JP7143609B2 - COMMUNICATION DEVICE, COMMUNICATION METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to a communication device, communication method, and program.

In recent years, due to the increasing need for enormous amount of calculation such as learning of deep learning, large-scale systems connecting many PCIe (PCI Express) devices such as GPGPU (General-purpose computing on graphics processing units) are being constructed. . It is required to prevent the failure from spreading to other PCIe devices when a failure occurs in one PCIe device. In general, when a failure occurs, the PCIe device notifies the host server of the failure, and the host server performs failure processing of the failed device. Japanese Patent Application Laid-Open No. 2002-200001 discloses a related technique for performing network control based on congestion on the network.

Paragraph 0005 of Patent Document 1 describes a credit-based system. The credit-based system, as described in paragraph 0005 of the document, is a technology that ensures a reduction in transmission data loss, and the state of the data receiver and the relationship between the sender and receiver. It is controlled by considering the state of all nodes. Specifically, the state of the node on the receiving side is notified to the transmitting side by sending credits. Credits indicate the ability to receive data and are initially generated by the receiving node. This credit is then returned upstream to the sending node. Then, at the sending node, the ability of the receiving node to receive the data is determined.

JP-A-8-56222

By the way, when a failure occurs in a credit-based system using a PCIe (PCI Express) device as described above, packets that perform failure-related processing flow through the same route as normal packets, so if a failure such as packet clogging occurs It is not possible to process failures. Also, in such a case, packets of normal PCIe devices are not processed, and if a failure occurs in one PCIe device, there is a problem that the failure spreads to other PCIe devices.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a communication device, a communication method, and a program that solve the above problems.

According to the first aspect of the present invention, a communication device sends transmission data of a data amount based on a receivable data amount of a counterpart communication device connected via a relay device to the counterpart communication device via a PCI (Peripheral) a data transmission unit that transmits based on the communication standard of Component Interconnect (Component Interconnect) Express; a failure detection unit that detects whether a failure has occurred in a communication path between the transmission data and the communication device on the opposite side ; and a disconnect determination unit that determines to disconnect the communication path when it is detected.

According to a second aspect of the present invention, a communication method transmits transmission data of a data amount based on a receivable data amount of a counterpart communication device connected via a relay device to the counterpart communication device via PCI (Peripheral). (Component Interconnect) Express communication standard, detects whether there is a failure in the communication path between the transmission data and the communication device of the other party, and if the failure is detected, the communication path It is characterized in that it is determined that the connection between the

According to the third aspect of the present invention, the program causes the computer of the communication device to send the transmission data of the data amount based on the receivable data amount of the communication device of the other party connected via the relay device to the communication device of the other party. data transmission means for transmitting data to a device based on the PCI (Peripheral Component Interconnect) Express communication standard; fault detection means for detecting whether a fault has occurred in a communication path between the transmission data and the other party communication device ; It is characterized by functioning as connection means for disconnecting the communication path when the failure is detected.

ADVANTAGE OF THE INVENTION According to this invention, it can suppress that a failure spreads to other communication apparatuses.

1 is a block diagram showing the configuration of a communication system according to one embodiment of the present invention; FIG. 1 is a diagram showing the hardware configuration of a communication device according to one embodiment of the present invention; FIG. FIG. 3 is a functional block diagram of a communication module according to one embodiment of the present invention; FIG. 4 is a first diagram showing functional blocks of a failure detection unit according to one embodiment of the present invention; FIG. 2 is a first diagram showing a processing flow of a client-side communication device according to an embodiment of the present invention; FIG. 10 is a second diagram showing functional blocks of the failure detection unit according to one embodiment of the present invention; FIG. 4B is a second diagram showing the processing flow of the client-side communication device according to the embodiment of the present invention; 1 is a diagram showing the minimum configuration of a communication device according to one embodiment of the present invention; FIG.

A communication system according to an embodiment of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing the configuration of a communication system according to the same embodiment.
As shown in FIG. 1 , in the communication system, a host-side communication device 100 and a client-side communication device 300 are connected via a relay device 200 . The communication device 100, the relay device 200, and the communication device 300 are each equipped with a communication device having a PCIe (PCI Express) communication function, and perform communication using the communication function defined by the PCIe (PCI Express). Note that the communication function specified by PCIe (PCI Express) performs credit-based communication. Specifically, the transmitting side detects the amount of data that can be received by the communication device of the other party, and transmits the transmission data of the amount based on the amount of data that can be received to the communication device of the other party. For example, the transmitting side detects the amount of data that can be received by the communication device of the other party, and transmits transmission data having a data amount equal to or less than the amount of data that can be received by the communication device of the other party. As a result, the partner communication device can receive and process only the amount of transmission data that can be received by itself.

A client-side communication device 300 constituting a communication system detects a failure in a communication path, and determines to disconnect from the communication path when a failure is detected. Then, the communication device 300 on the client side disconnects from the communication path. The communication system according to the present embodiment is a communication system in which, in principle, packets are not discarded because communication control is performed on a credit basis. However, in the communication system according to the present embodiment, the communication device 300 that detects a failure disconnects from the communication path, and the packet is discarded by the relay device 200 or the like, thereby resolving congestion such as packet clogging. . As a result, the communication device 300 on the client side is characterized by preventing the failure from spreading to other devices.

FIG. 2 is a diagram showing the hardware configuration of the communication device.
As shown in FIG. 2 , the client-side communication device 300 includes a CPU (Central Processing Unit) 301 , a ROM (Read Only Memory) 302 , a RAM (Random Access Memory) 303 and a communication module 304 . The communication module communicates with other devices by the communication function of PCIe (PCI Express).
It is assumed that the communication device 100 and the relay device 200 on the host side also have the same hardware configuration.

FIG. 3 is a functional block diagram of a communication module provided in the communication device.
As shown in FIG. 3, the communication module 304 includes functional units such as a data transmission unit 31, a link control unit 32, a failure detection unit 33, and a disconnection determination unit .
The data transmission unit 31 transmits to the communication device 100 the amount of transmission data based on the amount of data that can be received by the communication device 100 on the host side, which is the other communication device. In other words, the data transmission unit 31 performs communication control with the communication partner based on the credit-based communication regulations.
The fault detection unit 33 detects whether a fault has occurred in the communication path of the transmission data.
The disconnect determination unit 34 determines to disconnect the communication path when a failure is detected.
The link control unit 32 performs link control with a communication partner. For example, when a failure is detected by the disconnection determination unit 34, the connection with the communication path connected to the own device is disconnected.

FIG. 4 is a first diagram showing functional blocks of the failure detection unit.
The fault detector 33 includes a request signal transmitter 331 , a response signal determiner 332 and a monitor 333 .
The request signal transmission unit 331 transmits the first request signal to the communication device 100 on the host side, which is the other communication device.
The response signal determination unit 332 determines whether a response signal to the first request signal has been received from the communication device 100 on the host side.
The monitoring unit 333 determines whether the time from the transmission time of the first request signal to the reception time of the response signal is within a predetermined time measured by a timer.

FIG. 5 is a first diagram showing the processing flow of the communication device on the client side.
Next, the processing flow of the communication device 300 on the client side will be described step by step.
First, the monitoring unit 333 outputs to the request signal transmitting unit 331 a request to transmit the first request signal to the communication device 100 on the host side. The monitoring unit 333 resets the timer from the time when the transmission request is made, and then activates the timer. The timer is a timer that counts down a predetermined time. The predetermined time may be, for example, a time such as 3 seconds or 5 seconds.

The request signal transmission unit 331 acquires the transmission request of the first request signal. The request signal transmission unit 331 transmits the first request signal to the communication device 100 based on this transmission request (step S101). The first request signal is transmitted based on the PCI Express communication protocol used by the link control unit 32 .

The communication device 100 that has received the first request signal transmits a response signal to the communication device 300 . The response signal determination unit 332 receives the response signal based on the PCI Express communication protocol used by the link control unit 32 (step S102). When the response signal determining section 332 determines that the response signal has been received, it outputs a signal indicating that the response signal has been received to the monitoring section 333 . The monitoring unit 333 determines whether or not the response signal is received before the time counted by the timer expires (step S103). If the response signal is received before the time counted by the timer expires, the monitoring unit 333 repeats the process of outputting the transmission request of the first request signal. When the response signal is received after the time counted by the timer expires or when no response signal is received, the monitoring unit 333 determines that a failure has occurred (step S104). If the reception time of the response signal coincides with the count expiration time of the timer, it may be determined that no fault has occurred.

When the monitoring unit 333 determines that a failure has occurred, it notifies the link control unit 32 of the failure occurrence. Upon receiving the failure notification, the link control unit 32 generates a control signal for disconnecting the communication path of the communication network cable connected to the own device. The link control unit 32 disconnects the communication path based on the control signal (step S105). The link control unit 32 may generate a reset signal and temporarily disconnect the communication path based on the reset signal. It is desirable that the count time of the timer of the monitoring unit 333 is the same as the time indicated by the timeout value set in the link control unit 32 . When the monitoring unit 333 determines that a failure has occurred, the monitoring unit 333 may notify the communication device 100 on the host side of the failure by interrupting or the like.

The communication device 300 that performs such processing cuts off the communication connection of the communication path when detecting a link failure such as a packet clogging. As a result, a communication system having each communication device equipped with a communication device having a PCIe (PCI Express) communication function as shown in FIG. The device 200 discards the transmission data transmitted from the communication device 300 with which communication has been cut off. Therefore, it is possible to avoid problems such as clogging of packets, which are transmission data, and prevent the problems from spreading to other communication devices 300 .

Further, the communication device 300 attempts reconnection even after disconnecting from the communication path based on detection of the occurrence of a failure. In other words, the failure detection unit 33 repeatedly detects whether a failure has occurred in the communication path after the connection with the communication path is disconnected (step S106). The link control unit 32 controls reconnection with the host-side communication device 100 connected to the communication path when there is no failure in the communication path (step S107). As a result, the communication between the communication device 100 on the host side and the communication device 300 on the client side can be resumed. The failure detection unit 33 determines whether the processing of the communication device 300 is finished (step S108). The communication device 300 repeats the processing from step S101 if it is not finished.

FIG. 6 is a second diagram showing functional blocks of the failure detection unit.
The failure detection unit 33 includes a monitoring unit 333 , a request signal reception unit 334 and a storage unit 335 .
The request signal receiving unit 334 receives the second request signal from the communication device 100, which is the other communication device.
The storage unit 335 stores the identification information of the second request signal.
The monitoring unit 333 determines whether the time from the reception time of the second request signal to the reception time of the next second request signal is within the predetermined time measured by the timer.
The fault detection unit 33 detects that a fault has occurred in the communication path when the next second request signal cannot be received within a predetermined time.

The failure determination processing described with reference to FIGS. 4 and 5 may be replaced with the processing shown in FIGS. 6 and 7 as follows.
The host-side communication device 100 transmits a second request signal to the communication device 300 at preset time intervals. As for the second request signal, for example, an identification number is recorded in the signal. Each time the host-side communication device 100 transmits the second request signal, it increments the identification number included in the signal by one.

The request signal receiving unit 334 receives the second request signal (step S201). The second request signal is received based on the PCI Express communication protocol used by the link control unit 32 . The request signal receiving unit 334 records the identification number and the reception time included in the second request signal in the storage unit 335 (step S202).

The monitoring unit 333 periodically reads the identification number of the second request signal and the reception time recorded in the storage unit 335 . The monitoring unit 333 counts the timer based on the first reception of the second request signal, and determines whether the reception time of the next second request signal is before the expiration of the count time (step S203). If the second request signal is received before the time counted by the timer expires, the monitoring unit 333 repeats monitoring of the information included in the second request signal. The monitoring unit 333 determines that a failure has occurred when the reception time of the next second request signal is after the time counted by the timer expires or when the next second request signal is not received (step S204). If the reception time of the second request signal coincides with the count expiration time of the timer, it may be determined that no fault has occurred.

When the monitoring unit 333 determines that a failure has occurred, it notifies the link control unit 32 of the failure occurrence. Upon receiving the failure notification, the link control unit 32 generates a control signal for disconnecting the communication path of the communication network cable connected to the own device. The link control unit 32 disconnects the communication path based on the control signal (step S205).

The communication device 300 that performs such processing cuts off the communication connection of the communication path when detecting a link failure such as a packet clogging. As a result, a communication system having each communication device equipped with a communication device having a PCIe (PCI Express) communication function as shown in FIG. The device 200 discards the transmission data transmitted from the communication device 300 with which communication has been cut off. Therefore, it is possible to avoid problems such as clogging of packets, which are transmission data, and prevent the problems from spreading to other communication devices 300 .

Further, the communication device 300 attempts reconnection even after disconnecting from the communication path based on detection of the occurrence of a failure. In other words, the monitoring unit 333 of the failure detection unit 33 detects whether or not the identification number of the next second request signal is recorded in the storage unit after the connection with the communication path is disconnected (step S206). If the identification number of the next second request signal is recorded, the link control unit 32 determines that no failure has occurred in the communication path, and reconnection control (step S207). As a result, the communication between the communication device 100 on the host side and the communication device 300 on the client side can be resumed. The failure detection unit 33 determines whether the processing of the communication device 300 ends (step S208). The communication device 300 repeats the processing from step S201 if the processing is not finished.

The embodiment of the present invention has been described above. According to the above-described processing, it is possible to avoid problems such as jamming of packets, which are transmission data, and prevent the problems from spreading to other communication devices 300. .

FIG. 8 is a diagram showing the minimum configuration of a communication device.
The communication device 300 may include at least the data transmission unit 31, the failure detection unit 33, and the disconnection determination unit .
The data transmission unit 31 transmits to the communication device 100 the amount of transmission data based on the amount of data that can be received by the communication device 100 on the host side, which is the other communication device.
The fault detection unit 33 detects whether a fault has occurred in the communication path of the transmission data.
The disconnect determination unit 34 determines to disconnect the communication path when a failure is detected.

Each of the devices described above has an internal computer system. Each process described above is stored in a computer-readable recording medium in the form of a program, and the above process is performed by reading and executing this program by a computer. Here, the computer-readable recording medium refers to magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, and the like. Alternatively, the computer program may be distributed to a computer via a communication line, and the computer receiving the distribution may execute the program.

Further, the program may be for realizing part of the functions described above. Further, it may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system.

100, 300... Communication device 200... Relay device 31... Data transmission unit 32... Link control unit 33... Failure detection unit 34... Disconnection determination unit 331... Request signal transmission unit 332... Response signal determination unit 333... Monitoring unit 334... Request signal receiving unit 335... Storage unit

Claims (6)

A data transmission unit that transmits transmission data of a data amount based on the amount of data that can be received by a counterpart communication device connected via a relay device to the counterpart communication device based on the PCI (Peripheral Component Interconnect) Express communication standard. When,
a failure detection unit that detects whether a failure has occurred in a communication path between the transmission data and the communication device of the other party ;
a disconnect determination unit that determines to disconnect the communication path when the failure is detected;
A communication device comprising: a request signal transmission unit that transmits a first request signal to the counterpart communication device,
The communication device according to claim 1, wherein the failure detection unit detects that a failure has occurred in the communication path when a response signal to the first request signal cannot be received. a request signal receiving unit that receives a second request signal from the counterpart communication device,
The communication device according to claim 1, wherein the failure detection unit detects that a failure has occurred in the communication path when the second request signal cannot be received. The failure detection unit repeatedly detects whether a failure has occurred in the communication path after the connection with the communication path is disconnected,
4. The device according to any one of claims 1 to 3, further comprising a link control unit that controls reconnection with the communication device connected to the communication path when no failure has occurred in the communication path. The communication device according to . transmitting data amount based on the amount of data that can be received by a counterpart communication device connected via a relay device to the counterpart communication device based on PCI (Peripheral Component Interconnect) Express communication standards;
detecting whether a failure has occurred in a communication path between the transmission data and the communication device of the other party ;
A communication method that determines to cut the connection with the communication path when the failure is detected. the computer of the communication device,
Data transmission means for transmitting transmission data having a data amount based on the amount of data that can be received by the communication device of the other party connected via the relay device to the communication device of the other party based on the communication standard of PCI (Peripheral Component Interconnect) Express. ,
failure detection means for detecting whether or not a failure has occurred in a communication path between the transmission data and the counterpart communication device ;
connection means for disconnecting the communication path when the failure is detected;
A program that acts as a

Images