JPH11203221A - Data transfer device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To have an inter-process cooperation established even at the time of feed forward contention control where a high throughput can be obtained. SOLUTION: A timer control part 105 inspects a normal standby display signal supplied by a contention state collation part 106, a count stop signal is asserted when a standby is normal and the count stop signal is negated when it is not normal. An inter-process cooperation timer 102 sets a counter value to a normalized value when a reset signal is given from a process, and starts counting when a count start signal is given from the process. When the counter value becomes zero, an expiration notification signal is notified of to a process and an abnormal processing is transferred to. The inter-process cooperation timer 102 down counts a clock signal while the count stop signal is not supplied from the timer control part 105.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to coordination between processes when client-server processes communicate with each other via a packet switching network and perform a series of print processing and the like.

[0002]

2. Description of the Related Art In a packet switching network in which a plurality of computers are connected, processes on the computers via the network communicate with each other to perform a series of processing. In response to a "request" from one process, a "response" from the other process is confirmed, and the process proceeds to the next. In this way, confirming the response of the partner process between processes is called inter-process cooperation.

For the coordination between processes, the following two conditions are required. Data can be transferred without transmission errors. The other process is running. If these conditions are not met, no "response" will be returned, and inter-process coordination will not hold.

In order to satisfy the conditions, both processes assume the use of protocol software for error recovery. That is, even if a "request" or "response" may suffer a bit error in the packet switching network, it is reliably delivered to the partner process by error detection and retransmission by the protocol software. This protocol software detects network abnormalities such as disconnection
Detect autonomously. When the network is abnormal,
The protocol software rejects the request to transfer the “request” or “response” from the process. In this case, the process performs an abnormality handling process.

When one process issues a "request" to confirm a condition, that process starts a timer and counts the elapsed time. If there is no “response” even after the time T has elapsed, the process performs an abnormality handling process.

Here, how to determine the time T will be described. "Request" and "Response" have the data amount D by themselves. Therefore, the time T needs to be longer than the data transfer time of the data amount D. Usually, between the partner process
The data transfer time in a network with a sufficiently small transmission band is adopted so that there may be any packet switching network. The "request" data contains several GBy
There are also things that extend to tes. In some cases, the data amount is undecided before transmission. In such cases, the "request"
Data is usually divided into blocks of about 64 Kbytes, and a timer is started for each block. That is, the time T is
It can be determined by a transmission time of about 64 KB.

In a client-server type system using a packet switching network, "requests" from a client process on a client computer concentrate on the server process on the server computer. This concentration causes contention on the transmission media connecting the server computer to the network.

Here, the conflict will be described. The transmission band of the transmission medium is represented by the amount of transmission data per unit time. The maximum transmission bandwidth of the transmission medium (hereinafter Rmax [bit /
sec]) is a fixed finite value. The bandwidth usage (hereinafter, r (t)) of each data transmission accompanying the “request” from each client process is used by dividing Rmax. The sum of r (t) Σr (t) is Rmax of the transmission medium
If the number is larger, the transmission bandwidth is scrambled between the respective data transmissions. This is called a conflict. Data that requires a band exceeding Rmax is lost without being transmitted. Therefore, conflict control is performed to eliminate or avoid the conflict.

The competitive control includes a feedback control method and a feedforward control method. First, the feedback control will be described. What is feedback control?
When the transmission unit is transmitted to the network and feedback from the network is generated, the transmission of the next transmission unit is delayed to reduce the bandwidth usage r (t).
This is a method for resolving a race condition. Here, the transmission unit is a unit for performing transmission error detection or exchange in data transmission between computers. For example, packet or A
TM (Asynchronous Transfer)
Mode) cells.

Now, with the recent improvement in server performance and the increase in the amount of data exchanged between processes, there is a demand for an improvement in transmission throughput. In particular, in the case of a print service, the required transmission throughput is from several hundred mega bps to several giga bps due to the improvement of the print speed of the printer used for the server and the increase in the amount of image data per page.
ps, which is approaching the maximum transmission band Rmax.

In order to increase the transmission throughput, the bandwidth usage r (t) must be increased. Then, under the feedback control, the possibility of competition increases. Conflicts arise,
There is a problem that the delay is increased for each transmission unit, and as a result, the transmission throughput reaches a plateau.

Therefore, it has been proposed to apply feedforward control to transmission of a large amount of data such as a print service.

Next, the feed forward control will be described. In the feedforward control, when each client process generates multiple "requests" at the same time,
The data transmission is scheduled before the transmission is started so that the sum of the data transmissions accompanying the “request” does not exceed the maximum transmission band Rmax. Based on the scheduling result, the transmission start timing and the bandwidth usage for each data transmission r (t) is controlled, whereby the feed-forward control avoids the occurrence of contention during data transmission, and since no contention occurs for each transmission unit, a high transmission throughput is obtained.

On the other hand, in feedforward control,
When a plurality of clients simultaneously issue a "request" to the server, the "request" of each client waits for the transmission start timing until all of the previously scheduled data transmissions are completed.

When the time T for confirming the conditions for inter-process coordination is determined by the conventional method, the abnormal response processing is performed by the transmission start timing waiting time even though the partner process is operating. May be implemented. Also, the time required for the previously scheduled data transmission depends on the number of "requests" occurring simultaneously and the amount of data to be transferred. Even if the time T is increased, the upper limit cannot be specified.

The above-mentioned problem of the feedforward control will be specifically described.

FIG. 1 shows a conventional client-server type data transmission apparatus on a network for performing feed-forward type contention control. In FIG. 1, a conventional client-server type data transmission apparatus on a network for performing feedforward type competition control includes a client computer 10 and a server computer 20.
And a control device 30 for feedforward type competition control;
And a packet switching switch 40.
The client computer 10 includes a client process 101, an inter-process coordination timer 102, a request data transmission unit 103, and a response data reception unit 104. The server computer 20 has a server process 201, a response data transmitting unit 202, and a request data receiving unit 203. The control device 30 for feedforward control has a schedule unit 301. The packet switching switch 40 connects the client computer 10 and the server computer 20 to each other to execute data transfer by packet, and has a buffer 401.

When a data transfer request to the server process 201 occurs, the client process 101 passes the request data to the request data transmitting unit 103, and sets a specified value in the inter-process coordination timer 102 to reset the timer 102. to start. In addition, the client process 101 executes a server process 2 for request data.
When the response data from 01 is passed from the response data receiving unit 104, the inter-process cooperation timer 102 is reset. If the inter-process coordination timer 102 times out before receiving the response data, the client process 101 regards that a failure has occurred in the network or the server process 201.

Upon receiving the response data, the response data receiving unit 104 passes the response data to the client process 101.

The request data transmitting unit 103 is connected to the control unit 30
Is notified of the transmission request to the schedule unit 301.
When the transmission instruction is notified from the schedule unit 301,
Data transfer is performed in the transmission band of the instruction content.

When notified of the transmission request, the schedule unit 301 determines the transmission band and the transmission start timing from the request content and the transmission line band, and notifies the request data transmission unit 103 of this as a transmission instruction. I do. Transmission requests whose transmission start timings are delayed due to contention of a plurality of transmission requests are held in a queue.

Upon receiving the request data from the client process 101 from the request data receiving unit 203, the server process 201 processes the request data. At the same time, response data to the request data is passed to the response data transmission unit 202. The response data transmission unit 202 transmits the received response data to the client computer 10.

In this configuration, since the transfer timing of the request data and the response data changes according to the transmission request of another client, it is difficult to specify the timeout period. And failure of the network path cannot be detected, and as a result, inter-process coordination is not established.

As described above, there is a problem that coordination between processes cannot be realized when realizing a service requiring high throughput.

[0025]

SUMMARY OF THE INVENTION An object of the present invention is to establish inter-process coordination between a client process and a server process even during feed-forward contention control where a high throughput can be obtained.

[0026]

The inter-process coordination method according to the present invention controls a timer for inter-process coordination adaptively to a state of network contention. That is, according to the present invention, in order to achieve the above object,
Means for transmitting request data from one process to another process to a data transfer device for communicating between the processes on a plurality of computers connected via a packet switching network and performing inter-process coordination; Means for receiving response data sent from another process to the one process, an inter-process coordination timer for judging the success or failure of inter-process coordination, and checking the normality of the network for detecting occurrence of contention in the network Means and inter-process coordination timer control means for controlling the inter-process coordination timer based on the detection result of the normality checking means in a manner appropriate for the contention state of the network.

According to this configuration, the time measured by the timer for coordination between processes is varied according to the data transfer status of the network, so that coordination between processes can be ensured regardless of how the data transfer status of the network changes. I can do it.

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail.

FIG. 2 shows the overall configuration of the present invention. In FIG. 2, the present invention is constituted by the following two means. Network normality checking means 1 Inter-process coordination timer control means 2 First, the network normality checking means 1 will be described. Waiting for the transmission start timing is a normal state in the feedforward competition control. Therefore, a means for autonomously detecting the normality of the feedforward competition control is provided. This means is network normality checking means 1. This is a means of detecting abnormalities by yourself.
It may be possible to detect no abnormality. The process that has issued the "request" is notified of the normal period of waiting for the start of data transmission.

There are the following two methods for confirming the normality. Health Check The health check of the waiting time notification repeats an inquiry from the client to the data transfer scheduler and its response at regular intervals. In the waiting time notification, the data transfer scheduler notifies the client of the waiting time at regular intervals. Further, the data transfer scheduler may notify the client of the waiting time when the client makes an inquiry to the data transfer scheduler at an arbitrary timing.

Next, timer control means 2 for coordination between processes
Will be described. The process which has generated the "request" is provided with means for controlling a timer for counting the time T during this period. This means is the inter-process coordination timer control means 2.

There are the following three timer control methods. Suspension Reset The suspension of the setting of the waiting time detects the occurrence of network conflict and stops the timer for coordination between processes. When the conflict is resolved, the counting of the timer is restarted. Is to detect the occurrence of contention in the network and reset a timer for coordination between processes. When the conflict is resolved, the timer starts counting. The setting of the waiting time is to detect the transmission waiting time due to contention of the network, set the value to a timer for coordination between processes, and then start counting the timer.

[0033]

[Embodiment 1] FIG. 3 shows the overall configuration of a client-server type data transmission apparatus on a network for performing feed-forward type contention control according to Embodiment 1. 3, parts corresponding to those in FIG. 1 are denoted by corresponding reference numerals. 3 also includes a client computer 10, a server computer 20, a control device 30 for feedforward contention control, and a packet switching switch 40. Client computer 10
Has a client process 101, an inter-process coordination timer 102, a request data transmission unit 103, a response data reception unit 104, a timer control unit 105, and a competition status introduction unit 106. Server computer 2
0 is the server process 201 and the response data transmitting unit 2
02, and a request data receiving unit 203. The control device 30 of the feedforward control includes the schedule unit 3
01 and a competition status notification unit 302. The packet switching switch 40 is connected to the client computer 1
0 and the server computer 20 are connected to each other to execute data transfer by packets, and have a buffer 401.

In FIG. 3, client process 1
1, the request data transmitting unit 103, the response data receiving unit 102, the server process 201, the request data receiving unit 203, and the response data transmitting unit 202 are the same as those in the conventional example of FIG.

Next, the inter-process coordination timer 102 and the timer control unit 105 will be described. FIG.
Interprocess coordination timer 102 and timer control unit 10
5 shows the configuration of FIG. FIG. 5 shows an operation flow of the timer control unit 105.

In FIG. 4, an inter-process coordination timer 1
02 has an oscillator 111, a gate 112, and a counter 113. The oscillator 111 generates a clock signal. The gate 112 allows the clock signal to pass during the period when the count stop signal is not supplied from the timer control unit 105, and uses the count counter 113 as a count clock signal.
To supply. The clock signal is not passed while the count stop signal is being supplied. The count counter 113 is a down counter that uses a count clock signal as an operation clock. When a reset signal is given from the process, the counter value is set to a specified value. When a count start signal is given from the process, counting is started. When the counter value becomes 0, an expiration notification signal is sent to the process. While the count stop signal is being supplied from the timer control unit 105, the count is stopped because the operation clock is not supplied.

The timer control section 105 has a timer temporary stop function 114. Timer pause function 114
The normal standby display signal supplied from the competition status inquiry unit 106 is inspected (Step S10 in FIG. 5). In the case of normal standby, the count stop signal is asserted (S11, S12), and in the case of normal standby, the count stop signal is negated (S11).
13).

Next, the conflict status inquiry unit 106, the conflict status notification unit 302, and the schedule unit 301 will be described. FIG. 6 shows the configurations of the competition status inquiry unit 106, the competition status notification unit 302, and the schedule unit 301.

In FIG. 6, the competition status inquiry unit 106
Inquiry cycle generation function 115 and race condition request transmission function 11
6 and a competition status response receiving function 117. The inquiry cycle generation function 115 generates an inquiry timing signal at a predetermined fixed cycle, and supplies it to the competition situation request transmission function 116 and the competition situation response reception function 117. The race condition request transmission function 116 generates a race condition request when the inquiry timing signal is supplied from the inquiry cycle generation function 115, and notifies the race condition notification unit 302. The conflict status response receiving function 117 receives a conflict status response from the conflict status notification unit 302. The conflict status response is checked, and if a conflict occurs, a normal standby display signal is asserted to the inter-process coordination timer control unit 105 (FIG. 4). If no conflict occurs, the normal standby display signal is negated. Also, the normal standby display signal is negated when the competition status response is not received for a period equal to or longer than a predetermined integer multiple of the cycle time of the timing signal.

The competition status notification unit 302 has a competition status request reception function 311 and a competition status response transmission function 312. Upon receiving the competition status request from the competition status inquiry unit 106, the competition status request receiving function 311 notifies the schedule unit 301 of a waiting status inquiry. Upon being notified of the waiting status from the schedule unit 301, the contention response transmission function 312 transmits a contention response to the client computer 10 (FIG. 3) that transmitted the contention request.

The schedule section 301 has a transmission path band allocation function 313 and a data transfer queue function 314. When notified of the transmission request, the transmission path band allocation function 313 determines the contents of the transmission instruction including the transmission band from the contents of the transmission request and the transmission path band, and notifies the data transfer queue function 314. Data transfer queue function 314
Is notified of the content of the transmission instruction, and adds it to the queue. For the contents of the transmission instruction at the head of the queue, a transmission instruction is generated, and the request is transmitted to the request data transmission unit 103 of the client computer 10 which has issued the transmission request.
The transmission instruction is held in the queue until the data transfer indicated by the transmission instruction is completed. Further, when the data transfer queue function 314 is notified of the wait status inquiry from the contention status request reception function 311, it notifies the number of unexecuted transmission instruction contents in the queue as the wait status.

The control device 30 for feedforward control and the server computer 20 may be the same.

[Second Embodiment] FIG. 7 shows an overall configuration of a client-server type data transmission apparatus on a network for performing feedforward type competition control according to a second embodiment. 7, parts corresponding to those in FIG. 1 are denoted by corresponding reference numerals.

The client-server type data transmission device of FIG. 7 also includes a client computer 10 and a server
It is configured to include a computer 20, a control device 30 for feedforward type conflict control, and a packet switching switch 40. The client computer 10
It has a client process 101, an inter-process coordination timer 102, a request data transmission unit 103, a response data reception unit 104, a timer control unit 105, and a competition status introduction unit 106. Server computer 20
Is the server process 201 and the response data transmitting unit 20
2 and a request data receiving unit 203. The control device 30 for feedforward control includes a scheduler 30
1 and a competition status notification unit 302. The packet switching switch 40 is connected to the client computer 10.
The server computer 20 is connected to each other to execute data transfer by packet, and has a buffer 401.

In FIG. 7, client process 1
1, the request data transmitting unit 103, the response data receiving unit 102, the server process 201, the request data receiving unit 203, and the response data transmitting unit 202 are the same as those in the conventional example of FIG.

Next, the inter-process coordination timer 102 and the timer control unit 105 will be described. FIG.
Interprocess coordination timer 102 and timer control unit 10
5 shows the configuration of FIG. FIG. 9 is a flowchart showing the operation of the timer control unit 105.

In FIG. 8, timer 1 for coordination between processes is used.
02 has an oscillator 111 and a counter 112. The oscillator 111 generates a count clock signal.
The count counter 113 is a down counter that uses a count clock signal as an operation clock. When a reset signal is given from the process, the counter value is set to a specified value.
When a count start signal is given from the process, counting is started. When the counter value becomes 0, an expiration notification signal is sent to the process. Also, when a reset signal including a set value is provided from the timer control unit 105, the set value is compared with the specified value. If the set value is larger than the specified value, the counter value is set to the set value, and the counting is performed. Start. Ignore otherwise.

The timer control unit 105 has a timer reset function 118. The timer resetting function 118 checks the normal standby display signal supplied from the competition status inquiry unit 106 (FIG. 8) (S20), and stores the standby time value included in the normal standby display signal. If the standby time value included in the normal standby display signal is the same as the stored previous value, it is ignored (S21). If it is different from the previous value, the stored standby time value is updated, and a reset signal including the new standby time value as the installation value of the inter-process coordination timer 102 is generated, and the inter-process coordination timer is generated. 102 is notified (S22).

Next, the conflict status inquiry unit 106, the conflict status notification unit 302, and the schedule unit 301 will be described. FIG. 10 shows the configurations of the competition status inquiry unit 106, the competition status notification unit 302, and the schedule unit 301.

In FIG. 10, the competition status inquiry unit 106
Has a conflict status notification receiving function 119. The conflict status notification receiving function 119 receives a conflict status notification from the conflict status notification unit 302. The received competition status notification is checked, and a normal standby display signal including a standby time value is generated and supplied to the timer control unit 105 (FIG. 7).

The competition status notification section 302 has a notification cycle generation function 315 and a competition status notification transmission function 312. The notification cycle generation function 315 generates a notification timing signal at a predetermined fixed cycle, and transmits a race condition notification transmission function 312.
To supply. The race condition transmission function 312 notifies the schedule unit 301 of a standby time inquiry when the notification timing signal is supplied from the notification cycle generation function 315. When notified of the standby time from the schedule unit 301, a competition status notification including the standby time is generated,
Send to all client computers.

The schedule section 301 has a transmission path band allocation function 313 and a data transfer queue function 314. When notified of the transmission request, the transmission path band allocation function 313 determines the contents of the transmission instruction including the transmission band from the contents of the transmission request and the transmission path band, and notifies the data transfer queue function 314. Data transfer queue function 314
Is notified of the content of the transmission instruction, and adds it to the queue. A transmission instruction is generated for the contents of the transmission instruction at the head of the queue, and the request is transmitted to the request data transmission unit 103 (FIG. 7) of the client computer 10 that has issued the transmission request. The transmission instruction is held in the queue until the data transfer indicated by the transmission instruction is completed. In addition, when the data transfer queue function 314 is notified of the standby time inquiry from the contention status notification transmission function 312, the data transfer queue function 314 calculates the total data transfer required time from all the transmission instruction contents in the queue and notifies it as the standby time. .

The description of the embodiment has been completed. It should be noted that the present invention is not limited to the above-described embodiment, and various changes can be made without departing from the spirit of the present invention. For example, application of the present invention is not limited to feedforward type congestion control. The present invention can also be applied to data transfer for controlling the bandwidth in accordance with congestion and datagram transfer without error detection / retransmission.

[0054]

As described above, according to the present invention,
There is an effect of controlling a timer for inter-process coordination adaptively to the contention state of the network. As a result, inter-process cooperation between the client process and the server process is established even at the time of feedforward contention control at which a high throughput is obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the overall configuration of a conventional client-server data transfer device.

FIG. 2 is a diagram illustrating an overall configuration of the present invention.

FIG. 3 is a block diagram illustrating an overall configuration of a client-server data transfer device according to the first embodiment.

FIG. 4 is a block diagram illustrating a configuration of an inter-process coordination timer and a timer control unit according to the first embodiment.

FIG. 5 is a flowchart illustrating an operation of a timer control unit according to the first embodiment.

FIG. 6 is a block diagram illustrating a configuration of a competition status inquiry unit, a competition status notification unit, and a schedule unit according to the first embodiment.

FIG. 7 is a block diagram illustrating an overall configuration of a client-server data transfer device according to a second embodiment.

FIG. 8 is a block diagram illustrating a configuration of an inter-process coordination timer and a timer control unit according to a second embodiment.

FIG. 9 is a flowchart illustrating an operation of a timer control unit according to the second embodiment.

FIG. 10 is a block diagram illustrating a configuration of a competition status inquiry unit, a competition status notification unit, and a schedule unit according to the second embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Inter-process coordination timer control means 2 Network normality detection means 10 Client computer 20 Server computer 30 Feed-forward type contention control controller 40 Packet exchange switch 101 Client process 102 Inter-process coordination timer 103 Request data transmission Unit 104 response data receiving unit 105 timer control unit 106 competition status inquiry unit 201 server process 202 response data transmission unit 203 request data reception unit 301 schedule unit 302 competition notification unit

Claims (8)

[Claims] 1. A data transfer device for processes on a plurality of computers connected via a packet switching network to communicate with each other and perform inter-process coordination, wherein request data is transmitted from one process to another remote process. Means for receiving response data sent from the other process to the one process; a timer for inter-process coordination for judging the success or failure of inter-process coordination; A network normality check unit to be detected, and an inter-process coordination timer control unit that controls the inter-process coordination timer in a manner appropriate for the contention state of the network based on the detection result of the normality check unit. A data transfer device, characterized in that: 2. The data transfer device according to claim 1, wherein the normality check unit performs a health check in which an inquiry to a data transfer scheduler and a response to the inquiry are received at regular intervals. 3. The data transfer device according to claim 1, wherein the normality check unit performs a wait time notification for acquiring and notifying the wait time information from the data transfer scheduler at regular intervals. 4. The data transfer device according to claim 1, wherein the normality check unit notifies the data transfer scheduler of a waiting time upon an inquiry to the data transfer scheduler. 5. The inter-process coordination timer control means temporarily stops the inter-process coordination timer when the data transfer of the request is in a transfer standby state because the network is in a contention state. 5. The data transfer device according to claim 1, wherein the counting of the inter-process coordination timer is restarted after the conflict is resolved. 6. The inter-process coordination timer control means resets the inter-process coordination timer when a data transfer of a request is in a transfer standby state because the network is in a contention state, and 2. The counting of the inter-process coordination timer is started after the cancellation.
5. The data transfer device according to 2, 3, or 4. 7. The inter-process coordination timer control means detects a transmission waiting time when a data transfer of a request is in a transfer standby state because the network is in a contention state, and 5. The data transfer device according to claim 1, wherein the value is set in a timer, and counting of the inter-process coordination timer is started. 8. A data processing device in which processes on a plurality of computers connected via a packet switching network communicate with each other and cooperate between processes, wherein request data is transmitted from one process to another remote process. Means for receiving response data sent from the other process to the first process; an inter-process coordination timer for judging the success or failure of inter-process coordination; and detecting occurrence of contention in the network. A network normality check unit, and an inter-process coordination timer control unit that controls the inter-process coordination timer, based on a detection result of the normality check unit, in an adaptive manner to the contention state of the network. Characteristic data processing device.