JP2002149622A - Information processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information processing system which reduces burden on a CPU and remarkably improves communication performance between clusters by carrying out communication between the clusters continuously. SOLUTION: The information processing system for communication between clusters is provided with management tables 46-1 to 46-n obtained by adding a timer value for monitoring time to a cluster number and a communication instruction unit, and timer count control parts 43-1 to 43-n for updating all the timer values of the tables 46-1 to 46-n and performs time monitoring of the communication between consecutive clusters by each cluster number and communication instruction.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an information processing system, and more particularly to an inter-cluster communication command control method in an information processing system for performing inter-cluster communication.

[0002]

2. Description of the Related Art Conventionally, as an information processing system, FIG.
, CPUs (central processing units) 2-1 to 2-n
And a storage unit 3-1 to 3-n that stores communication information such as communication between the devices.
There is a system in which a plurality of clusters are provided, and these clusters (# 1 to #n) 5-1 to 5-n are connected by a transmission line 100.

In the above-mentioned system, these clusters 5-1 to 5-n are communication command controllers 6-1 to 6 connected to CPUs 2-1 to 2-n and storage devices 3-1 to 3-n.
-N, the communication command control units 6-1 to 6-n have timer units 61-1 to 61-n and communication management units 62-1 to 6-n.
62-n and response monitoring units 63-1 to 63-n are provided.

The timer sections 61-1 to 61-n can set an arbitrary monitoring time, and generate a timer interrupt when the monitoring time elapses. Communication management units 62-1 to 62-n
Starts the inter-cluster communication process in accordance with the instructions of the CPUs 2-1 to 2-n, starts the timer units 61-1 to 61-n in order to perform time monitoring assuming that there is no response inter-cluster communication, Time-out processing is performed by a timer interrupt from the units 61-1 to 61-n.

When the response inter-cluster communication is returned from the transmission destination in the inter-cluster communication, the response monitoring units 63-1 to 63-n identify the normal end or the abnormal end based on the response data and identify the CPUs 2-1 to CPU-2. -N and the transmission path 10
When an error occurs in 0 and there is no response inter-cluster communication and a timeout occurs, retry processing is performed and inter-cluster communication is executed again, and the result is notified to the CPUs 2-1 to 2-n.

The processing operation of the above system will be described with reference to FIG. When an inter-cluster communication process is requested from the CPU 2-1 to the cluster # 2, the communication command control unit 6
-1 is started, and the timer unit 61-1 is started to perform time monitoring assuming that there is no response cluster communication from the cluster # 2. At this time, the timer unit 61-1 starts time monitoring at an arbitrary time.

[0007] Thereafter, the communication management unit 62-1 sets the cluster #
2 and perform cluster communication. When the cluster # 2 receives the inter-cluster communication from the cluster # 1 and ends, the cluster # 2 performs the inter-cluster processing for the cluster # 1, and sends the response inter-cluster communication to the cluster # 1. At this time, the cluster # 1 performs time monitoring until the response inter-cluster communication comes. Therefore, unless there is a response inter-cluster communication from the cluster # 2, the inter-cluster communication to the other clusters # 3 to #n is not performed. Can't do it.

This is a section in which the subsequent inter-cluster communication shown in FIG. 7 cannot be executed. The above system is based on the premise that it does not process a plurality of inter-cluster communication commands at the same time, and thus adopts such an inter-cluster communication command control method.

In the inter-cluster communication command control method as described above, there is also a method of providing a management table 7 as shown in FIG. 8 so that monitoring can be performed on a command basis. FIG.
In the management table 7, a cluster number 71, a communication command 72, and a data section 73 are stored in correspondence with each other.

As a conventional inter-cluster communication command control method, there is a technique (first technique) disclosed in Japanese Patent Application Laid-Open No. 7-262110. The first technique is based on the premise that a plurality of inter-cluster communication commands are not processed at the same time, and is a 1: 1 inter-cluster communication command control method in which a transmission-side cluster and a reception-side cluster are processed.

The interruption of the above-mentioned communication command processing is detected when a communication failure occurs in the storage controller or when a reception wait timeout occurs. This is a method in which communication between other clusters is not possible until the inter-cluster communication is transmitted to one cluster and a response inter-cluster communication is returned or an inter-cluster communication is interrupted due to an abnormality in a transmission path.

Further, as a conventional inter-cluster communication command control method, there is a technique (second technique) disclosed in Japanese Patent Application Laid-Open No. Hei 8-190536. In the second technique, the configuration includes a plurality of system storage devices for controlling data transfer between clusters.

The system has a plurality of system storage devices for controlling the transmission line 100. When an abnormality occurs in the transmission line 100, the system storage device is dynamically switched to dynamically change the system storage device. In this case, the dynamic change is instructed by the CPU, and another system storage device used for the inter-cluster communication command is set. Also,
CP when retrying inter-cluster communication command
It is executed according to the instruction of U.

Further, as a conventional inter-cluster communication command control method, there is a technique (third technique) disclosed in Japanese Patent Application Laid-Open No. 9-101921. According to the third technique, when an abnormality occurs in the transmission line, data transmission and reception is momentarily stopped, and when the abnormality in the transmission line 100 is recovered, data communication can be continued without performing an operator operation operation. Like that. The third technique aims at providing an information processing apparatus capable of reducing the load on the EPU.

In the third technique, a timer is started for a fixed time,
Retry is repeatedly performed while the timer is running, and transmission / reception processing is stopped until the transmission line 100 is restored or the time is over, thereby instantaneously stopping data transmission / reception.

On the other hand, as a conventional inter-cluster communication command control method, there is a technique (fourth technique) disclosed in Japanese Patent Application Laid-Open No. 8-179969. In the fourth technique, a management table is provided so that monitoring can be performed for each instruction. In this management table, a job name, a notification condition, notification information, and a notification destination are registered. In this method, a method of notifying a failure when a job (business) running on the system has failed is managed by the management table.

As a conventional inter-cluster communication command control method, there is a technique (fifth technique) disclosed in Japanese Patent Laid-Open No. Hei 4-301945. In the fifth technique, multicast communication is performed in a computer system connected by a communication network that cannot perform multicast communication. The management table is used for confirming the destination when instructing data transmission and receiving the end report.

[0018]

In the above-mentioned conventional inter-cluster communication command control system, in the case of the first technique, in a system in which a plurality of clusters are connected by a transmission line,
When the number of connected clusters increases, asynchronous inter-cluster communication is frequently performed for a plurality of clusters in a logically divided configuration or the like.

With this configuration, it is impossible that there is no communication between a plurality of clusters. This can be an inter-cluster communication where the transmitting cluster and the receiving cluster are 1: n. In addition, since the response of the inter-cluster communication is extremely slow due to the performance of the transmission path and the like, the transmitting side sets the waiting time in consideration of the performance of the transmission path. At this time, when inter-cluster communication is sequentially performed on two or more clusters, if the previous inter-cluster communication is not completed, the subsequent inter-cluster communication is waited.

In this case, when the inter-cluster communication is continuously issued to a plurality of clusters, the subsequent inter-cluster communication is suspended. Therefore, even if no abnormality is detected in the transmission path, the cluster communication processing is suspended and slowed down, the inter-cluster communication performance is reduced, and the processing performance of the system is likely to be affected.

In the case of the second technique, when the system storage device is dynamically changed or when retry processing of inter-cluster communication is performed, the processing is executed by starting the CPU. This causes a problem that the load on the CPU increases each time an abnormality occurs in the inter-cluster communication processing.

In the case of the third technique, a timer is started for a fixed time, and while the timer is running, retry is repeatedly performed to restore the transmission path or stop transmission / reception until the time is over. , The transmission and reception of data is momentarily stopped. In this case, as in the first technique described above, if the inter-cluster communication is continuously issued, the inter-cluster communication is suspended every time the inter-cluster communication is issued, and the cluster communication processing is slowed down, thereby affecting the processing performance of the system. The problem occurs.

In the case of the fourth technique, the table is managed only when a failure of a job (business) is referred to when a failure notification method is referred to. Because there is no function, it is not possible to monitor the time in fine instruction units.

In the case of the fifth technique, a multicast communication is issued to a plurality of devices, but the communication is not terminated unless all the completion response notifications are completed or a timeout occurs in time monitoring. This is efficient when the same communication is performed for a plurality of devices, but if one device attempts to perform different communication, the efficiency is reduced. Further, it is not possible to quickly execute different communication continuously.

Therefore, an object of the present invention is to solve the above-mentioned problems and to execute inter-cluster communication continuously,
An object of the present invention is to provide an information processing system capable of reducing a load on a CPU and dramatically improving inter-cluster communication performance.

Another object of the present invention is to provide a method for easily changing the setting when the timeout time needs to be set in accordance with the priority of a communication command due to a problem in system operation. It is to provide a processing system.

[0027]

An information processing system according to the present invention comprises a central processing unit and a storage unit for storing communication information such as communication between the units, wherein the constituent units of the information processing unit are one cluster. Timer means for detecting a time-out by monitoring the time when there is no response inter-cluster communication when these clusters are connected by a transmission line and performing inter-cluster communication; Communication means for setting a communication partner cluster number, an inter-cluster communication command, and a data part of the inter-cluster communication process in a management table in order to cause a retry when a communication error occurs, and communication of the inter-cluster communication. A response for retrying the inter-cluster communication command when a timeout is notified from the timer means while monitoring an abnormality. And a management table in which the timer value for monitoring the time is added to the destination cluster number and the communication command unit, and all the timer values of the management table. Timer count control means for notifying the response monitoring means of a communication instruction that has been updated and timed out is provided for each of the plurality of clusters.

In another information processing system according to the present invention, in addition to the above-described configuration, time setting means having a table for storing information of an initial timer value which can be changed in units of the communication command, and refer to the communication command. The communication management means for setting the initial timer value of the time setting means to the timer value of the management table is provided for each of the plurality of clusters.

That is, the information processing system of the present invention comprises:
It has a management table in which a plurality of timer values can be set freely, and a timer value for time monitoring is added to each other's cluster number and communication command unit, and a timer count control unit that updates all timer values in the management table.

Thus, continuous inter-cluster communication can be monitored by a timer for each cluster and communication command, so that inter-cluster communication processing is not suspended without issuing communication, and inter-cluster communication performance is improved. There is no decrease, and there is no effect on the processing performance of the system, and the inter-cluster communication performance is dramatically improved.

Further, the inter-cluster communication command control method of the present invention has a time setting unit having a table of information on an initial timer value which can be changed in communication command units. When the time needs to be set according to the priority of the communication command, the setting can be easily changed.

Therefore, depending on the cluster of the communication partner, the system performance may be extremely slow, and the response of the inter-cluster communication may be slow. In some cases, an abnormality may occur in the transmission path during the inter-cluster communication. Since the inter-communication is not suspended and the inter-cluster communication can be continuously executed, the load on the CPU can be reduced.

If there is no communication between the response clusters of the receiving side cluster and a timeout occurs and an error occurs in the transmission line, the response monitoring unit of the communication command control unit makes a decision without starting the CPU and performs a retry process. After the execution again, only the final result is transmitted to the CPU, so that no load is imposed on the CPU.

Further, when the inter-cluster communication is executed,
Depending on the cluster of the communication partner, if the response is extremely slow due to the performance of the transmission path of inter-cluster communication, or if an error occurs on the transmission path, continuous inter-cluster communication can be executed and the cluster communication processing is suspended Without this, the inter-cluster communication performance is dramatically improved.

[0035]

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an information processing system according to one embodiment of the present invention. In FIG. 1, an information processing system according to an embodiment of the present invention includes CPUs (Central Processing Units) 2-1 to 2-n and storage devices 3-1 to 3 for storing communication information such as communication between the devices. −
n as a single cluster, and a plurality of such clusters (# 1 to #n) 1-1 to 1-n are connected to the transmission line 100.
The system is connected by.

In the above system, these clusters 1-1 to 1-n are communication command controllers 4-1 to 4 connected to the CPUs 2-1 to 2-n and the storage devices 3-1 to 3-n.
-N, and the communication command control units 4-1 to 4-n have timer units 41-1 to 41-n and time setting units 42-1 to 4-1-n.
42-n and timer / counter control units 43-1 to 43-n
Communication management units 44-1 to 44-n and a response monitoring unit 45
-1 to 45-n and management tables 46-1 to 46-n are provided.

The timer sections 41-1 to 41-n can set an arbitrary monitoring time, and generate a timer interrupt when the monitoring time elapses. Time setting units 42-1 to 42-
n has an initial timer value of a timeout time that can be changed for each communication command.

Timer / counter control units 43-1 to 43-n
Updates all the timer values in the management tables 46-1 to 46-n, and sends a communication command that has timed out to the response monitoring unit 45-n.
1-45-n. Communication management units 44-1 to 44-
“n” refers to the communication commands from the time setting units 42-1 to 42-n to the timer values in the management tables 46-1 to 46-n, and sets the timer values based on the initial timer values.

The response monitoring units 45-1 to 45-n start the inter-cluster communication process according to the instructions of the communication management units 44-1 to 44-n and the CPUs 2-1 to 2-n, and there is no response inter-cluster communication. Timer section 41 to monitor time assuming
-1 to 41-n are activated, and the timer units 41-1 to 41-n are activated.
Timeout processing is performed by a timer interrupt from.

When the inter-cluster communication returns a response inter-cluster communication from the transmission destination cluster in the inter-cluster communication, the response monitoring units 45-1 to 45-n identify the normal end or the abnormal end based on the response data, and determine whether the CPU 2 1 to 2-n, retry processing is performed when there is no response inter-cluster communication due to the occurrence of an abnormality in the transmission path 100 and a timeout occurs, and inter-cluster communication is executed again, and the result is transmitted to the CPU 2-1. ~
2-n is notified.

The management tables 46-1 to 46-n store timer values for time monitoring so that time monitoring can be performed in units of communication partner clusters and communication instructions even when a plurality of inter-cluster communications are executed simultaneously. Has been added.

FIG. 2 is a diagram showing the configuration of the table of the time setting unit 42 in FIG. In FIG. 2, the table of the time setting unit 42 includes a communication command 4 for each of the storage positions T1 to Tn.
2a and an initial timer value 42b of the waiting time in consideration of the priority for each communication command 42a are stored. FIG.
Has the same configuration as the table of the time setting unit 42 shown in FIG.

FIG. 3 is a diagram showing the configuration of the management table 46 of FIG. 3, the management table 46 stores a cluster number 46a, a communication command 46b, a timer value 46c, and a data section 46d for each of the storage locations K1 to Kn. The management tables 46-1 to 46-4 shown in FIG.
6-n has the same configuration as the management table 46 shown in FIG.

FIG. 4 is a diagram showing a simplified operation of the inter-cluster communication in the information processing system according to one embodiment of the present invention, and FIG. 5 is a diagram showing the inter-cluster communication of the inter-cluster communication command control system according to one embodiment of the present invention. It is a figure showing detailed operation. The inter-cluster communication command control in the information processing system according to the embodiment of the present invention will be described with reference to FIGS.

The CPU 2-1 of the cluster 1-1 sets data for inter-cluster communication in the storage device 3-1 (step S11 in FIG. 5), and thereafter, the communication management unit 44-1 of the communication command control unit 4-1. Request for inter-cluster communication (step S1 in FIG. 5). Upon receiving the request for inter-cluster communication from the CPU 2-1, the communication management unit 44-1 receives the management table 46-1.
Is created (step S52 in FIG. 5).

The communication management unit 44-1 starts with the management table 4
Referring to 6-1, the empty storage position (K1 to K in FIG. 3)
n) (step S53 in FIG. 5). When the storage position of the management table 46-1 is determined, the four items in the management table 46-1, the cluster number 46a, the communication command 46b, the timer value 46c, and the data section 46d are set (step S54 in FIG. 5).

In this case, since the information of the cluster number 46a and the communication command 46b is already known, they are set immediately.
For the timer value 46c, the communication command 42a of the time setting unit 44-1 and the management table 46b are referred to, and if they are the same value, the initial timer value 42b of the time setting unit 42 is read and set. CPU 2 is used for data section 46d.
-1 reads and sets the data written in the storage device 3-1.

The communication management section 44-1 has a management table 46-
When the information has been set to 1, the timer section 41-1 sets an arbitrary time and then instructs the timer to start (steps S55 and S21 in FIG. 5). The communication management unit 44-1 issues an inter-cluster communication process with a communication command to the cluster 1-2 (not shown) using the transmission path 100 (step S5 in FIG. 5).
6). At this time, if the inter-cluster communication is successful, the communication management unit 44-1 ends the process. At this time, if there is inter-cluster communication to other clusters # 3 to #n (clusters # 3 to # (n-1) are not shown), the same processing as above is performed again.

The timer section 41-1 activated by the communication management section 44-1 generates a timer interrupt to the timer count control section 43-1 after a set time elapses (step S22 in FIG. 5). Timer count control unit 43-1
Stores the timer value 46c of the management table 46-1 in the timer unit 4
Subtraction is performed for an arbitrary time set in 1-1.

Thereafter, the timer count control section 43-1
Sets an arbitrary time in the timer section 41-1 and instructs the timer to start again (step S41 in FIG. 5). as a result,
The timer count control unit 43-1 includes a management table 46-1.
It is determined whether or not a timeout has occurred due to the same time as the timer value 46c (step S42 in FIG. 5). If the result is not '0', the timer count control unit 43-1
Determines whether all timer values registered in the management table 46-1 have been subtracted (step S4 in FIG. 5).
3).

If there is a table registered in the management table 46-1, the timer count control unit 43-1 performs the above processing until there is no table that has not been decremented even after the timer interrupt has occurred. Repeated. If there is no registered one, timer count control section 43-1 ends the process. The timer count control unit 4
If the result of the subtraction is “0”, 3-1 notifies the response monitoring unit 45-1 of the timeout.

When the response monitor 45-1 receives the timeout notification, the timer value 46c in the management table 46-1 is received.
Is cleared to zero and deleted (step S61 in FIG. 5). Thereafter, the response monitoring unit 45-
1 executes a retry process (step S62 in FIG. 5).
When executing the retry processing, the response monitoring unit 45-1 checks whether the retry processing has already been executed (Step S63 in FIG. 5).

If the response monitoring unit 45-1 has not already performed the retry process, the response monitoring unit 45-1 notifies the communication management unit 44-1 of the retry activation, and the communication management unit 44-1 performs the retry process (step S57 in FIG. 5). ). Response monitoring unit 45-
If the retry processing has already been performed, the communication control unit 1 notifies the communication management unit 44-1 of the abnormal end, and notifies the CPU 2-1 of the abnormal end (step S58 in FIG. 5).

When the CPU 2-1 receives the abnormal end notification,
An abnormal end process is performed (step S3 in FIG. 5), and the process ends. When any abnormality occurs in the transmission path 100 at the time of issuing the inter-cluster communication from the communication management unit 44-1, the CPU 2-1 notifies the response monitoring unit 45-1 of a failure (step S81 in FIG. 5). Upon receiving the failure notification, the response monitoring unit 45-1 performs the same operation as when receiving the timeout notification (step S61 in FIG. 5).

When the inter-cluster communication processing from the CPU 2-1 to the cluster 1-2 is completed, the inter-cluster communication is issued from the cluster 1-2 to the response monitoring unit 45-1.
When the response monitoring unit 45-1 receives the response cluster communication,
The cluster number and the communication command are acquired from the information of the response cluster communication. The response monitoring unit 45-1 compares the acquired cluster number and communication command with the cluster number 46a and communication command 46b of the management table 46-1, and deletes the matched table (step S64 in FIG. 5).

Thereafter, the response monitoring unit 45-1 sends a normal end notification or an abnormal end notification to the communication management unit 44-1 based on the data of the response inter-cluster communication. Communication management unit 44-
1 receives a normal end notification or an abnormal end notification,
A normal end notification or an abnormal end notification is sent to U2-1 (steps S59 and S60 in FIG. 5), and a normal end process or an abnormal end process is executed by the CPU 2-1 (steps S4 and S in FIG. 5).
5).

A plurality of clusters 1-1 to 1-n are connected to the transmission line 1
00, the communication command 4 is stored in the table of the time setting units 42, 42-1 to 42-n.
Initial timer value of waiting time considering priority for each 2a 4
2b to execute the inter-cluster communication, and at the same time, the timer value 46c in which the value of the initial timer value 42b is set.
By setting and managing a plurality of management tables 46, 46-1 to 46-n having
2-1 → communication management unit 44-1 → management table 46-1,
Communication management unit 44-1 → timer unit 41-1; communication management unit 4
4-1 → cluster 1-2), a time monitoring process in the timer unit 41-1 and a process after receiving the response inter-cluster communication (cluster 1-2 → response monitoring unit 45-1 → CPU 2-1). Can be operated independently.

As a result, a new cluster is added, and inter-cluster communication becomes possible. Even if the number of clusters increases, it takes time for response cluster communication due to the performance of the transmission path 100 for inter-cluster communication or the like. Even if 100 becomes abnormal, continuous inter-cluster communication can be executed without any problem.

According to the above monitoring method, all continuous inter-cluster communication can be monitored by a timer for each cluster and communication command. Therefore, the inter-cluster communication processing is not suspended without issuing a communication, and the inter-cluster communication processing is not suspended. The communication performance does not decrease, and the processing performance of the system is not affected, and the inter-cluster communication performance can be dramatically improved.

[0060]

As described above, according to the information processing system of the present invention, a timer value for time monitoring is added to the management table for each destination cluster number and communication command unit, and all timer values in the management table are added. By notifying the response monitoring means of the updated and timed-out communication instruction, the inter-cluster communication can be executed continuously, the load on the CPU can be reduced, and the inter-cluster communication performance can be dramatically improved. There is an effect that it can be improved.

According to another information processing system of the present invention, the timer value of the management table is set by referring to the communication command.
By setting the initial timer value of the time setting means having a table for storing the information of the initial timer value that can be changed in communication instruction units, the timeout time is set in accordance with the priority of the communication instruction due to a problem in system operation. When setting is required, there is an effect that the setting can be easily changed.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an information processing system according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a table of a time setting unit in FIG. 1;

FIG. 3 is a diagram showing a configuration of a management table in FIG. 1;

FIG. 4 is a diagram showing a simplified operation of inter-cluster communication in the information processing system according to one embodiment of the present invention.

FIG. 5 is a diagram showing a detailed operation of the inter-cluster communication of the inter-cluster communication command control method according to one embodiment of the present invention.

FIG. 6 is a block diagram illustrating a configuration of an information processing system according to a conventional example.

FIG. 7 is a diagram showing a simplified operation of inter-cluster communication in an information processing system according to a conventional example.

FIG. 8 is a diagram showing a configuration of a management table of a conventional inter-cluster communication command control method.

[Explanation of symbols]

 1-1 to 1-n cluster (# 1 to #n) 2-1 to 2-n Central processing unit 3-1 to 3-n Storage device 4-1 to 4-n Communication command control unit 41-1 to 41 -N timer unit 42, 42-1 to 42-n time setting unit 43-1 to 43-n timer count control unit 44-1 to 44-n communication management unit 45-1 to 45-n response monitoring unit 46, 46 -1 to 46-n Management table 42a, 46b Inter-cluster communication command 42b Initial timer value 46a Cluster number 46c Timer value 46d Data section 100 Transmission path

Claims (2)

[Claims] An information processing apparatus including a central processing unit and a storage device that stores communication information such as communication between the devices is defined as one cluster, and a plurality of clusters are provided. A timer means for detecting a timeout by time monitoring when a cluster is connected by a transmission path and executing inter-cluster communication and there is no response inter-cluster communication, and a retry is performed by the timer means when a timeout or communication error occurs. A communication management means for setting a communication partner cluster number, an inter-cluster communication instruction, and a data part in the inter-cluster communication processing in a management table; and monitoring a communication abnormality of the inter-cluster communication and notifying a timeout from the timer means. And a response monitoring means for retrying the inter-cluster communication command when the plurality of clusters are executed. A management table in which the timer value for time monitoring is added to a destination cluster number and a communication command unit; and a communication command that has timed out by updating all timer values in the management table to the response monitoring unit. An information processing system comprising: a timer count control unit for notifying each of the plurality of clusters. 2. A time setting means having a table for storing information of an initial timer value which can be changed in units of said communication command, and an initial value of said time setting means in a timer value of said management table with reference to said communication command. 2. The information processing system according to claim 1, wherein communication management means for setting a timer value is included in each of the plurality of clusters.