not malfunction even if another module (not shown) connected to the system bus 100 receives noise generated when the bus switches 120A and 120B are electrically connected / disconnected. Connection / disconnection timing detection section 122A, 122B having a function of detecting the signal and transmitting it to the bus switch control sections 121A, 121B and a function of outputting the removal permission signal 124 of the module 130 after the bus switch 120A, 120B is turned off.
(Hereinafter, abbreviated as a timing detection unit) is provided.

FIG. 2 is a flowchart for removing the module 130 and FIG. 3 is a flowchart for inserting the module 130 in the above-mentioned structure.

When removing the module 130, the maintenance person first turns on the bus connection / disconnection request command 123 (step 200). When a voltage abnormality occurs, the voltage abnormality signal 125 (step 210) turns on, and when a clock abnormality occurs, the clock abnormality signal 126 (step 220) turns on.

Then, the timing detectors 122A and 122 which have received the bus connection / disconnection request command 123.
B monitors the system bus 100,
Another module (not shown) connected to 0 detects a timing at which the module 130 does not malfunction even if it receives noise generated when the module 130 is electrically disconnected, and transmits it to the bus switch control units 121A and 121B ( Step 230).

Bus switch control units 121A and 121B
Is the bus switch 120A, 120B at the above timing.
Is turned off (step 240). Timing detector 12
2A and 122B turn off the bus switches 120A and 120B and then turn on the removal permission signal 124 (step 2).
50). After confirming that the removal permission signal 124 is turned on, the maintenance worker removes the corresponding module 130 (260).

When inserting the module 130,
With the removal permission signal 124 turned on, the maintenance person first inserts the module 130 as shown in FIG. 3 (step 300), and then turns off the bus connection / disconnection request command 123 (step 310). .

The timing detectors 122A and 122B which have received the turn-off of the bus connection / disconnection request command 123
First turns off the removal permission signal 124 (step 32).
0). Then, even if the system bus 100 is monitored and the bus switches 120A and 120B are electrically connected as in the case of disconnection, another module connected to the system bus 100 occurs when the module 130 is electrically connected. Timing at which malfunction does not occur even if noise is received is detected and transmitted to the bus switch control units 121A and 121B (step 330). The bus switch control units 121A and 121B that received it receive the bus switches 120A and 120B.
B is turned on (step 340).

FIG. 4 shows an example of a timing at which another module does not malfunction even if it receives noise.
The transmitting side outputs data on the system bus 100 at the rising edge of the bus clock, and the receiving side takes in data at the next rising edge of the bus clock.

Next, the connection / disconnection timing detection unit 12
When the bus disconnection request command 123 is issued to
The connection / disconnection timing detection unit 122 turns on the bus switch control signal at the rising timing of the bus clock, turns on the bus switch disconnection signal, and turns off the bus switches 120A and 120B. At that time, noise is generated on the system bus 100.

The connection / disconnection timing detection unit 1
22 turns on the removal permission signal 124 at the next rise of the bus clock after turning on the bus switch control signal.
On the other hand, the operations on the transmitting side and the receiving side are stopped when the bus switch control signal is turned on, and restarted at the next bus clock after the removal permission signal 124 is turned on.

Connection / disconnection timing detection unit 122,
Since the elements having a small propagation delay are used for the bus switch control units 121A and 121B and the bus switches 120A and 120B, the noise has already converged when the extraction permission signal 124 turns on, and the receiving side receives the system bus 100.
You can capture the correct data from above.

Furthermore, connection / switching by one bus switch
When the disconnection operation fails, the bus switch failure signal 150B is supplied to the other bus switch control unit 121A, and the bus switch 120A causes the bus switch control unit 12 to operate.
It is forcibly turned off by 1A. When the other bus switch 120A fails, the bus switch failure signal 15
0A is supplied to the bus switch control unit 121B to forcibly turn off the bus switch 120B.

When in the connected state, either of the connecting / disconnecting bus switches 120A or 120
When B fails, it similarly generates a bus switch failure signal 150A or 150B, and is forcibly disconnected by the other bus switch 120B or 120A.

In this method, as described above, when the maintenance person removes / inserts the module 130, the module 130 has already been electrically disconnected from the system bus 100, and what kind of timing the maintenance person should take. Noise caused by human factors does not adversely affect the data processing system operating online even if it is inserted / removed at a relatively low speed.

Next, A of a multiplexed data processing system having a fault-tolerant function in which hardware is multiplexed so that the system does not stop even in the case of a single failure.
FIG. 5 shows another embodiment of the present invention in which a bus switch is provided in a multiplexed system bus that connects the system and the B system to each other, and online exchange is possible by disconnecting the system bus when exchanging modules.

In FIG. 5, the central processing unit (Cent
ral Processing Unit, abbreviated as CPU hereinafter) 44
The 0A and B system CPUs 440B include not only processors but also memories and the like, and perform the same processing in synchronization with each other via interface signals 452A and 452B between the CPUs.

Then, the processing result is transmitted to the multiplexing bus adapter 441 (Dual System bus Adapter, hereinafter abbreviated as DSBA). The DSBA 441 is divided into a master system and a slave system, and when the A system performs processing as the master system, the DSBA of the A system
Only 441-1A and 441-2A operate, CPU of A system
Data from the 440A is transferred to the system bus 450-1A, 4
A system input / output adapter (Input Output
bus adapter, abbreviated as IOBA hereinafter) 446A and B system IOBA 446B.

At this time, B system DSBAs 441-1B and 44
1-2B is A system CPU 440A or DSBA4
When 41-1A or 441-2A fails, a hot standby state is adopted in which the online processing is immediately taken over so that the online processing is not interrupted.

IOBA 446A and 446B
Accesses the input / output device via input / output adapters (hereinafter abbreviated as IOA) 447A and 447B connected to the respective input / output buses. In addition, power supply (Power Supply,
(Hereinafter abbreviated as PS) 448 has an A system PS 448A and a B system 448B separately, and supplies power to only each system.

In such a multiplexed data processing system, in order to realize online exchange, C of each system is used.
Module removal request signals 451A and 451B from other systems and removal permission signals 454A and 451B from other systems are added to PUs 440A and 440B, and DSBA44
By providing the bus switch 444, the bus switch control unit 460, and the timing detection unit 442 on the system buses 450-1B and 450-2A connecting the mutual systems, among the dual system buses 450 between 1 and IOBA 446, The system data processing system and the system B data processing system are completely separated, and the system in which the module to be replaced is present can be stopped to replace the module.

For example, an A system module IOA447A
When exchanging, the modules DSBA441-2B and IOBA446B connected to the system bus 450-2B receive the ON state of the removal request signal 451A of the A system,
By detecting the timing at which malfunction does not occur even if noise is generated when electrically connecting / disconnecting the bus switches 444-1B and 444-2B, and disconnecting the bus switches 444-1B and 444-2B at that timing. ,
It is possible to replace the module 447A by stopping only the A system while the B system remains in the online operating state. If you disconnect the bus, switch 4
Bus switch 444-2B even if 44-1B fails
Since it is separated by, the multiplexing system will be more reliable than before.

6 and 7 show detailed procedures for replacing the A system module. When removing the module,
First, the maintenance staff turns on the removal request signal 451A to turn on the B system CPU 440B and the multiplexed system bus 450-2.
Request the separation of A (step 500).

The CPU 440B determines whether the disconnection is possible (step 510), and if possible, turns on the bus disconnection signal, which is one of the bus switch control interface signals 453B, to the timing detection section 442B (step 520). . The timing detection unit 442B monitors the multiplexed system bus 450-2B, and the DSBA441
-2B and IOBA446-B are bus switches 444
-1B, 444-2B, the timing at which malfunction does not occur even if noise generated during electrical disconnection is received, and is transmitted to the bus switch control unit 460B (step 53).
0).

Bus switch control units 460-1B and 460
-2B turns off the bus switches 444-1B and 444-2B (step 540). Timing detector 442
B turns on the bus disconnection completion signal of the bus switch control interface signal 453B to turn on the CPU 44.
0B is notified that the disconnection of the system bus 450-2A is completed (step 550).

Receiving this, the CPU 440B turns on the removal permission signal 454A of the A system module (step 5).
60). The maintenance worker who confirms that the removal permission signal 454A is turned on turns off the switch of the A system PS448A (step 57).
0), the target module is removed (step 580).

When inserting the module, first, A
With the system stopped, the maintenance personnel inserts a normal module (step 600) and turns on the A system PS448A switch (step 610). Thereafter, the removal request signal 451A is turned off to request the B system CPU 440B to connect the multiplexing system bus 450-2A (step 620).

The CPU 440B judges whether the connection is possible (step 630), and if possible, first turns off the removal permission signal 454A of the A system module (step 64).
0). After that, the bus disconnection signal of the bus switch control interface signal 453B is turned off to the timing detection unit 442B (step 650). Timing detector 4
42B is a multiplexed system bus 450 operating online.
-Monitoring 2B, DSBA441-2B and IOBA
When the 446-B receives noise generated when the bus switch 460B is electrically connected, it detects a timing at which the bus switch 460B does not malfunction, and notifies the bus switch control unit 460B (step 660) of the bus switch control unit 460B. -1B and 444-2B are turned on (step 67).
0).

FIG. 8 shows a multiplexed bus switch failure detection circuit on the A system side, and FIG. 9 shows a timing chart when the bus switch 444-1A fails. Bus switch 44
When 4-1A and 444-2A are turned on, the comparator CMP-A in the bus switch control unit 460-1A outputs the signal 73 at the point a.
0 and the signal 740 at the point b are compared and the bus switch controller 4
The comparator CMP-B in 60-2A outputs the signal 750 at the point c.
And the signal 760 at point d are compared.

For example, the bus switch controller 460-1A
If the bus switch 444-1A fails in
A mismatch occurs between the signal 730 at point a and the signal 740 at point b,
At the rise of the bus clock 720, the bus switch failure signal 700-1A is output from the flip-flop FF-A in the bus switch control unit 460-1A.

Then, the bus switch control unit 460-2A
Detects the bus switch failure signal 700-1A, the bus disconnection signal 770-2A via the AND-B
Is turned on and the bus switch 444-2A is forcibly disconnected, the DSBA 441-1B and the system bus 45 due to the failure of the bus switches 444-1A and 444-2A.
It does not adversely affect the module on 0-1A.

Similarly, when the bus switch 444-2A fails, the comparator CMP-B and the flip-flop FF-
Bus disconnection signal 770 via B and AND-A
-2B is turned on and the bus switch 444-2B is forcibly disconnected, so that the bus switches 444-1B and 444-2 are connected.
The failure of B does not adversely affect the modules on the DSBA441-1A and the system bus 450-1B.

With the above-described multiplexed bus switch failure detection circuit, it is possible to realize more reliable online exchange without stopping both systems due to the bus switch failure.

Further, the timing detection unit 442B turns off the bus disconnection completion signal of the bus switch control interface signal 453B and notifies the CPU 440B that the connection of the system bus 450-2A is completed (step 680).

The timing detector 442 is a flip-flop which receives the bus disconnection signal and outputs the bus switch control signal at the rising timing of the bus clock of the duplex system bus. Further, the bus switch control unit 460 is configured by a driver that receives the bus switch control signal and outputs the bus switch disconnection signal.

Online data transmitted from the DSBA is, after the delay time d2 of the driver element at the rising timing of the clock of the system bus, as shown in FIG.
It is output to the system bus.

On the other hand, the timing detector which receives the ON input of the bus disconnection signal also turns on the bus switch control signal at the next rising timing of the bus clock, and after the delay time of the driver of the bus switch control unit (see FIG. 10). D3), the bus switch disconnection signal is turned on. Then, when the bus switch is turned off, noise is generated on the system bus.

However, by using the elements having a small propagation delay for the timing detector, the bus switch control section and the bus switch, the time taken for the noise generated on the system bus to disappear can be shortened. For example, if the period of the basic clock is 60 ns, 40n
It becomes possible to disappear within s.

In FIG. 10, the transmission data is <A>.
<B>.

After the data <A> has been sent to the bus,
The bus disconnection signal is turned on, and one cycle later, the bus switch disconnection signal is turned on. Then, after a slight delay, noise is generated, so that noise is eventually generated when the data <B> is present on the system bus.

When the noise convergence time is shortened as described above, the noise has already converged at the timing when IOBA receives the data <B> (note that the data is received at the rising edge of the clock). become. Therefore, IOBA can always receive correct data. In the figure, the timing detector, the bus switch control unit, and the bus switch are shown as separate modules, but they can be mounted in the A-system DSBA even considering their connection form.

In the above description, the A system is taken as an example, but the device configuration and operation of the B system can be considered to be exactly the same.

[0056]

EFFECTS OF THE INVENTION According to the present invention, (1) by multiplexing bus switches and bus switch control units, even if one switch fails, module replacement can be realized by the other switch, and reliability can be improved. A high multiplexing system can be realized.

(2) When the module is replaced, the module to be replaced is already in an electrically disconnected state, so that the module is inserted / removed online at any timing and speed. It does not affect the system at all, and more reliable online replacement can be realized compared to the conventional replacement method that depends on the insertion / removal timing.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a data processing system showing an embodiment of the present invention.

FIG. 2 is a flowchart for removing a module in the configuration of FIG.

FIG. 3 is a flowchart for inserting a module in the configuration of FIG.

FIG. 4 is a diagram illustrating a detection timing of a timing detector.

FIG. 5 is a diagram showing a configuration in a duplicated data processing system of the present invention.

FIG. 6 is a flowchart for removing a module in the configuration of FIG.

FIG. 7 is a flowchart for inserting a module in the configuration of FIG.

FIG. 8 is a diagram illustrating a hardware configuration example of a bus switch failure detection unit.

9 is a time chart illustrating the operation of FIG.

FIG. 10 is a time chart explaining the operation of the timing detector.

[Explanation of symbols]

100 ... System bus, 120 ... Bus switch, 121
... bus switch control unit, 122 ... connection / disconnection timing detection unit, 123 ... bus connection / disconnection request command, 124 ... removal permission signal, 130 ... module.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Eiki Shoko 5-2-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Process Computer Engineering Co., Ltd. (72) Inventor Yoshihiro Miyazaki 5-chome, Omika-cho, Hitachi-shi, Ibaraki No. 1 Incorporated company Hitachi Ltd. Omika Plant (72) Inventor Soichi Takatani 5-21 Omika-cho, Hitachi City, Ibaraki Prefecture Inventor Hitachi Ltd. Omika Plant (72) Naoto Miyazaki Omika Hitachi City, Ibaraki Prefecture 7-1-1, Machi, Hitachi, Ltd., Hitachi Research Laboratory (72) Inventor, Akihisa Nakamura 5-2-1, Omika-cho, Hitachi, Hitachi, Ibaraki (72) Incorporated, Hitachi, Ltd., Omika Plant (72), Masahiro Yoshinuma, Ibaraki 2-5-1, Omika-cho, Hitachi-city, Hitachi, Ltd. Inside the Omika factory, Hitachi Ltd.

Claims (6)

[Claims] 1. A system bus and a data processing system in which a module can be online replaced by inserting / removing the system bus into / from the system bus, which enables electrical connection / disconnection at a connection portion between the system bus and the module. When two sets of bus switches are provided in series, when disconnection operation by one bus switch fails at disconnection, disconnection by the other bus switch, connection operation by one bus switch fails at connection state,
A data processing system capable of online module replacement, comprising a switch control unit forcibly disconnecting a failure system by the other bus switch. 2. The module according to claim 1, wherein when electrically connecting / disconnecting the module to / from the system bus to replace the module, another module connected to the system bus electrically connects to the module. / A timing detection unit is provided to detect a timing that does not cause an erroneous operation even when noise generated when disconnecting is received, and the switch control unit controls the bus switch at the detected timing to perform electrical connection / disconnection. A data processing system that allows online replacement of modules to be performed. 3. A multiplexed data system, wherein one system and the other system are connected by a system bus, and one system can substitute for the other system. In a data processing system capable of online replacement of modules by insertion / removal, two sets of bus switches that enable electrical connection / disconnection at each connection portion of the system bus are provided in series, and one of the bus switches is connected at the time of disconnection. When the disconnection operation by the bus switch fails, the other bus switch disconnects, and in the connected state, when the connection operation by one bus switch fails, the other bus switch forcibly disconnects the failed system. A data processing system capable of online module exchange, which is characterized in that 4. An online system for disconnecting a non-replaceable module from a system bus according to claim 3, wherein each connecting / disconnecting unit is provided with a failure detecting means and the multiplexed bus switches are electrically disconnected when a failure is detected. A replaceable data processing system. 5. The system according to claim 3, wherein each system is provided with a multiplexed failure detection circuit, and when the bus switch failure of one system is detected, the data processing is taken over by the other system.
A data processing system that allows online replacement of modules that forcibly disconnects the failed switch, replaces the failed switch, and restores. 6. The module according to claim 3, wherein when the module is electrically connected / disconnected to / from the system bus, another module connected to the system bus electrically connects / disconnects the module. A module that provides a timing detection unit that detects a timing that does not cause an erroneous operation even when receiving noise that occurs when performing a switch, and the switch control unit controls the bus switch at the detected timing to electrically connect / disconnect the module. A data processing system that allows online exchange.