JP5605279B2 - Inter-process synchronization method - Google Patents

Description

  The present invention relates to an inter-process synchronization method for a computer system that performs processing by synchronizing a plurality of processes.

  The conventional inter-process synchronization method sets the total number of synchronization points in advance in the synchronization manager for the processors participating in barrier synchronization, and reaches the synchronization point by a counter each time a barrier synchronization request from each processor occurs. When the count value of the counter reaches the total number of synchronization points, a barrier synchronization establishment notification is issued to release the synchronization wait in each processor (see, for example, Patent Document 1).

JP 2000-215182 A

  However, the conventional inter-process synchronization method is not supposed to be applied to a computer system having a fault tolerant system configuration including a normal system and a standby system host processor. For this reason, there is a problem that the synchronization wait is not released unless all the host processors in the standby system are activated.

  In addition, when the active and standby host processors are switched, the host processor to be switched is restarted and initialized. This causes a problem that the restarted host processor is not released from the synchronization wait state. To solve this problem, set a predetermined time-out period, and when the time-out period is reached, determine that the active system and standby system have switched and restarted, and forcibly release the synchronization wait. It is necessary to make it.

  However, the time-out period is normally set to a sufficiently long period in consideration of the arrival time until the synchronization point is reached after the host processor is restarted. For this reason, since no host processor is released from waiting for synchronization until the timeout time is reached, there is a problem that the time until synchronization is completed is unnecessarily extended.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to obtain an interprocess synchronization method that does not depend on timeout in a computer system having a fault-tolerant system configuration of a normal system and a standby system.

  The inter-process synchronization method according to the present invention includes a plurality of sets including a normal system and a standby system host, and in a computer system including a synchronization monitoring device, a synchronization wait release state from a synchronization wait state between different sets of processes. Is an inter-process synchronization method that synchronizes between processes by obtaining timing adjustments, and synchronization generated by each host according to the progress of a predetermined number of tasks executed by each host before releasing the wait for synchronization. The step of counting the number of calls by distinguishing between the active system and the standby system for each group, and the total number of the total number of synchronous call counts issued by the above-mentioned active system host coincided with a predetermined value. Sometimes, the step of canceling the standby wait of the active host and the count value of the number of synchronous calls issued by the standby host are the same as the active host of the standby host. When it becomes equal to the synchronous call number count value when the synchronous waiting list is canceled, in which and a step of releasing a synchronization wait in standby host.

  According to the present invention, the synchronization waiting cancellation condition of the active host does not depend on the standby host synchronization waiting cancellation condition, so unnecessary timeout waiting can be eliminated, and an interprocess synchronization method that does not depend on timeout is obtained. be able to.

1 is a diagram illustrating a configuration of a computer system that constitutes a normal and standby fault tolerant system according to Embodiment 1. FIG. 4 is a flowchart illustrating an operation example of the synchronization monitoring apparatus according to the first embodiment. It is a figure which shows the processing flow when the synchronous monitoring apparatus by Embodiment 1 detects regular standby switching. 10 is a flowchart illustrating an operation example of the synchronization monitoring apparatus according to the second embodiment. FIG. 10 is a diagram illustrating a processing flow of the synchronization monitoring apparatus according to the third embodiment.

Embodiment 1 FIG.
FIG. 1 is a diagram showing a configuration of a computer system that constitutes a fault-tolerant system of a normal system and a standby system according to Embodiment 1 of the present invention, (a) shows the overall configuration of the computer system, and (b) 2 shows a configuration of a synchronization monitoring device provided in the computer system, and FIG. 3C is a diagram showing a configuration example of a partial data area of a storage unit provided in the synchronization monitoring device.

  In FIG. 1A, the computer system of the first embodiment forms a fault tolerant system composed of an active host group 3 a, a standby host group 3 b, a synchronization monitoring device 2, and a network 6. A fault-tolerant system is a system that continues processing normally even if some of its components fail. In this example, a pair of regular and standby hosts are provided, Even if a failure occurs, the system is configured to switch to the other host that has not failed and continue processing normally. Here, the host is a computer or processor that provides services, processing capabilities, and the like to another device, computer, or processor via the network 6. Each host of the active host group 3a and the standby host group 3b performs inter-process communication between different belonging groups (groups) via the network 6 so that there is no collision when accessing shared information. Synchronize and exchange information with each other. The network 6 may be a LAN (Local Area Network) cable, a telephone line, an optical line, or a data bus that connects processors.

In the example of FIG. 1A, description will be made on the assumption that there are three hosts in the normal system as a form of the computer system that constitutes the fault tolerant system of the normal system and the standby system. In this case, the standby system inevitably has three hosts that are paired with the regular system.
The regular host group 3a is defined as a host 1a, a host 1c, and a host 1e. Further, the standby host group 3b that forms a pair with the regular system is referred to as a host 1b, a host 1d, and a host 1f.
The hosts 1a, 1c, and 1e of the active host group 3a and the hosts 1b, 1d, and 1f of the standby host group 3b are connected to the network 6, respectively.
The standby hosts that are paired with the regular hosts are grouped into the same group (group), the host 1a and the host 1b are group 4a, the host 1c and the host 1d are group 4b, and the host 1e and the host 1f are group 4c. As shown in FIG.
Each group is composed of one standby host that is paired with one regular host.
Two or more regular and standby hosts belonging to each group may be provided, and the number of groups may be two or three or more.

Each of the hosts 1a, 1c, 1e of the active host group 3a and each of the hosts 1b, 1d, 1f of the standby host group 3b is provided with a control unit ( (Not shown) detects the abnormality and is automatically controlled to switch between the active host and the standby host in the same group.
When the active host and the standby host are switched, the host that is switched due to an abnormality is restarted, and the progress of the process being executed is initialized (restarted).

  Each of the hosts 1a to 1f sequentially performs a preset process according to a program set in the processor. A program set in each of the hosts 1a to 1f includes a plurality of tasks composed of a group of processes. Each of the hosts 1a to 1f issues a synchronous call when reaching a preset “synchronization point” by completing the processing of a predetermined task. This synchronization point is set each time a prescribed task is completed in each of the hosts 1a to 1f. That is, each host 1a to 1f processes tasks sequentially and issues a synchronous call whenever it reaches the synchronization point.

Until the inter-process synchronization is performed between the hosts 1a to 1f, each host obtains a synchronization waiting state. For example, when each host 1a to 1f issues a required number of synchronous calls and reaches a "synchronization wait point" that is a timing at which inter-process synchronization should be executed between different hosts, each host 1a to 1f receives host information. A message containing is sent to the synchronization wait state. Thereafter, the synchronization monitoring device 2 described later releases the synchronization wait of each of the hosts 1a to 1f.
When the host 1a to 1f is released from the synchronization wait state, it synchronizes between the processes processed by each host belonging to different groups and sets the synchronization timing so that there is no collision in access to the shared information. However, by performing parallel processing, a predetermined series of tasks is performed by the cooperative work of the hosts belonging to different groups.
Note that the number of synchronous calls issued by each host (hereinafter referred to as the number of synchronous calls) before the inter-process synchronization is performed between the hosts is usually different for each host, and the synchronization waiting time of each host is not necessarily uniform. is not.

  The synchronization monitoring device 2 is connected to each host (1a to 1f) via the network 6. 2B, the synchronization monitoring device 2 includes a main control unit (not shown) that performs main processing, a synchronous call recognition unit 21, a storage unit 22, and a restart recognition unit 23. , A writing unit 24, a synchronization waiting determination unit 25, and a synchronization waiting release instruction unit 26.

  The synchronous call recognition unit 21 of the synchronous monitoring device 2 is means for recognizing a synchronous call from each host (1a to 1f). The synchronous call recognizing unit 21 recognizes the synchronous call from each of the hosts 1a to 1f by distinguishing between the normal host group 3a and the standby host group 3b.

  The storage unit 22 held by the synchronization monitoring device 2 is connected to the active host group 3a and the standby host group 3b, and records and reproduces the system state of each host. Further, as shown in FIG. 2C, a part of the data area held by the storage unit 22 includes a matrix table composed of counters divided for the normal system and the standby system of each group (4a to 4c). 7 is configured. In the counter of the matrix table 7, for each group (4a to 4c), a count value of “synchronous call” for each of the active system and the standby system (a count value obtained by accumulating the number of generated synchronous calls) is recorded. Retained. Writing of values to the matrix table 7 is performed by the writing unit 24.

  As shown in FIG. 2C, the synchronous call total area 8 is set in the data area held by the storage unit 22. The total number of synchronous calls 8 stores and holds the value of “total number of synchronous calls” set as the sum of the numbers of calls only for the regular system. The writing unit 24 writes a value in the synchronous call total area 8.

  The restart recognition unit 23 of the synchronization monitoring device 2 is means for recognizing that each of the hosts 1a to 1f has been restarted. When the restart recognition unit 23 recognizes the restart of the original active host that becomes the standby system at the timing of switching of the host from the active system to the standby system, it accesses the data area of the matrix table 7. The number of inter-process synchronous calls of the standby host in the group to which the original active host to be switched belongs is reset to zero. When switching between the active system and the standby system occurs, the host that has been switched becomes a new active system, and the host that is switched and restarted becomes a new standby system.

  The synchronization wait determination unit 25 of the synchronization monitoring device 2 is means for determining that a task has been waiting for synchronization. For example, when a task of any host reaches the synchronization waiting point, the task of the host transmits a message including its host information to the synchronization waiting determination unit 25 of the synchronization monitoring device 2, thereby causing the synchronization waiting determination unit 25. Recognizes that the task is waiting for synchronization and makes a determination.

  The synchronization wait release instructing unit 26 of the synchronization monitoring device 2 is means for instructing an arbitrary host to issue a synchronization wait release instruction when the conditions for releasing the synchronization wait are satisfied. For example, when the synchronization monitoring apparatus 2 determines that the task is waiting for synchronization and the synchronization waiting determination unit 25 recognizes that the condition for releasing synchronization waiting is satisfied, the synchronization monitoring apparatus 2 sends a message to the host to release the synchronization waiting. Send. When receiving this message, the host releases the synchronization wait.

Next, an example of the operation of the synchronization monitoring device 2 according to the first embodiment will be described below.
In this operation example, the active host 1a has three tasks that need to wait for synchronization (hereinafter referred to as the number of synchronization waits), the host 1c has four synchronization waits, and the host 1e has two synchronization waits. Assume that there are two cases.
That is, the “synchronization wait release condition” in which the synchronization wait state is released when the timing for performing the inter-process synchronization between the hosts is reached is when the total number of tasks in the synchronization wait state becomes nine. This value corresponds to the sum of the synchronization waiting number 3 of the active host 1a, the synchronization waiting number 4 of the active host 1b, and the synchronization waiting number 2 of the active host 1a (that is, 9 = 3 + 4 + 2).
The fact that the total number of synchronous calls is 9, for example, is information determined at the time of system design. Therefore, the synchronization monitoring device 2 uses the write unit 24 to synchronously call the total number of synchronous calls for inter-process synchronization as 9 in advance. The value 9 of the total number of calls for interprocess synchronization is written in the total number area 8.
The number of synchronization waits corresponds to the count value of the number of synchronous calls required by each of the hosts 1a to 1f until the synchronization wait is canceled. When the task waits for synchronization, the count value of “synchronous call count” matches the synchronization wait number.

FIG. 2 is a flowchart showing an operation example of the synchronization monitoring apparatus 2 according to the first embodiment.
In the figure, the synchronization monitoring device 2 performs a process of monitoring whether or not a synchronous call from each of the hosts 1a to 1f has been made by the synchronous call recognition unit 21 (step S100).
As a result of monitoring, when it is detected that a synchronous call has been made, the storage unit 22 is accessed, and the count value of “synchronous call number” of the corresponding counter in the matrix table 7 of the host that has executed the synchronous call is set to one. Increment (count up) (step S101).

Next, the synchronization monitoring device 2 determines whether the host belongs to the active host group 3a or the standby host group 3b (step S102).
If the result of determination in step S102 is that the active host belongs to the active host group 3a, the matrix table 7 is accessed and the number of synchronous calls of the counters corresponding to the active hosts 1a, 1c, and 1e of all groups is determined. The sum is obtained, and based on the obtained sum, it is determined whether or not the “synchronization waiting release condition” of the active host is satisfied (step S103).
Here, “the total sum of the synchronous calls of the active hosts of all groups matches the value stored in the synchronous call total number area 8 (the preset total number of synchronous calls)” This is a determination condition that satisfies the synchronization wait cancellation condition.

As a result of the determination in step S103, the release condition for waiting for synchronization of the active hosts is satisfied, that is, “the sum of the synchronous calls of the active hosts in all groups matches the value in the synchronous call total area 8” Thus, the synchronization wait cancellation instructing unit 26 of the synchronization monitoring device 2 instructs all hosts 1a, 1c, and 1e in the active host group 3a to release the synchronization wait (step S104).
If “the total number of synchronous calls of the active hosts in all groups does not match the value in the synchronous call total area 8”, the process returns to step S100, and the monitoring process by the synchronous call recognition unit 21 is continued.

On the other hand, if the result of determination in step S102 is that the host is a standby host belonging to the standby host group 3b, the synchronization monitoring device 2 performs the process of step S105.
In step S105, it is determined whether or not the standby host satisfies the “synchronization wait cancellation condition”.

The “synchronization wait release condition” of this standby host is different from that of the regular host.
The "synchronization wait release condition" here means "count value of the number of synchronous calls of the active host that is paired with the group to which the standby host belongs in the state where the active host has been released from synchronization" The count value of the number of synchronous calls equal to (in this case, the number of waiting for synchronization) is set in the counter of the matrix table 7 ”.
Further, in this case, when the synchronization waiting determination unit 25 of the synchronization monitoring device 2 recognizes that the task is waiting for synchronization, the state becomes the state that “the active host has been released from synchronization”.

  If the standby host satisfies the “synchronization waiting cancellation condition” as a result of the determination, the synchronization is canceled only for the standby host (step S106).

Here, since the synchronization monitoring device 2 counts the count value (count value) of the number of regular synchronization calls for each group, the synchronization monitoring device 2 recognizes the number of synchronization waits required for the standby system to release the synchronization state. can do.
For example, the breakdown of the count value of the number of synchronous calls of each host (that is, the number of synchronization waits) when the total number of synchronous calls is nine, which is a condition for releasing synchronization of the standby system, is 3 for group 4a, The group 4b is 4, and the group 4c is 2.
Therefore, the standby host 1b is released from the synchronization wait when the synchronization with the active host 1a is released and the number of synchronization waits is 3. Similarly, the standby host 1d has a synchronization wait number of 4, and the standby host 1f has a synchronization wait release instruction unit 26 of the synchronization monitoring device 2 when the synchronization wait number is 2, respectively. Is used to cancel the synchronization wait.

  Once the normal system synchronization release condition is satisfied, the value of the number of synchronization waits for obtaining the total number of synchronous calls in each normal system host in the matrix table 7 is fixed. When the host is restarted due to switching between the active system and the standby system, the restarting host starts as a new standby system.

FIG. 3 shows processing when the synchronization monitoring device 2 detects switching between the normal system and the standby system.
The restart recognition unit 23 of the synchronization monitoring device 2 resets the number of synchronous calls of the standby system to 0 when it detects the restart of the original active host that becomes the standby system by switching between the active system and the standby system. (Step S200).
When the original active host that newly becomes the standby system is restarted, the count value of the number of synchronous calls of the new standby system becomes equal to the release condition for the standby system synchronization wait by the same processing as in FIG. Then, when the count value of the number of synchronous calls of the standby host becomes the same value as the synchronization wait number of the active host, the synchronization wait is released.

Here, the distinction between the active system and the standby system is pre-allocated to the host at the first startup, but it is not always fixed to the host. As a result, the roles of hosts in the same group are switched.
Since the standby wait cancellation condition of the primary host does not depend on the standby wait cancellation condition of the standby host, even if the standby host becomes abnormal, the synchronization wait is not released. For this reason, as long as the conditions for the regular host are satisfied, the cancellation of the synchronization wait can be achieved, and the unnecessary timeout wait can be eliminated.

Note that the standby wait release of the standby host may be delayed from the release of the standby wait of the active host. For this reason, in the first embodiment, if an abnormality occurs immediately after the active host cancels the synchronization wait, the standby host does not reach the synchronization wait cancellation condition when switching to the standby host. There is a possibility that the switching may be hindered.
However, because the same processing is being executed on the hosts in the same group, the standby host may reach the condition for releasing the synchronization wait without lagging behind the timing when the active host reaches the release of the synchronization wait. is assumed. For this reason, such an event will not cause a problem unless the primary host fails and the primary host and standby host are switched immediately after the standby synchronization of the primary host is released. The possibility is small and there is no substantial problem.

  As described above, the inter-process synchronization method according to the first embodiment includes a plurality of sets including the normal system and the standby system host, and in the computer system including the synchronization monitoring device, the synchronization wait is performed between different sets of processes. This is an inter-process synchronization method that obtains the synchronization wait release state from the status and adjusts the timing to synchronize the processes, and according to the progress of a predetermined number of tasks executed by each of the hosts before releasing the synchronization wait And counting the number of synchronous calls issued by each host by distinguishing between the active system and the standby system for each group, and the total number of the total number of synchronous call counts issued by the above-mentioned active host When the value matches the specified value, the step of canceling the synchronization wait of the active host and the count value of the number of synchronous calls issued by the standby host And a step of releasing the standby waiting of the standby host when the same value as the count of the number of synchronous calls when the standby waiting of the active host in the same group as the host is released is provided. .

  As a result, the waiting condition for releasing the synchronization wait for the active host does not depend on the waiting condition for waiting for the synchronization of the standby host, so that unnecessary timeout waiting can be eliminated and an inter-process synchronization method that does not depend on timeout can be obtained. The effect is obtained.

Embodiment 2. FIG.
In the computer system synchronization monitoring apparatus 2 according to the second embodiment of the present invention, the synchronization wait for the standby hosts in the same group is set as the condition for releasing the synchronization wait for the active host in addition to the condition for the first embodiment. The fact that the release conditions are complete is added to the conditions.

FIG. 4 is a flowchart showing the operation of the synchronization monitoring apparatus 2 according to the second embodiment.
Compared with the flowchart of the first embodiment described in FIG. 2, in the second embodiment, step S104 is replaced with step S104-2. Since other steps are the same, description thereof is omitted.

  In step S104-2 in FIG. 4, in addition to the condition in step S103 being satisfied, this is applicable only when the condition for canceling the synchronization wait of the standby host that is paired with the regular host in the same group is satisfied. Condition determination processing is performed so that the synchronization wait of the active host is released. That is, in this case, the count value (count value) of the number of synchronous calls coincides with the standby host paired with the regular host in the same group.

  In the second embodiment, the active host waits until the synchronization waiting of the paired standby host is released. Therefore, when an error occurs in the active host immediately after the synchronization waiting is released, the standby host Even if the switch to the host occurs, it can be ensured that the standby host has reached the synchronization point.

  As described above, in the second embodiment, assuming that the standby host cannot reach the synchronization point due to an abnormality or the like, the maximum timeout period during which the active host can wait for the standby host to reach the synchronization wait number is set. It is characterized by having a mechanism for canceling the waiting for synchronization without waiting for the waiting for synchronization of the standby system to be completed when a timeout occurs.

  By doing so, it is possible to avoid the hang-up of the fault tolerant system due to an unnecessarily long timeout time. Further, in the second embodiment, compared to the conventional inter-process synchronization method, the time-out waiting range can be limited to the same group, so a shorter appropriate time-out time can be set. Will be able to do.

  As described above, the inter-process synchronization method according to the second embodiment includes a plurality of sets composed of a plurality of hosts, and in a computer system including a synchronization monitoring device, a state of waiting for synchronization between different sets of processes. Is an inter-process synchronization method that adjusts timing by obtaining the state of synchronization wait release from the host, and synchronizes between processes, according to the progress of a predetermined number of tasks executed by each host before releasing the synchronization wait The step of counting the number of synchronous calls issued by each host separately for each group and for each host, and selecting the maximum value of the number of synchronous calls issued by each host in the same group, When the sum of all sets is taken for each set, and when the sum matches a predetermined value, each host that gives the maximum value for each set is defined as a regular host. The step of canceling the synchronization wait of the primary host and the host other than the primary host in the same group as the standby host, and the count value of the number of synchronous calls issued by the standby host is the standby host And a step of releasing the standby wait of the standby host when the same value as the count of the number of synchronous calls when the standby wait of the same set of active hosts is released. .

  As a result, the time-out waiting range can be limited to the same group, so that an effect that a shorter appropriate time-out time can be set can be obtained.

Embodiment 3 FIG.
In the computer system according to the third embodiment of the present invention, when the host is started, the synchronization monitoring device 2 starts synchronization waiting without predetermining which host in the group will be the active system and the standby system. When it is detected that the sum of the count values (count values) of the number of synchronous calls of any host in all groups has reached the preset total number of synchronous calls, the maximum number of synchronous calls in each group (this In this case, the host that has reached the number of waiting for synchronization) is defined as the active host, and the waiting for synchronization of the active host is released. By doing so, it is possible to minimize the time required for synchronization.

  In other words, in the third embodiment, the active host and the standby host are dynamically determined, and the host that has the conditions for releasing the synchronization wait earliest is used as the active host. This time can be the fastest.

FIG. 5 is a flowchart showing processing of the synchronization monitoring device 2 of the computer system according to the third embodiment.
In step S300, upon receiving a synchronous call from each host and receiving the synchronous call, the counter of the number of synchronous calls of the corresponding host in the matrix table 7 is incremented by one.

However, since it is not determined whether the corresponding host is the regular system or the standby system, the counter is incremented as a temporary regular system or a provisional standby system (step S301).
Next, the value that maximizes the count value (count) of the counter for the number of synchronous calls in each group is extracted, and the sum of the maximum values of each count value extracted from all groups is calculated. It is confirmed whether or not the total value matches the value of the prescribed number of synchronous call total area 8 (step S302).

  If they match as a result of confirmation, this indicates that there is a combination necessary for canceling the synchronization wait, so that the host that gives the maximum count value of the number of synchronous calls in each group is treated as a regular host.

At this time, if the counter value of the temporary standby host is the maximum value, the value of the normal system and the value of the standby system in the matrix table 7 are exchanged (step S303).
Now that the conditions for releasing the waiting for synchronization of the normal system are met, the waiting for synchronization of the host determined as the active system is released (step S304).
Since the subsequent processing is the same processing flow as the flowchart of FIG. 2 described in the first embodiment, description thereof is omitted.

  As described above, the inter-process synchronization method according to the third embodiment includes a plurality of sets composed of a plurality of hosts, and in a computer system including a synchronization monitoring device, a state of waiting for synchronization between different sets of processes. Is an inter-process synchronization method that adjusts timing by obtaining the state of synchronization wait release from the host, and synchronizes between processes, according to the progress of a predetermined number of tasks executed by each host before releasing the synchronization wait The step of counting the number of synchronous calls issued by each host separately for each group and for each host, and selecting the maximum value of the number of synchronous calls issued by each host in the same group, When the sum of all sets is taken for each set, and when the sum matches a predetermined value, each host that gives the maximum value for each set is defined as a regular host. The step of canceling the synchronization wait of the primary host and the host other than the primary host in the same group as the standby host, and the count value of the number of synchronous calls issued by the standby host is the standby host And a step of releasing the standby wait of the standby host when the same value as the count of the number of synchronous calls when the standby wait of the same set of active hosts is released. .

  As a result, in addition to the effect of the first embodiment, the effect that the time required until synchronization can be minimized can be obtained.

In a computer system that requires reliability, a fault tolerant system is configured so that processing can be continued even if some computers fail. On the other hand, in a large-scale computer system, it is necessary to perform parallel processing while maintaining synchronization because the programs are linked between the computers.
The inter-process synchronization method described in the first to third embodiments can be used for performing a synchronization process on a program between computers constituting a computer system that forms a fault-tolerant system.

  1a to 1f Host, 2 Synchronization monitoring device, 3a Host group constituting the active system, 3b Host group constituting the standby system, 4a to 4c Host group constituting the normal system and standby system group (set), 5a Regular system Synchronization release instruction, 5b Standby system synchronization release instruction, 6 Synchronous call instruction, 7 Matrix table, 8 Synchronous call total area.

Claims (3)

A computer system having a plurality of sets of active and standby hosts, and a computer system having a synchronization monitoring device, an inter-process synchronization method for synchronizing the processes of different sets of active and standby hosts ,
The number of synchronous calls issued by each process of the active system and the standby system host until each of the active system and the standby system host cancels the synchronization standby state from the synchronization standby state, respectively, A step of distinguishing and counting a normal system and a standby system for each set of
When the total number of all counts of the synchronous call counts issued by the active host process matches the predetermined value, the synchronization monitoring device waits for synchronization of all the active host processes. The step of releasing
The synchronization monitoring device, process synchronization call number of counts that originated in the standby host, synchronous call number when a synchronization wait process of the standby host and the same set of the conventional system host is released when it becomes equal to the count value, and releasing the synchronization wait process of the standby host,
With inter-process synchronization. In a computer system provided with a plurality of sets composed of a plurality of hosts and provided with a synchronization monitoring device, an inter-process synchronization method that synchronizes the processes of the respective hosts in different sets,
The synchronization monitoring device determines the number of synchronous calls issued by the processes of each host for each set and host until each of the active and standby hosts releases the synchronization waiting state from the synchronization waiting state. A step of counting each separately,
The synchronization monitoring device selects the maximum count value of the number of synchronous calls issued by each host in the same group, takes the sum of all the groups for the maximum value, and the sum matches the predetermined value. Sometimes, each host giving the maximum value for each group as an active host, releasing the process of waiting for synchronization of the process of the active host,
The synchronization monitoring device uses a host other than the regular host in the same group as a standby host, and the count value of the number of synchronous calls issued by the process of the standby host is the same as that of the standby host. when synchronization wait in the conventional system host is the same as the synchronization number of calls counted value when it is released, and releasing the synchronization wait in the standby host,
With inter-process synchronization. A computer system having a plurality of sets of active and standby hosts, and a computer system having a synchronization monitoring device, an inter-process synchronization method for synchronizing the processes of different sets of active and standby hosts ,
The synchronization monitoring device uses the number of synchronous calls issued by the processes of each host for each set until the respective active and standby hosts release the synchronization waiting state from the synchronization waiting state. Distinguishing and counting the system and the standby system;
The synchronization monitoring device, the total number of the entire set of process synchronous call number of the count values emitted in the conventional system host, matches the predefined value, yet the conventional system host the same set of standby host and synchronous call number of the count value of the process, if they match the synchronous call number of the count value of the conventional system host, and releasing the synchronization wait process of the conventional system host,
The synchronization monitoring device counts the number of synchronous calls issued by the standby host, and the count of synchronous calls when the process of the standby host in the same set as the standby host is released from the synchronization wait state. when it is the same, the step of releasing the synchronization wait in the standby host,
With inter-process synchronization.

Images