JP2008148181A - Communication apparatus and communication control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To easily achieve overhead reduction in software. <P>SOLUTION: The disclosed method includes the steps of: generating address information containing information about an address of a first buffer for storing data to be transmitted and about a length of the data to be transmitted, and storing the generated information in a second buffer; copying the address information from the second buffer to a third buffer, adding a first control header and generating a first packet; copying the first packet from the third buffer to a fourth buffer within a communication apparatus without interposing a processor; copying the data to be transmitted from the first buffer to the fourth buffer without interposing the processor on the basis of the address information, adding a second control header, and generating a second packet; and transmitting the first packet and the second packet stored in the fourth buffer to a transmission destination terminal. Thus, information is copied without interposing the processor, so that the overhead of software can be reduced. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a communication control technique, and more particularly to a communication technique for reducing software overhead.

  In a computer system in which a plurality of computers are connected to a network, data may be lost during transmission. In this case, a mechanism for guaranteeing reliability on the network as to whether or not the data has been reliably transmitted is required. For example, the reliability on the network is ensured by discarding or correcting the data when an error is detected from the received data, or retransmitting the data when the data does not reach the transmission destination. Methods for realizing this reliability guarantee are roughly divided into a method realized by hardware and a method realized by software. InfiniBand (registered trademark), FibreChannel, and the like are used as a method for realizing reliability assurance by hardware. However, these have the problem that the hardware is complicated and expensive.

  On the other hand, in the case of using hardware whose reliability is not guaranteed (for example, Ethernet (registered trademark)), the reliability of communication is realized by software. Protocols such as TCP / IP (Transmission Control Protocol / Internet Protocol) are widely used as a method for realizing communication reliability assurance by software. In TCP / IP, etc., when data arrives, an acknowledgment (ACK packet) is returned from the transmission destination to the transmission source. If no ACK packet is returned within a predetermined time, the data is retransmitted. Thus, communication reliability is guaranteed. Although communication reliability is not guaranteed, Ethernet has the advantage of being cheaper than InfiniBand. However, as described below, the conventional method for realizing communication reliability guarantee by software has a problem that software overhead increases.

  A conventional system for realizing communication reliability guarantee by software will be briefly described with reference to FIGS. 1 and 2. FIG. 1 shows an outline of conventional packet transmission / reception processing. In FIG. 1, it is assumed that the transmission side computer and the reception side computer are connected by, for example, Ethernet. Each of the transmission side computer and the reception side computer includes a user program, an OS (Operating System), and a NIC (Network Interface Card). The user program, OS, and NIC in the transmission side computer are referred to as a transmission side user program, a transmission side OS, and a transmission side NIC, respectively. In addition, the user program, OS, and NIC in the receiving computer are referred to as a receiving user program, a receiving OS, and a receiving NIC, respectively. In FIG. 1, it is assumed that communication reliability is guaranteed by the TCP / IP protocol implemented in the OS.

  First, an outline of processing for conventional packet transmission will be described. The transmission-side user program requests the transmission-side OS to transmit user data (FIG. 1: step (1)). At this time, the transmission-side user program designates an address indicating an area in which user data to be transmitted is stored, a length of the user data, and an end point. Then, the transmission side OS copies the user data to be transmitted to the internal buffer of the transmission side OS to generate a packet frame (step (2)). At this time, the transmission side OS adds a control header.

  2A and 2B show an example of a format of a packet frame transmitted / received on the Ethernet. FIG. 2A shows an example of a basic format of a packet frame transmitted / received on the Ethernet. In FIG. 2A, a packet frame is composed of an Ethernet header and a data part. The Ethernet header includes information on a transmission destination, a transmission source, and a protocol type, and the data part includes a data part. User data is included from the beginning. In the example of FIG. 2A, the transmission destination, transmission source, and protocol type are collectively referred to as a control header. On the other hand, FIG. 2B shows an example of a packet frame format when “IP” is set as the protocol type of the Ethernet header. The point composed of the Ethernet header and the data part is the same as the example of FIG. 2A, but differs from FIG. 2A in that an IP header and a TCP header are included at the head of the data part. In the example of FIG. 2B, the transmission destination, transmission source, protocol type, IP header, and TCP header are collectively referred to as a control header. Note that FIG. 2B is an example in which a TCP header is included, and may include a header related to a protocol other than TCP.

  After generating the packet frame, the transmission side OS notifies the transmission side NIC of the packet transmission request. When receiving the transmission request from the transmission side OS, the transmission side NIC reads the packet frame from the internal buffer of the transmission side OS by DMA transfer (step (3)). Then, the transmission side NIC transmits a packet frame (step (4)).

  Next, an outline of conventional packet reception processing will be described. The receiving-side user program requests the receiving-side OS to receive user data (step (5)). At this time, the receiving-side user program specifies the address of the data reception buffer, the data length, and the end point. Then, the reception-side OS waits for a reception notification from the reception-side NIC. Further, when receiving the packet frame, the receiving side NIC writes the packet frame in the internal buffer of the receiving side OS by DMA transfer, and notifies the receiving side OS that the packet frame has been received (step (6)). When receiving the reception notification from the receiving-side NIC, the receiving-side OS copies the data (that is, user data) excluding the control header from the packet frame stored in the internal buffer of the receiving-side OS to the data receiving buffer (Step S1). (7)).

  Also, the receiving side OS generates an ACK packet frame, stores it in the internal buffer of the receiving side OS, and notifies the receiving side NIC of the packet transmission request (step (8)). When receiving the transmission request from the receiving OS, the receiving NIC reads the packet frame (ACK) from the internal buffer of the receiving OS by DMA transfer (step (9)). Then, the receiving side NIC transmits a packet frame (ACK) (step (10)). Upon receiving the packet frame (ACK), the transmission side NIC writes the packet frame (ACK) in the internal buffer of the transmission side OS by DMA transfer, and notifies the transmission side OS that the packet frame (ACK) has been received ( Step (11)). When the transmission side OS receives the reception notification from the transmission side NIC, the transmission side OS confirms the ACK packet frame (step (12)). Since the transmission-side OS has confirmed that the packet frame has reached the transmission destination without any problem, the transmission-side OS performs processing such as deleting the packet frame from the internal buffer of the transmission-side OS.

  As described above, in the conventional method for guaranteeing communication reliability by software, a copy of the user data (step (2) and step (7)) is generated between the user program and the OS. Increases the overhead. In particular, when a large amount of user data is transmitted, a CPU (Central Processing Unit) load due to copy processing increases. In recent years, although the communication speed on the network has been greatly improved, the inherent performance of the network cannot be exhibited due to the large software overhead.

  For example, there is a TOE (TCP Offload Engine) as a conventional technique for dealing with such a problem. In the TOE, since the copy process is performed by the NIC, the CPU load due to the copy process can be reduced. However, the TOE does not need to change the user program at all, but must change the OS to correspond to the TOE. Furthermore, there is a problem that the NIC is complicated and expensive.

  Further, as another conventional technique for dealing with the above problem, for example, there is RDMA (Remote Direct Memory Access). In RDMA, since the NIC directly reads out user data, copy processing can be reduced. However, the user program must be changed to support RDMA. Furthermore, since the reliability of communication is not guaranteed in RDMA, a mechanism for ensuring the reliability of communication must be installed in the NIC, and there is still a problem that the NIC becomes complicated and expensive.

  Also, for example, Japanese Patent Laid-Open No. 2005-100230 discloses a device interface that includes a physical layer, a link layer, a transport layer, and an application layer, and transfers commands and data in a packet format with a host by serial transmission. A device that is provided in the transport layer and stores a command packet or a data packet received from the host via the physical layer and the link layer in a first-in first-out manner, and is stored in the reception FIFO during data transfer. A command detection circuit that detects a received command and outputs a command detection signal; a reception task file register that loads a command content of a reception FIFO provided in an application layer; and a transport layer that is provided for a packet transmission Load command or data A task file register for transmission and a transmission FIFO provided in the transport layer, storing the contents of the task file register for transmission in a first-in first-out manner, and transmitting a command packet or a data packet to the host via the link layer and the physical layer; During the data transfer, a free time generation unit that generates a free time for receiving another command packet from the host, and when a command packet is received from the host during the free time, the data transfer is suspended and the received command is An interface device is disclosed that includes an in-transfer command processing unit that resumes data transfer after decoding and executing processing. However, a technique that solves the above problem is not disclosed.

Also, for example, Japanese Patent Publication No. 10-505700 discloses a network device that interfaces between a local area network and a network peripheral device, from the physical network layer to the peripheral device via an interrupt-driven data link layer. With a protocol stack that extends to the serving application layer, the data link layer is activated by a unique dump / debug packet header, interprets dedicated debug commands contained in the data portion of the packet, and in interrupt deactivation mode A network device having a module for executing a dump / debug command is disclosed. However, no technology that solves the above problem is disclosed.
JP 2005-100230 A Japanese National Patent Publication No. 10-505700

  As described above, according to the conventional technique, it is possible to reduce the overhead of software generated when the reliability of communication is ensured by software. However, the conventional technique has a problem that the NIC is complicated and expensive. Furthermore, in the prior art, since the conventional OS cannot be used as it is, it is necessary to change to a dedicated configuration. In general, it is not easy to change the OS, and labor and knowledge are required.

  Therefore, an object of the present invention is to provide a technique for easily realizing a reduction in software overhead that occurs when the reliability of communication is ensured by software.

  Another object of the present invention is to provide a technique for realizing a reduction in software overhead that occurs when a communication reliability guarantee is realized by software without changing a conventional OS.

  A communication control method according to a first aspect of the present invention is a communication control method executed by a computer having a processor and a communication device, and includes an address of a first buffer for storing data to be transmitted and data to be transmitted. Address information including the length information, and storing the address information in the second buffer, copying the address information from the second buffer to the third buffer, adding a first control header, A first packet generating step for generating a packet; a copy processing step for copying the first packet from the third buffer to the fourth buffer in the communication device without using a processor; and data to be transmitted based on the address information. The second packet is copied from the first buffer to the fourth buffer without passing through the processor, and the second control header is added. It includes a second packet generating step of generating, and transmitting step of transmitting a first packet and a second packet stored in the fourth buffer to the destination terminal.

  In this way, for example, even when a large amount of data is transmitted, since the address information is copied to the third buffer, the CPU load due to the copy process can be reduced. Further, since the data to be transmitted is copied from the first buffer to the fourth buffer in the communication device without passing through the processor (for example, CPU), no load is applied to the CPU. Furthermore, since the communication device (for example, NIC) does not have a mechanism for guaranteeing communication reliability, it is possible to prevent the communication device from becoming complicated. Therefore, it is possible to easily reduce the software overhead.

  Further, the second packet generation step may include a step of copying the first packet and adding it to the second packet. In this way, the transmission destination terminal can detect the second packet corresponding to the first packet or the first packet corresponding to the second packet.

  A communication control method according to a second aspect of the present invention is a communication control method executed by a computer having a processor and a communication device, the address of a first buffer storing data to be transmitted and the data to be transmitted. Address information including the length information, and storing the address information in the second buffer, copying the address information from the second buffer to the third buffer, adding a first control header, A first packet generating step for generating a packet; a copy processing step for copying the first packet from the third buffer to the fourth buffer in the communication device without using a processor; and transmission from the first buffer based on the address information. Data to be read without going through the processor, added to the tail of the first packet stored in the fourth buffer, It includes a second packet generating step of generating a packet, and a transmission step of transmitting an integrated packet stored in the fourth buffer to the destination terminal.

  In this way, for example, even when a large amount of data is transmitted, since the address information is copied to the third buffer, the CPU load due to the copy process can be reduced. Further, since the data to be transmitted is copied from the first buffer to the fourth buffer in the communication device without passing through the processor (for example, CPU), no load is applied to the CPU. Furthermore, since the communication device (for example, NIC) does not have a mechanism for guaranteeing communication reliability, it is possible to prevent the communication device from becoming complicated. Therefore, similarly to the first aspect of the present invention, it is possible to easily reduce the software overhead.

  The first or second aspect of the present invention may further include a hook processing step for hooking a data transmission request to the computer operating system. Then, the address information generation step may be performed after the hook processing step. In this way, for example, changes to the conventional program and the conventional OS that are the request source of the data transmission request become unnecessary. Therefore, software overhead can be reduced without changing the conventional OS.

  Furthermore, in the first or second aspect of the present invention, the first packet generation step may be executed by an operating system of a computer. The operating system may further include a step of notifying the communication device of a transmission request for the first packet. Further, the copy processing step, the second packet generation step, and the transmission step may be executed by the communication device. Since the processing of the first packet generation step is not different from the processing of the conventional OS, the communication reliability assurance mechanism (for example, TCP / IP protocol) that the conventional OS has can be used as it is.

  A communication control method according to a third aspect of the present invention is a communication control method executed by a computer having a processor and a communication device, which receives a packet from a transmission source terminal and stores it in a first buffer in the communication device. When the receiving step to store and the received packet is a first packet including information on a predetermined address and information on the length of specific data stored in an area indicated by the predetermined address, A first determination step for determining whether or not a second packet including information indicating the correspondence between the second packet and the specific data has already been received; and if the received packet is a second packet, the second packet corresponds to the second packet A second determination step for determining whether or not the first packet to be received has already been received, and when it is determined in the first determination step that the second packet has already been received, If it is determined in the second determination step that the first packet corresponding to the second packet has already been received, the specific data contained in the second packet is transferred from the first buffer to the second buffer without going through the processor. A copy processing step of copying and copying the first packet from the first buffer to the third buffer without going through the processor; and information on a predetermined address included in the first packet and information on the length of specific data. And a response step of copying an acknowledgment to the first packet to the transmission source terminal while copying from the third buffer to the fourth buffer.

  In this way, for example, even when large data is received, information to be copied to the fourth buffer is information on a predetermined address and information on the length of specific data. CPU load can be reduced. Further, the specific data is copied from the first buffer to the second buffer in the communication device without passing through the processor (for example, CPU), so that no load is applied to the CPU. Furthermore, since the communication device (for example, NIC) does not have a mechanism for guaranteeing communication reliability, it is possible to prevent the communication device from becoming complicated. Therefore, it is possible to easily reduce the software overhead.

  In the third aspect of the present invention, the reception step, the first determination step, the second determination step, and the copy processing step may be executed by the communication device. Further, the response step may be performed by a computer operating system. Since the response step processing is not different from the processing of the conventional OS, the communication reliability assurance mechanism (for example, TCP / IP protocol) of the conventional OS can be used as it is.

  Furthermore, in the third aspect of the present invention, the information indicating the association with the first packet may be information obtained by copying the corresponding first packet. In this case, the second packet corresponding to the first packet or the first packet corresponding to the second packet can be detected.

  A communication control method according to a fourth aspect of the present invention is a communication control method executed by a computer having a processor and a communication device, and includes information on a predetermined address and an area indicated by the predetermined address from a transmission source terminal. Receiving the integrated packet including the information on the length of the specific data stored in the data and the specific data, and storing the integrated packet in the first buffer in the communication device; and specifying the specific data included in the integrated packet in the first A copy processing step of copying data from the buffer to the second buffer without using a processor, and copying data obtained by removing specific data from the integrated packet as a first packet from the first buffer to the third buffer without using a processor; Copying information of a predetermined address and specific data length included in one packet from the third buffer to the fourth buffer In, and a response step of transmitting the acknowledgment for the first packet to the source terminal.

  In this way, for example, even when large data is received, information to be copied to the fourth buffer is information on a predetermined address and information on the length of specific data. CPU load can be reduced. Further, the specific data is copied from the first buffer to the second buffer in the communication device without passing through the processor (for example, CPU), so that no load is applied to the CPU. Furthermore, since the communication device (for example, NIC) does not have a mechanism for guaranteeing communication reliability, it is possible to prevent the communication device from becoming complicated. Therefore, similarly to the third aspect of the present invention, it is possible to easily reduce the software overhead.

  In the fourth aspect of the present invention, the reception step and the copy processing step may be executed by the communication device. Further, the response step may be performed by a computer operating system.

  Furthermore, the third or fourth aspect of the present invention may further include the step of hooking a data reception request to the computer operating system and notifying the communication device of the address of the second buffer. Furthermore, a step of discarding information on a predetermined address and information on the length of specific data stored in the fourth buffer may be included. In this way, for example, changes to the conventional program and the conventional OS that are the request source of the data reception request become unnecessary. Therefore, software overhead can be reduced without changing the conventional OS.

  Note that a program for causing a computer to execute the communication control method according to the present invention can be created. The program can be a storage medium such as a flexible disk, a CD-ROM, a magneto-optical disk, a semiconductor memory, a hard disk, or the like. It is stored in a storage device. In some cases, digital signals are distributed over a network. Note that data being processed is temporarily stored in a storage device such as a computer memory.

  According to the present invention, it is possible to easily realize a reduction in software overhead that occurs when a communication reliability guarantee is realized by software.

  In addition, according to another aspect of the present invention, it is possible to realize a reduction in software overhead that occurs when a communication reliability is ensured by software without changing a conventional OS.

[Embodiment 1]
A first embodiment of the present invention will be described with reference to FIGS. First, FIG. 3 shows a system outline according to the embodiment of the present invention. In FIG. 3, a computer 3A and a computer 3B are connected by an interface 1 such as Ethernet. FIG. 3 shows two examples of the computer 3A and the computer 3B. However, the number is not limited to two. For example, three or more computers 3 may be connected via a router or the like. . In the present embodiment, it is assumed that the computer 3A and the computer 3B perform packet communication via the interface 1.

  Each computer 3 is divided into a software level and a hardware level. In the software level, a user program 31 that requests transmission / reception of user data, a hook library 33 that hooks a predetermined instruction from the user program 31, and The OS 35 is included, and the hardware level includes an NIC 37 that is connected to the interface 1 to generate and transmit / receive packet frames. Although not shown, the user program 31, the hook library 33, and the OS 35 are stored in the storage device, and are read from the storage device to the main memory when executed by the CPU.

  FIG. 4 shows a functional block diagram of the NIC 37 in the first embodiment. The NIC 37 according to the first embodiment includes a memory buffer 371 for storing a packet frame, a transmission determination unit 372 that determines the contents of the packet frame at the time of packet transmission, and a zero copy that generates a zero copy packet frame described in detail below. A packet construction unit 373, a DMA processing unit 374 that performs data transfer by DMA (Direct Memory Access), a transmission unit 375 that transmits a packet frame stored in the memory buffer 371, and a received packet frame that is stored in the memory buffer 371 Receiving unit 376, receiving determination unit 377 for determining the contents of the packet frame when receiving the packet, and zero copy packet for notifying the DMA processing unit 374 to write the user data included in the zero copy packet frame to the data reception buffer Processing unit 3 And a 8.

  The outline of packet transmission processing in the first embodiment will be described with reference to FIGS. The user program 31, the hook library 33, the OS 35, and the NIC 37 in the transmission side computer are referred to as a transmission side user program, a transmission side hook library, a transmission side OS, and a transmission side NIC, respectively. In addition, the user program 31, the hook library 33, the OS 35, and the NIC 37 in the receiving computer are referred to as a receiving user program, a receiving hook library, a receiving OS, and a receiving NIC, respectively.

  First, the transmission-side user program requests the transmission-side OS to transmit user data (FIG. 5: Step (21)). The transmission side user program designates the address indicating the area where the user data to be transmitted is stored, the length of the user data, and the end point as in the conventional case. At this time, in this embodiment, the transmission side hook library hooks a request to the transmission side OS from the transmission side user program, and performs the following processing. That is, the transmission side hook library generates address information including the address and length specified by the transmission side user program, and stores them in the internal buffer of the transmission side hook library (step (22)). Then, the transmission side hook library requests the transmission side OS to transmit address information. That is, the transmission-side user program requests transmission of user data as in the conventional case, but the transmission-side OS transmits address information instead of user data. The transmission side OS copies the address information to the internal buffer of the transmission side OS to generate a packet frame (step (23)). Here, for convenience of explanation, a packet frame including address information is referred to as an address information packet frame. However, in actuality, there is no distinction between a conventional packet frame and an address information packet frame on the transmission side OS, and both are the same. Be treated.

  FIG. 6 shows an example of the format of the address information packet frame. In the example of FIG. 6, the address information packet frame indicates a control header (that is, transmission destination, transmission source, protocol type (IP), IP header, and TCP header) and address information (that is, an area in which user data is stored). Address and user data length). The point that address information is set instead of user data is different from the conventional packet frame (FIG. 2B).

  After generating the address information packet frame, the transmission side OS notifies the transmission side NIC of a packet transmission request. Upon receiving the transmission request from the transmission side OS, the transmission side NIC reads the packet frame from the internal buffer of the transmission side OS by DMA transfer (step (24)). Then, the transmission side NIC determines whether or not the read packet frame is an address information packet frame (step (25)). Note that the processing for checking the contents of the packet frame is the same as in the prior art, and a detailed description thereof will be omitted. If the read packet frame is determined to be a conventional packet frame, the transmission side NIC transmits the packet frame as in the conventional case. On the other hand, when it is determined that the read packet frame is an address information packet frame, the transmitting-side NIC reads the user data from the area indicated by the address included in the address information by DMA transfer, and generates a zero copy packet frame. (Step (26)). At this time, the transmitting NIC adds a copy of the control header and the corresponding address information packet frame.

  FIG. 7 shows an example of the format of the zero copy packet frame. In the example of FIG. 7, the zero copy packet frame includes a control header (that is, transmission destination, transmission source, protocol type (dedicated number) and zero copy information) and a corresponding copy of the address information packet frame (that is, transmission destination, A transmission source, a protocol type (IP), an IP header, a TCP header, an address indicating the area where user data is stored, and the length of user data) and user data. The protocol type included in the control header is set with a dedicated number indicating that it is a zero copy packet frame. Further, control information necessary for transmission / reception of a zero copy packet frame is set in the zero copy information included in the control header. For example, in Ethernet, the size of a packet frame that can be transmitted is limited. Therefore, when transmitting large user data, it is necessary to divide the packet frame into a plurality of packet frames. In conventional packet transmission, division or the like is performed by the transmission side OS. However, in this embodiment, the transmission side OS transmits address information (that is, small data) instead of user data. It is assumed that division by the transmission side OS is not necessary. Therefore, in this embodiment, it is necessary for the transmitting NIC to divide user data into a plurality of packet frames and transmit a plurality of zero copy packet frames. In this case, information indicating the position of the zero copy packet frame in the whole (that is, information corresponding to the sequence number) is set in the zero copy information. Note that processing such as division is not the main part of the present embodiment, and thus will not be described here.

  After generating the zero copy packet frame, the transmission side NIC transmits the address information packet frame and the zero copy packet frame (step (27)). Further, as will be described later, an outline of packet reception processing in this embodiment will be described later. When the address information packet frame and the zero copy packet frame arrive at the reception side computer, the transmission side NIC A packet frame (ACK) is received (step (28)). Then, the transmission side NIC writes the packet frame (ACK) in the internal buffer of the transmission side OS by DMA transfer, and notifies the transmission side OS that the packet frame (ACK) has been received (step (29)). When receiving the reception notification from the transmission side NIC, the transmission side OS confirms the ACK packet frame (step (30)). Since the transmission-side OS has confirmed that the packet frame has reached the transmission destination without any problem, the transmission-side OS performs processing such as deleting the packet frame from the internal buffer of the transmission-side OS. At this time, the transmission side OS actually deletes the address information packet frame from the internal buffer.

  In this way, since the transmission side NIC directly reads out the user data by DMA transfer, the user data is not copied at the software level. Further, the transmission side OS copies only the address and length information to the internal buffer. For example, even when a large amount of user data is transmitted, the CPU load due to the copy process can be minimized. Further, this can be realized without changing the transmission side user program and the transmission side OS.

  Next, an outline of packet reception processing in the first embodiment will be described with reference to FIG. First, the receiving-side user program requests the receiving-side OS to receive user data (FIG. 8: Step (31)). The receiving side user program designates the address of the data receiving buffer, the length of the data, and the end point as in the conventional case. At this time, in this embodiment, the receiving side hook library hooks a request to the receiving side OS from the receiving side user program, and performs the following processing. That is, the receiving side hook library notifies the receiving side NIC of the address of the data receiving buffer designated by the receiving side user program (step (32)). Thereby, as will be described later, the receiving-side NIC can write user data to the data reception buffer by DMA transfer. The receiving hook library reserves an area for storing address information (hereinafter referred to as an address information receiving buffer) in the internal buffer of the receiving hook library. As will be described later, the address information is copied to the address information reception buffer by the receiving OS. Then, the receiving side hook library requests the receiving side OS to receive the address information. That is, the receiving-side user program requests reception of user data as in the conventional case, but the receiving-side OS receives address information instead of user data. Then, the reception-side OS waits for a reception notification from the reception-side NIC.

  Further, when the receiving side NIC receives the packet frame, the receiving side NIC determines whether the received packet frame is any one of an address information packet frame, a zero copy packet frame, and another packet frame (that is, a conventional packet frame) ( Step (33)). If the received packet frame is determined to be a conventional packet frame, the conventional processing is performed. On the other hand, when it is determined that the received packet frame is an address information packet frame or a zero copy packet frame, it is determined whether a corresponding zero copy packet frame or a corresponding address information packet frame has already been received. If the corresponding zero copy packet frame or the corresponding address information packet frame has not been received yet, the subsequent processing is suspended. When the corresponding zero copy packet frame or the corresponding address information packet frame is not received for a predetermined time or longer, the received address information packet frame or zero copy packet frame is deleted from the internal buffer.

  On the other hand, if the corresponding zero copy packet frame or the corresponding address information packet frame has already been received, the receiving side NIC writes the user data included in the zero copy packet frame to the data reception buffer by DMA transfer. (Step (34)). The receiving NIC writes the address information packet frame in the internal buffer of the receiving OS by DMA transfer, and notifies the receiving OS that the packet frame has been received (step (35)). When the reception side OS receives the reception notification from the reception side NIC, the reception side OS copies the data (that is, the address information) excluding the control header from the address information packet frame stored in the internal buffer of the reception side OS to the address information reception buffer. (Step (36)).

  Also, the receiving side OS generates an ACK packet frame, stores it in the internal buffer of the receiving side OS, and notifies the receiving side NIC of the packet transmission request (step (37)). When receiving the transmission request from the receiving OS, the receiving NIC reads out the packet frame (ACK) from the internal buffer of the receiving OS by DMA transfer (step (38)). Then, the receiving side NIC transmits a packet frame (ACK) (step (39)).

  In addition, when the receiving process is completed, the receiving OS notifies the user program of a completion response. At this time, in this embodiment, the receiving side hook library hooks a completion response to the user program from the receiving side OS, and performs the following processing. That is, the receiving side hook library discards the address information stored in the address information receiving buffer (step (40)).

  In this way, since the receiving NIC directly writes the user data to the data reception buffer by DMA transfer, the user data is not copied at the software level. Further, only the address and length information is copied to the internal buffer by the receiving-side OS. For example, even when a large amount of user data is received, the CPU load due to the copy process can be minimized. Furthermore, since the address information is discarded by the hook library, it can be realized without changing the transmission side user program and the transmission side OS.

  Next, the process of the NIC 37 for performing the process as shown in FIG. 5 will be described in more detail. FIG. 9 shows a processing flow of the NIC 37 transmission processing. First, when receiving a packet transmission request from the transmission side OS (FIG. 9: step S1), the DMA processing unit 374 reads a packet frame to be transmitted from the internal buffer of the transmission side OS by DMA transfer, and stores it in the memory buffer 371. Store (step S3). The transmission determination unit 372 refers to the memory buffer 371 and determines the content of the packet frame to be transmitted (step S5). That is, it is determined whether the packet frame to be transmitted is an address information packet frame. If the packet frame to be transmitted is not an address information packet frame (step S7: No route), the transmission unit 375 transmits the packet frame as in the conventional case (step S9). Then, the process returns to the original process.

  On the other hand, when the packet frame to be transmitted is an address information packet frame (step S7: Yes route), the transmission determination unit 372 sends an instruction to the zero copy packet construction unit 373 to generate a zero copy packet frame. Output. Then, the zero copy packet construction unit 373 outputs a command for reading the user data to the DMA processing unit 374. The DMA processing unit 374 reads data by DMA transfer based on the address information included in the address information packet frame, and stores it in the memory buffer 371 (step 11). As described above, since the address information includes an address indicating the area where the user data is stored and the length of the user data, the user data is read based on these information. Then, the zero copy packet construction unit 373 adds a control header and a copy of the address information packet frame to the user data, and generates a zero copy packet frame (step S13). The transmission unit 375 transmits the address information packet frame and the zero copy packet frame stored in the memory buffer 371 (step S15). Then, the process returns to the original process.

  Next, the process of the NIC 37 for performing the process as shown in FIG. 8 will be described in more detail. FIG. 10 shows a processing flow of the NIC 37 reception processing. First, the DMA processing unit 374 receives the address of the data reception buffer from the reception side hook library and stores it in the memory buffer 371 (FIG. 10: step S17). The receiving unit 376 receives the packet frame and stores it in the memory buffer 371 (step S19). Then, the reception determining unit 377 refers to the memory buffer 371 and determines the content of the received packet frame (step S21). That is, it is determined whether the received packet frame is an address information packet frame, a zero copy packet frame, or another packet frame.

  If the received packet frame is an address information packet frame (step S23: Yes route), the reception determining unit 377 sends a command to the corresponding zero copy packet frame to the zero copy packet processing unit 378. Output. The zero copy packet processing unit 378 compares the address information packet frame with the copy of the address information packet frame included in the zero copy packet frame, and determines whether the corresponding zero copy packet frame is received (step S25). . If the corresponding zero copy packet frame has not been received yet (step S25: No route), the process returns to the process of step S19 via the terminal A. On the other hand, when the corresponding zero copy packet frame is received (step S25: Yes route), the process proceeds to step S33.

  On the other hand, when the received packet frame is not an address information packet frame (step S23: No route) but a zero copy packet frame (step S27: Yes route), the reception determination unit 377 displays the corresponding address information packet frame. An instruction to search is output to the zero copy packet processing unit 378. The zero copy packet processing unit 378 compares the address information packet frame with the copy of the address information packet frame included in the zero copy packet frame, and determines whether or not the corresponding address information packet frame is received (step S31). . If the corresponding address information packet frame has not been received yet (step S31: No route), the process returns to step S19 via the terminal A. On the other hand, when the corresponding address information packet frame is received (step S31: Yes route), the process proceeds to step S33.

  If the received packet frame is not an address information packet frame (step S23: No route) and is not a zero copy packet frame (step S27: No route), the received packet frame is a conventional packet frame. Therefore, the same processing as the conventional one is performed. That is, the DMA processing unit 374 writes the received packet frame in the internal buffer of the reception-side OS by DMA transfer (step S29).

  When the corresponding zero copy packet frame is received (step S25: Yes route) or when the corresponding address information packet frame is received (step S31: Yes route), the zero copy packet processing unit 378 A write command for user data and address information packet frame included in the zero copy packet frame is output to the DMA processing unit 374. Based on the address of the data reception buffer stored in the memory buffer 371, the DMA processing unit 374 writes the user data included in the zero copy packet frame to the data reception buffer by DMA transfer (step S33). Further, the DMA processing unit 374 writes the address information packet frame in the internal buffer of the receiving-side OS by DMA transfer (step S35). Then, the process returns to the original process.

  Before the receiving NIC receives the address of the data reception buffer from the receiving hook library, a packet frame may be transmitted from the transmitting NIC. In this case, an address information packet frame and zero are transmitted. The copy packet frame may be held in the memory buffer 371, and the processes in steps S33 and S35 may be performed at a predetermined timing.

  As described above, according to this embodiment, copy of user data does not occur at the software level of the transmission side computer and the reception side computer. For example, even when a large amount of user data is transmitted / received, the CPU by the copy process The load can be minimized, and a decrease in communication speed can be prevented. Further, the conventional user program 31 and the OS 35 can be used as they are, and the communication control function of the OS 35 (for example, retransmission when a packet is lost, etc.) can be used as it is. That is, since the reliability on the network is guaranteed at the software level, it is not necessary to guarantee the reliability on the network at the hardware level.

[Embodiment 2]
Next, a second embodiment of the present invention will be described with reference to FIGS. The system configuration in the present embodiment is basically the same as that shown in FIG. 1, but the NIC 37 b performs processing instead of the NIC 37.

  FIG. 11 is a functional block diagram of the NIC 37b according to the second embodiment. The NIC 37b in the second embodiment includes a memory buffer 371, a transmission determination unit 372, an integrated packet frame construction unit 373b, a DMA processing unit 374, a transmission unit 375, a reception unit 376, and a reception determination unit 377. And an integrated packet frame processing unit 378b. The integrated packet frame construction unit 373b generates an integrated packet frame described in detail below when transmitting a packet. Further, the integrated packet frame processing unit 378b notifies the DMA processing unit 374 to write the user data included in the integrated packet frame to the data reception buffer when receiving the packet. The memory buffer 371, transmission determination unit 372, DMA processing unit 374, transmission unit 375, reception unit 376, and reception determination unit 377 are basically the same as those in the first embodiment.

  An outline of packet transmission processing in the second embodiment will be described with reference to FIGS. First, the transmission-side user program requests the transmission-side OS to transmit user data (FIG. 12: step (51)). The transmission side user program designates the address indicating the area where the user data to be transmitted is stored, the length of the user data, and the end point as in the conventional case. At this time, as in the first embodiment, the transmission-side hook library hooks a request to the transmission-side OS from the transmission-side user program, and performs the following processing. That is, the transmission side hook library generates address information including the address and length specified by the transmission side user program, and stores them in the internal buffer of the transmission side hook library (step (52)). Then, the transmission side hook library requests the transmission side OS to transmit address information. That is, the transmission-side user program requests transmission of user data as in the conventional case, but the transmission-side OS transmits address information instead of user data. The transmission side OS copies the address information to the internal buffer of the transmission side OS to generate a packet frame (step (53)).

  After generating the address information packet frame, the transmission side OS notifies the transmission side NIC of a packet transmission request. When receiving the transmission request from the transmission side OS, the transmission side NIC reads the packet frame from the internal buffer of the transmission side OS by DMA transfer (step (54)). Then, the transmitting-side NIC determines whether or not the read packet frame is an address information packet frame (step (55)). If the read packet frame is determined to be a conventional packet frame, the transmission side NIC transmits the packet frame as in the conventional case. On the other hand, when it is determined that the read packet frame is an address information packet frame, in this embodiment, the transmission-side NIC reads user data from the area indicated by the address included in the address information by DMA transfer, and integrates the packet frame. Is generated (step (56)). The transmitting-side NIC adds zero copy information and the read user data to the end of the address information packet frame to generate an integrated packet frame.

  FIG. 13 shows an example of the format of the integrated packet frame. In the example of FIG. 13, the integrated packet frame includes an address information packet frame (FIG. 6), zero copy information, and user data.

  After generating the integrated packet frame, the transmitting side NIC transmits the integrated packet frame (step (57)). The packet reception processing outline in this embodiment will be described later. When the integrated packet frame arrives at the reception side computer, the transmission side NIC receives the packet frame (ACK) from the reception side computer. (Step (58)). Then, the transmission side NIC writes the packet frame (ACK) in the internal buffer of the transmission side OS by DMA transfer, and notifies the transmission side OS that the packet frame (ACK) has been received (step (59)). When the transmission side OS receives the reception notification from the transmission side NIC, the transmission side OS confirms the ACK packet frame (step (60)). Since the transmission-side OS has confirmed that the packet frame has reached the transmission destination without any problem, the transmission-side OS performs processing such as deleting the packet frame from the internal buffer of the transmission-side OS. At this time, the transmission side OS actually deletes the address information packet frame from the internal buffer.

  In the first embodiment, two packets of an address information packet frame and a zero copy packet frame are transmitted. In the second embodiment, necessary information is collected into one packet (integrated packet frame). Send.

  In this way, since the transmission side NIC directly reads out the user data by DMA transfer, the user data is not copied at the software level. Further, the transmission side OS copies only the address and length information to the internal buffer. For example, even when a large amount of user data is transmitted, the CPU load due to the copy process can be minimized. Further, this can be realized without changing the transmission side user program and the transmission side OS.

  Next, an outline of packet reception processing in the second embodiment will be described with reference to FIG. First, the receiving-side user program requests the receiving-side OS to receive user data (FIG. 14: step (61)). The receiving side user program designates the address of the data receiving buffer, the length of the data, and the end point as in the conventional case. At this time, as in the first embodiment, the receiving-side hook library hooks a request for the receiving-side OS from the receiving-side user program, and performs the following processing. That is, the receiving side hook library notifies the receiving side NIC of the address of the data receiving buffer designated by the receiving side user program (step (62)). Thereby, as will be described later, the receiving-side NIC can write user data to the data reception buffer by DMA transfer. The receiving hook library reserves an area for storing address information (hereinafter referred to as an address information receiving buffer) in the internal buffer of the receiving hook library. As will be described later, the address information is copied to the address information reception buffer by the receiving OS. Then, the receiving side hook library requests the receiving side OS to receive the address information. That is, the receiving-side user program requests reception of user data as in the conventional case, but the receiving-side OS receives address information instead of user data. Then, the reception-side OS waits for a reception notification from the reception-side NIC.

  Further, when receiving the packet frame, the receiving-side NIC determines whether or not the received packet frame is an integrated packet frame (step (63)). If the received packet frame is not an integrated packet frame, the conventional processing is performed. On the other hand, if the received packet frame is an integrated packet frame, the receiving-side NIC writes the user data included in the integrated packet frame to the data reception buffer by DMA transfer (step (64)). Also, the receiving NIC writes the control header and address information (that is, the contents of the address information packet frame) included in the integrated packet frame to the internal buffer of the receiving OS by DMA transfer, and receives the fact that the packet frame has been received. Notification is made to the side OS (step (65)). When the reception side OS receives the reception notification from the reception side NIC, the reception side OS copies the data (that is, the address information) excluding the control header from the address information packet frame stored in the internal buffer of the reception side OS to the address information reception buffer. (Step (66)).

  Further, the receiving side OS generates an ACK packet frame, stores it in the internal buffer of the receiving side OS, and notifies the receiving side NIC of the packet transmission request (step (67)). When receiving the transmission request from the receiving OS, the receiving NIC reads out the packet frame (ACK) from the internal buffer of the receiving OS by DMA transfer (step (68)). Then, the receiving side NIC transmits a packet frame (ACK) (step (69)).

  In addition, when the receiving process is completed, the receiving OS notifies the user program of a completion response. At this time, in this embodiment, a completion response to the user program is hooked from the receiving-side OS, and the following processing is performed. That is, the receiving side hook library discards the address information stored in the address information receiving buffer (step (70)).

  In the second embodiment, since it is transmitted as one packet (integrated packet frame), whether both packet frames (address information packet frame and zero copy packet frame) are received as in the first embodiment. It is not necessary to determine whether or not.

  In this way, since the receiving NIC directly writes the user data to the data reception buffer by DMA transfer, the user data is not copied at the software level. Further, only the address and length information is copied to the internal buffer by the receiving-side OS. For example, even when a large amount of user data is received, the CPU load due to the copy process can be minimized. Furthermore, since the address information is discarded by the hook library, it can be realized without changing the transmission side user program and the transmission side OS.

  Next, the process of the NIC 37b for performing the process as shown in FIG. 12 will be described in more detail. FIG. 15 shows a processing flow of transmission processing of the NIC 37b. First, when receiving a packet transmission request from the transmission side OS (FIG. 15: step S41), the DMA processing unit 374 reads the packet frame to be transmitted from the internal buffer of the transmission side OS by DMA transfer, and stores it in the memory buffer 371. Store (step S43). The transmission determination unit 372 refers to the memory buffer 371 and determines the content of the packet frame to be transmitted (step S45). That is, it is determined whether the packet frame to be transmitted is an address information packet frame. If the packet frame to be transmitted is not an address information packet frame (step S47: No route), the transmission unit 375 transmits the packet frame as in the conventional case (step S49). Then, the process returns to the original process.

  On the other hand, when the packet frame to be transmitted is an address information packet frame (step S47: Yes route), the transmission determination unit 372 outputs an instruction to generate an integrated packet frame to the integrated packet frame construction unit 373b. To do. Then, the integrated packet frame construction unit 373b outputs an instruction to read user data to the DMA processing unit 374. The DMA processing unit 374 reads data by DMA transfer based on the address information included in the address information packet frame, and stores it in the memory buffer 371 (step 51). As described above, since the address information includes an address indicating the area where the user data is stored and the length of the user data, the user data is read based on these information. Then, the integrated packet frame construction unit 373b adds the zero copy information and the read user data to the tail of the address information packet frame to generate an integrated packet frame (step S53). The transmission unit 375 transmits the integrated packet frame stored in the memory buffer 371 (step S55). Then, the process returns to the original process.

  Next, the process of the NIC 37b for performing the process as shown in FIG. 14 will be described in more detail. FIG. 16 shows a process flow of the reception process of the NIC 37b. First, the DMA processing unit 374 receives the address of the data reception buffer from the reception side hook library and stores it in the memory buffer 371 (FIG. 16: step S57). The receiving unit 376 receives the packet frame and stores it in the memory buffer 371 (step S59). Then, the reception determining unit 377 refers to the memory buffer 371 and determines the content of the received packet frame (step S61). That is, it is determined whether the received packet frame is an integrated packet frame.

  If the received packet frame is not an integrated packet frame (step S63: No route), the same processing as the conventional one is performed. That is, the DMA processing unit 374 writes the received packet frame in the internal buffer of the receiving-side OS by DMA transfer (step S65). On the other hand, when the received packet frame is an integrated packet frame (step S63: Yes route), the reception determination unit 377 sends an instruction to the integrated packet frame processing unit 378b to execute processing of the integrated packet frame. Output. The integrated packet frame processing unit 378b sends a write command for user data included in the integrated packet frame and control header and address information (that is, the contents of the address information packet frame) included in the integrated packet frame to the DMA processing unit 374. Output. The DMA processing unit 374 writes the user data included in the integrated packet frame to the data reception buffer by DMA transfer based on the address of the data reception buffer stored in the memory buffer 371 (step S67). Also, the DMA processing unit 374 writes the control header and address information included in the integrated packet frame in the internal buffer of the receiving OS by DMA transfer (step S69). Then, the process returns to the original process.

  As described above, according to this embodiment, copy of user data does not occur at the software level of the transmission side computer and the reception side computer. For example, even when a large amount of user data is transmitted / received, the CPU by the copy process The load can be minimized, and a decrease in communication speed can be prevented. Further, the conventional user program 31 and the OS 35 can be used as they are, and the communication control function of the OS 35 (for example, retransmission when a packet is lost, etc.) can be used as it is. That is, since the reliability on the network is guaranteed at the software level, it is not necessary to guarantee the reliability on the network at the hardware level.

  Although the embodiment of the present invention has been described above, the present invention is not limited to this. For example, the functional block diagram described above does not necessarily correspond to an actual program module configuration. The OS 35 is not necessarily an OS, and may be a dedicated library that controls communication in the same manner as the OS.

  In the example described above, the hook library 33 performs processing such as generation of address information. However, the hook library 33 may not be prepared and may be performed by the user program 31 or the OS 35. When using the user program 31, it is necessary to change a part of the user program 31, but it is not necessary to change the main structure of the user program 31 or the OS 35. In addition, in order to perform with the OS 35, it is necessary to change a part of the OS 35, but it is not necessary to change the processing related to the reliability guarantee of communication and the user program 31 that the OS 35 has.

  Further, the format of the address information packet frame, the zero copy packet frame, and the integrated packet frame described above is an example, and the content of the data part can be changed in the data arrangement, or data can be added or deleted as necessary. It may be deleted. For example, the control header may not include an IP header or a TCP header.

  Also in the processing flow, if the processing result does not change, the processing order can be changed. Further, it may be executed in parallel.

(Appendix 1)
A communication control method executed by a computer having a processor and a communication device,
An address information generating step for generating address information including information on an address of a first buffer for storing data to be transmitted and information on a length of the data to be transmitted, and storing the address information in a second buffer;
A first packet generating step of copying the address information from the second buffer to the third buffer, adding a first control header, and generating a first packet;
A copy processing step of copying the first packet from the third buffer to a fourth buffer in the communication device without using the processor;
Based on the address information, the data to be transmitted is copied from the first buffer to the fourth buffer without going through the processor, and a second control header is added to generate a second packet. Steps,
A transmission step of transmitting the first packet and the second packet stored in the fourth buffer to a destination terminal;
Including a communication control method.

(Appendix 2)
A communication control method executed by a computer having a processor and a communication device,
An address information generating step for generating address information including information on an address of a first buffer for storing data to be transmitted and information on a length of the data to be transmitted, and storing the address information in a second buffer;
A first packet generating step of copying the address information from the second buffer to the third buffer, adding a first control header, and generating a first packet;
A copy processing step of copying the first packet from the third buffer to a fourth buffer in the communication device without using the processor;
Based on the address information, the data to be transmitted is read from the first buffer without passing through the processor, and added to the end of the first packet stored in the fourth buffer to generate an integrated packet. A second packet generation step;
A transmission step of transmitting the integrated packet stored in the fourth buffer to a destination terminal;
Including a communication control method.

(Appendix 3)
A hook processing step of hooking a data transmission request to the operating system of the computer,
The communication control method according to claim 1 or 2, wherein the address information generation step is performed after the hook processing step.

(Appendix 4)
The communication control method according to claim 1 or 2, wherein the first packet generation step is executed by an operating system of the computer.

(Appendix 5)
The communication control method according to claim 4, further comprising a step of notifying the communication device of a transmission request for the first packet by the operating system.

(Appendix 6)
The communication control method according to claim 1 or 2, wherein the copy processing step, the second packet generation step, and the transmission step are executed by the communication device.

(Appendix 7)
The second packet generating step comprises:
The communication control method according to supplementary note 1, including a step of copying the first packet and adding it to the second packet.

(Appendix 8)
A communication control method executed by a computer having a processor and a communication device,
A reception step of receiving a packet from a transmission source terminal and storing the packet in a first buffer in the communication device;
When the received packet is the first packet including information on the predetermined address and information on the length of the specific data stored in the area indicated by the predetermined address, association with the first packet is performed. A first determination step of determining whether or not a second packet including information to be represented and the specific data has already been received;
If the received packet is the second packet, a second determination step of determining whether or not the first packet corresponding to the second packet has already been received;
When it is determined in the first determination step that the second packet has already been received, or in the second determination step, it is determined that the first packet corresponding to the second packet has already been received. In this case, the specific data included in the second packet is copied from the first buffer to the second buffer without using the processor, and the first packet is transferred from the first buffer to the third buffer through the processor. Copy processing steps to copy without
The information on the predetermined address and the information on the length of the specific data included in the first packet are copied from the third buffer to the fourth buffer, and a confirmation response to the first packet is sent to the transmission source terminal. A response step to send to
Including a communication control method.

(Appendix 9)
A communication control method executed by a computer having a processor and a communication device,
An integrated packet including information on a predetermined address, information on the length of specific data stored in an area indicated by the predetermined address, and the specific data is received from a transmission source terminal, A receiving step for storing in one buffer;
The specific data included in the integrated packet is copied from the first buffer to the second buffer without going through the processor, and data obtained by removing the specific data from the integrated packet is used as the first packet as the first buffer. A copy processing step of copying from the first to the third buffer without going through the processor;
The information on the predetermined address and the information on the length of the specific data included in the first packet are copied from the third buffer to the fourth buffer, and a confirmation response to the first packet is sent to the transmission source terminal. A response step to send to
Including a communication control method.

(Appendix 10)
The communication control method according to claim 8 or 9, further comprising the step of hooking a data reception request to the operating system of the computer and notifying the communication device of the address of the second buffer.

(Appendix 11)
The communication control method according to appendix 8 or 9, further comprising: a step of discarding the information on the predetermined address and the information on the length of the specific data stored in the fourth buffer.

(Appendix 12)
The communication control method according to claim 8, wherein the reception step, the first determination step, the second determination step, and the copy processing step are executed by the communication device.

(Appendix 13)
The communication control method according to claim 9, wherein the reception step and the copy processing step are executed by the communication device.

(Appendix 14)
The communication control method according to appendix 8 or 9, wherein the response step is executed by an operating system of the computer.

(Appendix 15)
The communication control method according to claim 8, wherein the information indicating the association with the first packet is information obtained by copying the corresponding first packet.

(Appendix 16)
A communication device connected to a computer having a processor,
A storage device for storing packets to be transmitted;
Means for copying the first packet from the first buffer storing the first packet related to the packet transmission request to the storage device without going through the processor when a packet transmission request is received;
When the first packet stored in the storage device includes information on the address of a second buffer for storing specific data and information on the length of the specific data, the specific data is stored in the second buffer. Means for copying from the storage device without passing through the processor;
Packet generating means for generating a second packet by adding a control header to the specific data stored in the storage device;
Means for transmitting the first packet and the second packet stored in the storage device to a destination terminal;
A communication device.

(Appendix 17)
A communication device connected to a computer having a processor,
A storage device for storing packets to be transmitted;
Means for copying the first packet from the first buffer storing the first packet related to the packet transmission request to the storage device without going through the processor when a packet transmission request is received;
When the first packet stored in the storage device includes information on the address of a second buffer for storing specific data and information on the length of the specific data, the specific data is stored in the second buffer. Means for reading out without passing through the processor and adding to the tail of the first packet stored in the storage device to generate an integrated packet;
Means for transmitting the integrated packet stored in the storage device to a destination terminal;
A communication device.

(Appendix 18)
The packet generation means
The communication device according to appendix 16, further comprising means for copying the first packet stored in the storage device and adding the first packet to the second packet.

(Appendix 19)
A communication device connected to a computer having a processor,
A storage device for storing received packets;
Means for receiving a packet from a source terminal and storing it in the storage device;
When the received packet is a first packet including information on a predetermined address and information on the length of specific data stored in an area indicated by the predetermined address, it represents an association with the first packet. First determination means for determining whether a second packet including information and the specific data has already been received;
If the received packet is the second packet, second determination means for determining whether or not the first packet corresponding to the second packet has already been received;
When it is determined by the first determination means that the second packet has already been received, or when the second determination means has determined that the first packet corresponding to the second packet has already been received. The specific data contained in the second packet is copied from the storage device to the first buffer without passing through the processor, and the first packet is copied from the storage device to the second buffer without passing through the processor. Means for copying;
A communication device.

(Appendix 20)
A communication device connected to a computer having a processor,
A storage device for storing received packets;
Means for receiving, from the transmission source terminal, information on a predetermined address, information on the length of specific data stored in the area indicated by the predetermined address, and an integrated packet including the specific data, and storing them in the storage device When,
The specific data included in the integrated packet is copied from the storage device to the first buffer without passing through the processor, and the data obtained by removing the specific data from the integrated packet is stored as the first packet from the storage device. Means for copying to two buffers without going through the processor;
A communication device.

(Appendix 21)
The communication device according to appendix 19 or 20, further comprising: means for receiving information on the address of the first buffer and storing the information in the storage device.

It is a figure for demonstrating the process outline of the conventional packet transmission / reception. (A) And (b) is a figure which shows an example of the format of the packet frame transmitted / received on Ethernet. It is a system outline figure concerning an embodiment of the invention. It is a functional block diagram of NIC in the 1st Embodiment of this invention. It is a figure which shows the process outline | summary of the packet transmission in the 1st Embodiment of this invention. It is a figure which shows an example of a format of an address information packet frame. It is a figure which shows an example of a format of a zero copy packet frame. It is a figure which shows the process outline | summary of the packet reception in the 1st Embodiment of this invention. It is a figure which shows the processing flow of the transmission process of NIC in the 1st Embodiment of this invention. It is a figure which shows the processing flow of the reception process of NIC in the 1st Embodiment of this invention. It is a functional block diagram of NIC in the 2nd Embodiment of this invention. It is a figure which shows the process outline | summary of the packet transmission in the 2nd Embodiment of this invention. It is a figure which shows an example of a format of an integrated packet frame. It is a figure which shows the process outline | summary of the packet reception in the 2nd Embodiment of this invention. It is a figure which shows the processing flow of the transmission process of NIC in the 2nd Embodiment of this invention. It is a figure which shows the processing flow of the reception process of NIC in the 2nd Embodiment of this invention.

Explanation of symbols

1 Interface 3 Computer 31 User Program 33 Hook Library 35 OS 37 NIC
371 Memory buffer 372 Transmission determination unit 373 Zero copy packet construction unit 373b Integrated packet frame construction unit 374 DMA processing unit 375 Transmission unit 376 Reception unit 377 Reception determination unit 378 Zero copy packet processing unit 378b Integrated packet frame processing unit

Claims (8)

A communication control method executed by a computer having a processor and a communication device,
An address information generating step for generating address information including information on an address of a first buffer for storing data to be transmitted and information on a length of the data to be transmitted, and storing the address information in a second buffer;
A first packet generating step of copying the address information from the second buffer to the third buffer, adding a first control header, and generating a first packet;
A copy processing step of copying the first packet from the third buffer to a fourth buffer in the communication device without using the processor;
Based on the address information, the data to be transmitted is copied from the first buffer to the fourth buffer without going through the processor, and a second control header is added to generate a second packet. Steps,
A transmission step of transmitting the first packet and the second packet stored in the fourth buffer to a destination terminal;
Including a communication control method. A communication control method executed by a computer having a processor and a communication device,
An address information generating step for generating address information including information on an address of a first buffer for storing data to be transmitted and information on a length of the data to be transmitted, and storing the address information in a second buffer;
A first packet generating step of copying the address information from the second buffer to the third buffer, adding a first control header, and generating a first packet;
A copy processing step of copying the first packet from the third buffer to a fourth buffer in the communication device without using the processor;
Based on the address information, the data to be transmitted is read from the first buffer without passing through the processor, and added to the end of the first packet stored in the fourth buffer to generate an integrated packet. A second packet generation step;
A transmission step of transmitting the integrated packet stored in the fourth buffer to a destination terminal;
Including a communication control method. A hook processing step of hooking a data transmission request to the operating system of the computer,
The communication control method according to claim 1, wherein the address information generation step is performed after the hook processing step. The communication control method according to claim 1, wherein the first packet generation step is executed by an operating system of the computer. The communication control method according to claim 1, wherein the copy processing step, the second packet generation step, and the transmission step are executed by the communication device. A communication device connected to a computer having a processor,
A storage device for storing received packets;
Means for receiving a packet from a source terminal and storing it in the storage device;
When the received packet is a first packet including information on a predetermined address and information on the length of specific data stored in an area indicated by the predetermined address, it represents an association with the first packet. First determination means for determining whether a second packet including information and the specific data has already been received;
If the received packet is the second packet, second determination means for determining whether or not the first packet corresponding to the second packet has already been received;
When it is determined by the first determination means that the second packet has already been received, or when the second determination means has determined that the first packet corresponding to the second packet has already been received. The specific data contained in the second packet is copied from the storage device to the first buffer without passing through the processor, and the first packet is copied from the storage device to the second buffer without passing through the processor. Means for copying;
A communication device. A communication device connected to a computer having a processor,
A storage device for storing received packets;
Means for receiving, from the transmission source terminal, information on a predetermined address, information on the length of specific data stored in the area indicated by the predetermined address, and an integrated packet including the specific data, and storing them in the storage device When,
The specific data included in the integrated packet is copied from the storage device to the first buffer without passing through the processor, and the data obtained by removing the specific data from the integrated packet is stored as the first packet from the storage device. Means for copying to two buffers without going through the processor;
A communication device. The communication device according to claim 6 or 7, further comprising: means for receiving information on an address of the first buffer and storing the information in the storage device.

Images