JP3612512B2 - Network relay device, network relay method, and program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a network relay technology applied to bridges, routers, multilayer switches, and the like, and more particularly to a network relay technology capable of eliminating packet loss with an economical configuration.
[0002]
[Prior art]
In network relay devices such as bridges, routers, and multilayer switches, generally, packets received via input ports are temporarily stored in an input buffer, and the stored packets are sequentially processed in the order of storage. . Therefore, if the processing time for the packet becomes longer than the packet arrival interval, the received packet gradually stays in the input buffer, and eventually overflow (packet loss) occurs. Even in a network relay device of a type that does not include an input buffer for each input port, a packet loss occurs when the processing time is longer than the packet arrival interval.
[0003]
When packet loss occurs, packet retransmission processing is performed in an upper layer of the network. When packet retransmission processing is performed, there may be a vicious circle of traffic increase → load increase of network relay device → packet loss → retransmission processing. Also, the client process often stops until the upper layer retransmission process is completed.
[0004]
In order to prevent the occurrence of such a situation, conventionally, the operation clock of the network relay device is made high so that the processing time for the packet becomes shorter than the arrival interval of the packet.
[0005]
[Problems to be solved by the invention]
However, the method of preventing packet loss by increasing the operation clock of the network relay device has to use expensive hardware capable of high-speed operation, which increases the price and power consumption. There is also the problem of increasing.
[0006]
In recent years, the amount of data flowing through network relay devices such as bridges, routers, and multilayer switches has increased explosively with the expansion of networks and the improvement of line speeds such as LAN and WAN. Furthermore, in a network relay device, in addition to basic processing such as bridging processing and routing processing, encapsulation processing, address conversion processing, filtering processing, and the like have been performed. It tends to become more complex. Therefore, it is fully predicted that it will not be possible to cope with packet loss only by increasing the operation clock of the network relay device in the future.
[0007]
Therefore, an object of the present invention is to enable packet loss to be prevented with an economical configuration.
[0008]
[Means for Solving the Problems]
The network relay device of the present invention achieves the above-described object,
A plurality of input ports and a plurality of output ports are provided, relay processing including conversion processing is executed for packets received via the input ports, and the packets for which relay processing has been completed are output to the output port corresponding to the destination. An output network relay device,
The relay process is divided into a plurality of partial processes, and when the plurality of partial processes are sequentially performed on the received packet, the processing result is obtained from the plurality of input ports after the completion of one partial process. Means is provided for looping back to an input port assigned for the next partial processing and causing the next partial processing to perform processing using the looped-back processing result.
[0009]
According to this configuration, partial processing is sequentially performed on packets received via the input port. Therefore, even if the processing time of the entire relay processing is longer than the packet arrival interval, The processing time can be made shorter than the packet arrival time, and as a result, the occurrence of packet loss can be prevented. For example, if the processing time of the entire relay processing is 2T and the packet arrival interval is 1T, the relay processing is divided into two partial processes, and the processing time of each of the divided processing is 1T, thereby preventing packet loss. be able to. In addition, since the operation clock of the network relay device can be kept low compared to when the relay processing is not divided into a plurality of partial processes, the network relay device can be configured using low-speed and inexpensive parts. In addition, power consumption can be kept low.
[0010]
More specifically, the network relay device of the present invention is:
A plurality of input ports, an input buffer for each of the plurality of input ports, and a plurality of output ports, and an output port corresponding to a transmission destination of a packet stored in the input buffer via the input port A network relay device that executes relay processing including conversion processing on the accumulated packets and outputs the packets that have undergone relay processing to the specific output port when the specific output port is a port,
When the output port corresponding to the transmission destination of the packet stored in the input buffer is the specific output port, a part of the relay processing is executed on the stored packet, and the part Looping back a packet for which the processing of step 1 is completed by using a loopback output port and a loopback input port among the plurality of input ports and output ports, for loopback corresponding to the loopback input port Means for storing in the input buffer, executing the remaining processing of the relay processing on the packet stored in the loopback input buffer, and outputting the packet for which the remaining processing has been completed to the specific output port It has.
[0011]
DETAILED DESCRIPTION OF THE INVENTION
Next, embodiments of the present invention will be described in detail with reference to the drawings.
[0012]
Referring to FIG. 1, there is shown a configuration example of a multilayer switch as an embodiment of the present invention. The multi-layer switch of the present embodiment includes a LAN line processing unit 1, a loopback circuit 2, a packet conversion processing unit 3, a packet buffer management unit 4, and a buffer memory 5.
[0013]
The LAN line processing unit 1 includes a plurality of input / output ports 11-1 to 11-n that are pairs of input ports and output ports, and input buffers 12-1 to 12-1 for each of the input / output ports 11-1 to 11-n. 12-n. Among the input / output ports 11-1 to 11-n, the jth input / output port 11-j (loopback input / output port 11-j) is connected to the loopback circuit 2, and the nth The input / output port 11-n (specific input / output port 11-n) is connected to the Internet 6. Input / output ports other than the loopback input / output port 11-j and the specific input / output port 11-n are connected to different LANs (not shown). The LAN line processing unit 1 is configured using, for example, an LSI called a MAC chip.
[0014]
The loopback circuit 2 connects an output port (referred to as a loopback output port) and an input port (loopback input port) constituting the loopback input / output port 11-j, and from the loopback output port. Loop the output packet to the loopback input port.
[0015]
The packet conversion processing unit 3 is configured by, for example, a network processing processor, and includes a first processing unit 31, a second processing unit 33, a forward table 33, and a recording medium K1.
[0016]
The packet conversion processing unit 3 executes relay processing including encapsulation processing, and the first and second processing means 31 and 32 divide and perform the relay processing.
[0017]
The first processing means 31 performs the following processes A to C on the input buffers 12-1 to 12-n other than the loopback input buffer 12-j as processing targets.
[0018]
A. When the packet is accumulated in the input buffer to be processed, the output port of the packet output destination is determined based on the packet destination address and the contents of the forward table 33.
B. If the output port determined in A is a specific output port for Internet connection, the packet is passed to the packet buffer management unit 4 and the port number of the loopback output port is passed as information indicating the output destination of the packet. .
C. If the output port determined in A is not a specific output port for Internet connection, the packet is transferred to the packet buffer management unit 4 and the port number of the output port determined in A is used as information indicating the output destination of the packet. hand over.
[0019]
The second processing unit 32 sets the loopback input buffer 12-j as a processing target and performs the following processes D and E.
[0020]
D. The packet stored in the loopback input buffer 12-j is taken out and encapsulated.
E. The encapsulated packet is passed to the packet buffer management unit 4, and the port number of the specific output port for Internet connection is passed as information indicating the output port of the packet.
[0021]
The recording medium K1 provided in the packet conversion processing unit 3 is a disk, a semiconductor memory, or other recording medium, and a program for causing the network processing processor to function as the packet conversion processing unit 3 is recorded. This program is read by the network processing processor, and controls the operation thereof, thereby realizing the first and second processing means 31 and 32 on the network processing processor.
[0022]
The buffer memory 5 is divided into a plurality of partial areas, and each partial area stores a packet and the port number of the output port of the output destination. The packet buffer management unit 4 stores the packet and port number pair passed from the packet conversion processing unit 3 in the partial area next to the partial area in which the previous pair is stored, and the storage order from the oldest The packet / port number pair is read out, and the packet is output to the output port of the port number via the packet conversion processing unit 3 → the LAN line processing unit 1. That is, the packet buffer management unit 4 and the buffer memory 5 operate as a queue.
[0023]
[Description of operation of embodiment]
Next, the operation of this embodiment will be described in detail.
[0024]
Now, for example, as shown in FIG. 2, it is assumed that the received packet 7 is input to the input / output port 11-2 and stored in the input buffer 12-2. In FIG. 2, the same reference numerals as those in FIG. 1 denote the same parts. As shown in FIG. 3, the received packet 7 includes a transmission destination address (DA), a transmission source address (SA), a type (TYPE), an IP header (IP Header), and information (Information). In addition, the IP header includes an existing time (TTL) and a checksum (CS).
[0025]
The first processing means 31 in the packet conversion processing unit 3 sequentially targets input buffers other than the loopback input buffer 12-j. For example, if the input buffer 12-2 is set as a processing target, the received packet 7 is stored in the input buffer 12-2. Therefore, the processing shown in the flowchart of FIG. 4 is performed.
[0026]
First, the first processing means 31 extracts the transmission destination address (DA) from the received packet 7 (step A1). Next, based on the transmission destination address (DA) and the contents of the forward table 33, the output port of the output destination of the received packet 7 is determined (step A2).
[0027]
Thereafter, the existence time (TTL) is decremented by 1, and the checksum (CS) is recalculated (steps A3 and A4). Next, it is determined whether or not the output port obtained in step A2 is a specific output port (nth output port) for Internet connection (step A5).
[0028]
If it is determined that the port is a specific output port, that is, if the received packet 7 is to be sent to the Internet 6 (YES in step A5), the received packet 7 and loopback output are sent to the packet buffer management unit 4. A pair with the port number (referred to as #j) of the port is passed (step A6). Accordingly, the packet buffer management unit 4 stores the pair of the received packet 7 and the port number #j in the storage target area of the buffer memory 5. Thereafter, the packet buffer management unit 4 reads the pair of the received packet 7 and the port number #j from the buffer memory 5, and sends the received packet 7 to the port via the packet conversion processing unit 3 → the LAN line processing unit 1. Output to the loopback output port number #j. The received packet 7 output from the loopback output port is stored in the loopback input buffer 12-j via the loopback circuit 2 → the loopback input port.
[0029]
On the other hand, when it is determined that the port is not a specific output port, that is, when the received packet 7 is to be sent to another LAN (step A5 is NO), the packet buffer management unit 4 receives the received packet 7 and the step. A pair with the port number (referred to as #x) of the output port determined in A2 is passed (step A7). Thus, the packet buffer management unit 4 stores the pair of the received packet 7 and the port number #x in the storage target area of the buffer memory 5. Thereafter, the packet buffer management unit 4 reads the pair of the received packet 7 and the port number #x from the buffer memory 5, and sends the received packet 7 to the port via the packet conversion processing unit 3 → the LAN line processing unit 1. Output to the output port number #x.
[0030]
On the other hand, the second processing means 32 targets only the loopback input buffer 12-j, and when the received packet 7 is stored in the loopback input buffer 12-j, it is shown in the flowchart of FIG. Process.
[0031]
First, an additional IP header is created (step B1). Next, as shown in FIG. 3, a transmission packet 8 is created by adding an additional IP header to the reception packet 7 (step B2). Thereafter, a pair of the transmission packet 8 and the port number (referred to as #n) of the specific output port is passed to the packet buffer management unit 4 (step B3). Thus, the packet buffer management unit 4 stores the pair of the transmission packet 8 and the port number #n in the storage target area of the buffer memory 5. Thereafter, the packet buffer management unit 4 reads the pair of the transmission packet 8 and the port number #n from the buffer memory 5, and transmits the transmission packet 8 to the port via the packet conversion processing unit 3 → the LAN line processing unit 1. Output to the specific output port number #n.
[0032]
In the above-described embodiment, the relay process performed by the network relay device is divided into two partial processes, but may be divided into three or more partial processes. In that case, it is necessary to provide processing means and a loopback circuit according to the number of divisions. For example, when the relay process is divided into the first to m-th m partial processes, m processing means for executing the first to m-th partial processes, the first to (m-1) th (m-1) th processes. It is necessary to provide (m−1) loopback circuits for looping back the processing results of the processing means, and (m−1) input buffers corresponding to the first to (m−1) th loopback circuits. is there.
[0033]
Other Embodiments of the Invention
FIG. 6 is a block diagram of another embodiment of the present invention. The difference from the embodiment shown in FIG. 1 is that a packet conversion processing unit 3a is provided instead of the packet conversion processing unit 3. The loopback circuit 2k is connected to the input / output port 11-k (loopback input / output port 11-k).
[0034]
The packet conversion processing unit 3a executes relay processing including encapsulation processing and filter processing, and is realized by, for example, a network processing processor.
[0035]
The difference between the packet conversion processing unit 3a and the packet conversion processing unit 3 shown in FIG. 1 is that the second processing unit 32a is provided instead of the second processing unit 32, and the recording is performed instead of the recording medium K1. A point that the medium K1a is provided and a third processing unit 34 are added.
[0036]
The second processing means 32a has substantially the same function as the second processing means 32 shown in FIG. 1, but instead of outputting the encapsulated packet to a specific output port for Internet connection, The difference is that the data is output to a loopback output port to which the loopback circuit 2k is connected.
[0037]
The third processing means 34 performs a filtering process on the packets accumulated in the loopback input buffer 12-k as a processing target.
[0038]
The recording medium k1a is a disk, semiconductor memory, or other recording medium, and stores a program for causing the network processing processor to function as the packet conversion processing unit 3a. This program is read by the network processing processor, and controls the operation thereof, thereby realizing the first, second, and third processing means 31, 32a, and 34 on the network processing processor.
[0039]
Next, the operation of this embodiment will be described.
[0040]
When the packet P input from the input / output port 11-1 is accumulated in the input buffer 12-1, the first processing means 31 performs the processing shown in the flowchart of FIG. Now, for example, if the output port corresponding to the transmission destination of the packet P is a specific output port for Internet connection (YES in step A5), the packet P is input to the loopback via the loopback circuit 2. Stored in the buffer 12-j.
[0041]
Thereafter, the second processing means 32a performs the processing shown in steps B1 and B2 in the flowchart of FIG. 5 on the packet P stored in the loopback input buffer 12-j as a processing target. Thereafter, the second processing means 32a performs processing for outputting the packet P that has undergone the encapsulation processing to the loopback output port to which the loopback circuit 2k is connected, instead of performing the processing of step B3. Do. As a result, the packet P is stored in the loopback input buffer 12-k.
[0042]
Thereafter, the third processing means 34 performs the processing shown in the flowchart of FIG. 7 with the packet P stored in the loopback input buffer 12-k as a processing target.
[0043]
In the first step C1, the source address (SA) is extracted from the packet P. In the next step C2, it is determined whether or not the transmission source of the packet P is permitted to use the Internet 6 based on the transmission source address (SA) and the address designated by the administrator.
[0044]
If it is determined that the use of the Internet 6 is prohibited, the packet P is discarded (step C2 is YES, step C3). On the other hand, when it is determined that the use of the Internet is permitted, a pair of the packet P and the port number #n of the specific output port for Internet connection is passed to the packet buffer management unit 4 (Step C2 NO, step C4). Thereby, the packet buffer management unit 4 stores the pair of the packet P and the port number #n in the storage target area of the buffer memory 5. Thereafter, the packet buffer management unit 4 reads the pair of the packet P and the port number #n from the buffer memory 5, and sends the packet P to the port number # via the packet conversion processing unit 3 a → the LAN line processing unit 1. Output to n specific output ports.
[0045]
【The invention's effect】
As described above, according to the present invention, partial processing is sequentially performed on a packet received via an input port. Therefore, even if the processing time of the entire relay processing is longer than the packet arrival interval, The processing time of each partial process can be made shorter than the packet arrival interval, and as a result, occurrence of packet loss can be prevented. In addition, since the operation clock of the network relay device can be kept low compared with the case where the relay processing is not divided into a plurality of partial processes, the network relay device can be configured using low-speed and inexpensive parts and consumed. Electric power can be kept low.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of the present invention.
FIG. 2 is a diagram for explaining the operation of an embodiment of the present invention.
3 is a diagram illustrating a configuration example of a reception packet 7 and a transmission packet 8. FIG.
FIG. 4 is a flowchart showing a processing example of the first processing means 31;
FIG. 5 is a flowchart showing a processing example of the second processing means 32;
FIG. 6 is a block diagram of another embodiment of the present invention.
FIG. 7 is a flowchart showing a processing example of the third processing means 34;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... LAN line process part 11-1 to 11-n ... Input-output port 12-1 to 12-n ... Input buffer 2, 2k ... Loopback circuit 3, 3a ... Packet conversion process part 31 ... 1st process means 32 ... second processing means 33 ... forward table 34 ... third processing means K1, K1a ... recording medium 4 ... packet buffer management unit 5 ... buffer memory 6 ... internet 7 ... received packet 8 ... transmitted packet DA ... destination address SA ... Source address TYPE ... Type IP Header ... IP header Information ... Information TTL ... Existing time CS ... Checksum

Claims (13)

A plurality of input ports and a plurality of output ports are provided, relay processing including conversion processing is executed for packets received via the input ports, and the packets for which relay processing has been completed are output to the output port corresponding to the destination. An output network relay device,
The relay process is divided into a plurality of partial processes, and when the plurality of partial processes are sequentially performed on the received packet, the processing result is obtained from the plurality of input ports after the completion of one partial process. A network relay device comprising means for looping back to an input port assigned for the next partial process and causing the next partial process to perform processing using the looped back processing result. A plurality of input ports, an input buffer for each of the plurality of input ports, and a plurality of output ports, and an output port corresponding to a transmission destination of packets accumulated in the input buffer via the input ports A network relay device that executes relay processing including conversion processing on the accumulated packets and outputs a packet for which relay processing has been completed to the specific output port when the port is a specific output port;
When the output port corresponding to the transmission destination of the packet accumulated in the input buffer is the specific output port, a part of the relay processing is executed on the accumulated packet, and the part Looping back the packet that has been processed in the loopback using the loopback output port and the loopback input port among the plurality of input ports and output ports. Means for accumulating in the input buffer, executing the remaining processing of the relay processing on the packet accumulated in the loopback input buffer, and outputting the completed packet to the specific output port A network relay device comprising: In the network relay device according to claim 2,
A loopback circuit that connects the loopback input port and a loopback output port of the plurality of output ports;
When the output port corresponding to the transmission destination of the packet extracted from the input buffer is the specific output port, a part of the relay processing is performed on the extracted packet, and the partial processing is By outputting the completed packet to the loopback output port, it is accumulated in the loopback input buffer, and the remaining processing of the relay processing is executed on the packet accumulated in the loopback input buffer. And a packet conversion processing unit for outputting the packet for which the remaining processing is completed to the specific output port. In the network relay device according to claim 3,
The specific output port is an output port connected to the Internet, the loopback output port is an output port connected to the loopback circuit, and the remaining output ports are output ports connected to different LANs. A network relay device characterized by being. The network relay device according to claim 4, wherein
A network relay device, wherein the conversion process is an encapsulation process. A plurality of input ports and a plurality of output ports are provided, relay processing including conversion processing is executed for packets received via the input ports, and the packets for which relay processing has been completed are output to the output port corresponding to the destination. A network relay method in an output network relay device,
Dividing the relay process into a plurality of partial processes;
When the plurality of partial processes are sequentially performed on the received packet, the result of processing after the completion of one partial process is assigned to an input port assigned to the next partial process among the plurality of input ports. A network relay method, wherein the next partial process is caused to perform a process using the looped back process result. A network relay method in a network relay device comprising a plurality of input ports, an input buffer for each of the plurality of input ports, and a plurality of output ports,
When the output port corresponding to the transmission destination of the packet stored in the input buffer via the input port is a specific output port of the plurality of output ports, the stored packet is converted. Execute some of the relay processing,
A loop corresponding to the loopback input port by looping back the packet that has been partially processed using the loopback output port and loopback input port of the plurality of input ports and output ports. Accumulate in the input buffer for back,
The remaining processing of the relay processing is executed on the packets stored in the loopback input buffer,
A network relay method, comprising: outputting the packet for which the remaining processing is completed to the specific output port. The network relay method according to claim 7, wherein
The network relay method, wherein the specific output port is an output port connected to the Internet, and output ports other than the specific output port and the loopback output port are output ports connected to different LANs, respectively. . The network relay method according to claim 8, wherein
A network relay method, wherein the conversion process is an encapsulation process. A program for causing a computer having a plurality of input ports and a plurality of output ports to function as a network relay device,
The computer,
The relay process is divided into a plurality of partial processes, and when the plurality of partial processes are sequentially performed on the received packet, the processing result is obtained from the plurality of input ports after the completion of one partial process. A program for causing a loopback to an input port assigned for the next partial process and for causing the next partial process to perform a process using the looped-back processing result. A program for causing a computer having a plurality of input ports, an input buffer for each of the plurality of input ports, and a plurality of output ports to function as a network relay device,
The computer,
When the output port corresponding to the transmission destination of the packet stored in the input buffer via the input port is a specific output port of the plurality of output ports, the stored packet is converted. A part of the relay process is executed, and the packet after the part of the process is completed is looped back using the loopback output port and the loopback input port among the plurality of input ports and output ports. Is stored in the loopback input buffer corresponding to the loopback input port, and the remaining processing of the relay processing is executed on the packets stored in the loopback input buffer. A program for functioning as a means for outputting a packet for which the above processing has been completed to the specific output port. The program according to claim 11, wherein
The specific output port is an output port connected to the Internet, and the output port other than the specific output port and the loopback output port is an output port connected to a different LAN. In the program according to claim 12,
A program characterized in that the conversion process is an encapsulation process.

Images