JP2013138290A - Relay device and failure processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To increase reliability of a packet buffer which stores packet data in a network relay device.SOLUTION: The relay device has: a data error detection part 201 which manages an area of a packet buffer 10 by divided entry units E(0)-E(N-1) and can detect a failure for every entry; and an error frequency table 302 in which a number of times of failure for every entry is counted. When the number of times of failure for every entry exceeds a threshold value, a failure entry is excluded to continue an operation.

Description

    The present invention relates to a relay device and a failure processing method, and more particularly to a network relay device and a failure processing method for a packet buffer mounted on the network relay device.

  In recent years, the dependence on networks in businesses and homes has increased, and high reliability is required for network relay devices that are components of the networks. In general, a network relay device is provided with a buffer (hereinafter referred to as a packet buffer) for storing relay packets. The packet input to the network relay device is temporarily stored in the packet buffer, and when the device outputs the packet, the packet is extracted from the packet buffer and sent to the subsequent device. That is, the packet buffer is a memory that is most frequently used in such a network relay device, and the failure rate of the packet buffer greatly contributes to the reliability of the network relay device.

  The packet buffer is generally managed as a shared buffer. Packets input from multiple lines share one buffer system virtually divided. When a failure occurs in a packet buffer managed as a shared buffer, it may affect the communication of all lines, so it is easily determined that the failure of the network relay device due to the failure of the packet buffer Instead, it continues the operation while discarding the packet that was being processed when the failure was detected, and takes a technique that does not cause the network relay device to stop as much as possible. For example, the number of packet buffer failures per unit time is monitored, and if the failure repeats, the network relay device is stopped. However, with the recent increase in the speed and capacity of networks, the speed and integration of the packet buffer itself are remarkable, and the failure rate of the packet buffer tends to increase in proportion. For this reason, the failure processing method of the packet buffer is an important technique in aiming for high reliability of the apparatus.

  High-speed and highly integrated packet buffer failures are classified into two cases: soft errors and hard errors. The soft error is a failure in which data in the memory is temporarily destroyed due to a neutron beam colliding with a memory cell, and can be recovered from the failure by rewriting correct data. A hard error is a failure in which data in a memory is permanently destroyed due to physical memory cell damage or the like.

  In a network relay device, discarding a packet is often within an expected range. Data arrival guarantee as a communication system is realized by retransmitting a discarded packet in a higher layer than the network relay device. Therefore, a temporary packet buffer failure due to a soft error or the like may be continued by discarding the failed packet.

  A packet buffer managed as a shared buffer is virtually shared by a plurality of lines. In general, the total amount of buffers virtually allocated to each line is larger than the entire capacity of the packet buffer. This is because it is statistically rare that all lines use a large amount of packet buffers simultaneously. This is called the statistical multiplexing effect of the packet buffer. Therefore, in the shared buffer system, even if the packet buffer area is damaged to some extent, it is assumed that there is no change in the amount of buffer that can be used by each line, and that communication is not affected.

  As a buffer failure processing technique other than the network relay device, there is a method as described in Patent Document 1. In Patent Document 1, “Even if a 1-bit error occurs in the memory, the data in a certain unit of the memory including the portion where the 1-bit error has occurred is saved in another storage area, and the memory in which the error has occurred is fixed. By disabling the unit area, the fault tolerance of the system is improved and the reliability is improved "(summary).

Japanese Patent Laid-Open No. 11-03290

  Conventionally, as a method for handling a failure of a packet buffer mounted on a network relay device, a failure is detected in units of physical buffers using a data protection code such as ECC, and the number of failures occurring per unit time is used. Network relay device failure. In such a case, in the case where only a specific area of the buffer has failed, the network relay apparatus has failed even if the other area can operate normally. Also, in the case where a specific area of the packet buffer is read continuously over multiple times, such as a multicast packet, even if the area causes a temporary failure due to a soft error, The network relay device failure occurred because multiple failures were detected.

  As a buffer failure processing technique other than the network relay device, there is a method as described in Patent Document 1. However, in the memory control system described in Patent Document 1, it is necessary to interrupt read / write access to the packet buffer while the buffer failure area is copied and saved to another area. When this method is applied to the packet buffer of the network relay device, the packet relay performance is deteriorated, which is not necessarily appropriate.

  In view of the following points, the present invention reduces the occurrence probability of a failure of the network relay device by managing the failure that occurs in the packet buffer in divided packet units in the packet buffer mounted on the network relay device, The purpose is to increase reliability.

  According to one aspect of the failure handling method in the packet buffer provided by the present invention, the following configuration is provided. That is, a packet buffer for temporarily storing packets received by the network relay device, a packet buffer entry management means for managing the packet buffer area in divided entries, and taking out from the packet storage in entries. A data error detecting means capable of detecting a data error occurring in a period up to and including an error occurrence frequency table having a counter for storing the number of error occurrences detected by the data error detection means in units of entries, and the error occurrence frequency table A means for detecting a failure for each entry by comparing the number of error occurrences for each entry stored in and a preset threshold, and a counter for counting the number of failure entries detected by the failure detection means for each entry; Threshold value in which the failure entry number counter is set in advance A means for reporting the packet buffer failure detected to the host control section by comparison, and means for outputting a counter for counting said fault entry and the error count table as needed.

According to the first solution of the present invention,
A relay device for relaying packets,
A packet buffer that includes a plurality of entries divided into regions, and temporarily stores received packets in each of the entries;
A free entry management unit for managing the packet buffer in units of entries;
A packet buffer control unit that extracts a packet from the packet buffer and detects a data error in units of entries;
A plurality of counters corresponding to each of the entries of the packet buffer, and an error occurrence number table for holding the number of occurrences of data errors for each entry;
An entry free entry buffer that pools the entry numbers of available free entries,

The empty entry management unit receives from the packet buffer control unit a first entry number that is an entry number in which a transmission packet is stored and a data error check result of the transmission packet,
When there is no data error, the empty entry management unit subtracts or sets the first counter corresponding to the first entry number in the error occurrence count table to 0, and sets the first entry number as empty. Pool it in the free entry buffer as an entry,
The empty entry management unit increments the first counter when there is a data error, and sets the first entry number when the value of the first counter is less than a preset error occurrence threshold. When the first entry number is pooled in the empty entry buffer as the empty entry buffer and the first counter value is equal to or larger than the error occurrence threshold, the first entry number is pooled in the empty entry buffer. Is excluded,
The empty entry management unit refers to the empty entry buffer and notifies the packet buffer control unit of any of the pooled entry numbers,
The packet buffer control unit is provided with a relay device that stores a received packet in the entry of the notified entry number.

According to the second solution of the present invention,
A failure processing method in a relay device that relays packets,
The relay device is
A packet buffer that includes a plurality of entries divided into regions, and temporarily stores received packets in each of the entries;
A free entry management unit for managing the packet buffer in units of entries;
A packet buffer control unit that extracts a packet from the packet buffer and detects a data error in units of entries;
A plurality of counters corresponding to each of the entries of the packet buffer, and an error occurrence number table for holding the number of occurrences of data errors for each entry;
An entry free entry buffer that pools the entry numbers of available free entries,

The empty entry management unit receives from the packet buffer control unit a first entry number that is an entry number in which a transmission packet is stored and a data error check result of the transmission packet,
When there is no data error, the empty entry management unit subtracts or sets the first counter corresponding to the first entry number in the error occurrence count table to 0, and sets the first entry number as empty. Pool it in the free entry buffer as an entry,
The empty entry management unit increments the first counter when there is a data error, and sets the first entry number when the value of the first counter is less than a preset error occurrence threshold. When the first entry number is pooled in the empty entry buffer as the empty entry buffer and the first counter value is equal to or larger than the error occurrence threshold, the first entry number is pooled in the empty entry buffer. Is excluded,
The empty entry management unit refers to the empty entry buffer and notifies the packet buffer control unit of any of the pooled entry numbers,
The packet buffer control unit is provided with a failure processing method for storing a received packet in the entry of the notified entry number.

  According to the present invention, in the packet buffer mounted on the network relay device, the failure occurring in the packet buffer is managed in divided entry units, so that the probability of failure of the network relay device can be reduced and the reliability is improved. Can be increased.

It is a block diagram which shows embodiment of this invention. It is a flowchart explaining a packet transmission process. It is a flowchart explaining the entry management method of a packet buffer. It is a flowchart explaining the failure determination method of a packet buffer.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing an embodiment of the present invention.
In FIG. 1, the network relay device 1 includes a packet buffer 10, a packet buffer control unit 20, an entry management unit 30, and a control unit 40 that controls various operations of the network relay device 1.

The packet buffer 10 is a buffer for temporarily storing packets received by the network relay device 1 and is managed by dividing the buffer area into N entries. Each entry is received by the network relay device 1. Packet data is stored. In the following description, the divided entry areas are represented as “E (0) to E (N−1) (N is an integer of 2 or more)”.
The packet buffer control unit 20 includes a data error detection unit 201, a packet reception unit 202, and a packet transmission unit 203.

  The packet reception unit 202 receives an entry number for storing the received packet in the packet buffer 10 from the entry management unit 30, and instructs the data error detection unit 201 to store the packet in the entry number (packet storage instruction) Put out. The data error detection unit 201 receives the above-described packet storage instruction from the packet reception unit 202, adds a data protection code to the packet, and stores the packet in the packet buffer 10. When the network relay device 1 transmits a packet, the data error detection unit 201 extracts the transmission packet from the packet buffer 10 and transmits it to the packet transmission unit 203. Further, when the data error detection unit 201 extracts a packet from the packet buffer 10, the data error detection unit 201 checks a data error using a data protection code attached to the packet data, and transmits the check result to the packet transmission unit 203. . The packet transmission unit 203 receives the transmission packet, the packet buffer entry number in which the transmission packet is stored, and the data error check result from the data error detection unit 201, and if the data error check result is an error, The transmission packet is discarded, otherwise the transmission packet is transmitted to the output line. Further, the packet transmission unit 203 transfers the entry number received from the data error detection unit 201 and the data error detection result to the entry management unit 30.

The entry management unit 30 includes a free entry management unit 301, an error occurrence count table 302, an error occurrence count threshold register 303, a fault entry count counter 304, a fault entry count threshold register 305, and a free entry buffer 306.
The empty entry management unit 301 instructs the packet reception unit 202 about the entry number when the packet received by the network relay device 1 is stored in the packet buffer 10. When the network relay device 1 transmits a packet, the packet transmission unit 203 receives the entry number in which the transmission packet is stored and the data error check result when the packet is extracted from the entry. The error occurrence count table 302 includes N counters equal to the number of entries in the packet buffer 10 and holds the error occurrence count for each entry in the packet buffer 10. In the following description, the counters are represented as C (0) to C (N-1) (N is an integer of 2 or more). The error occurrence number threshold register 303 is a register in which a threshold value is set in advance by the control unit 40, and this threshold value is set by the empty entry management unit 301 in counters C (0) to C (N−1) of the error occurrence number table 302. ) Is compared with the value held in The operations of the error occurrence count table 302 and the error occurrence count threshold register 303 will be described in detail in the description of FIG. 3 to be described later. The failure entry number counter 304 is a counter that holds the counter value of the error occurrence number table 302 and the threshold value set in the error occurrence number threshold register 303, and the number of entries exceeding this threshold value. The counter is compared with the threshold value of the failure entry number threshold register 305 whose value is preset by the control unit 40. The operations of the failure entry number counter 304 and the failure entry number threshold register 305 will be described in detail in the description of FIG. 4 to be described later.

  The control unit 40 is connected to the packet buffer 10, the packet buffer control unit 20, and the entry management unit 30, and sets the threshold values of the error occurrence number threshold value register 303 and the failure entry number threshold value register 305, as necessary. The function of reading the contents of the failure occurrence counter 302 and the failure entry counter 304 is provided. Further, when receiving a failure notification of the packet buffer 10 from the empty entry management unit 301, the control unit 40 performs a blocking process on the packet buffer 10.

FIG. 2 is a flowchart of packet transmission processing of the network relay processing device 1. In the following flowcharts, a case where a transmission packet is stored in the entry number M of the packet buffer 10 (M is one of the entry numbers 0 to N-1 obtained by dividing the packet buffer 10) will be described as an example.
In the first step S200, the data error detection unit 201 extracts packet data from the entry number M in which the packet is stored, and checks the data error using a data protection code assigned to the packet data. In step S201, the packet transmission unit 203 determines whether or not the data error check result is an error. If no error is detected, the packet transmission unit 203 proceeds to step S202 and transmits the transmission packet to the output line. If an error is detected, the process proceeds to step S203, where the transmission packet is discarded without being transmitted to the output line. Thereafter, the packet transmission unit 203 transmits the entry number M and the data error detection result to the entry management unit 30 in step S204.

FIG. 3 is a flowchart for explaining processing of the entry management unit 30.
In first step S301, the empty entry management unit 301 determines whether the data error check result of the entry number M received from the packet transmission unit 203 is an error. If it is not an error, the process proceeds to step S302, and the empty entry management unit 301 subtracts the counter C (M) corresponding to the entry number M in the error occurrence number table 302 or sets 0. Whether to subtract or set 0 can be determined in advance by the control unit 40 or the like. Thereafter, in step S305, the free entry management unit 301 pools the entry number M as a free entry. Note that the empty entries pooled in step S305 are stored in the empty entry buffer 306 of the empty entry management unit, and are reused to store other packets. If there is a data error in the determination in step S301, the process proceeds to step S303, and the empty entry management unit 301 adds 1 to the value of the counter C (M) and writes it back. Thereafter, in step S304, the empty entry management unit 301 compares the value of the counter C (M) with the threshold value of the error occurrence number threshold register 303 set in advance by the control unit 40, and determines the value of the counter C (M). Is smaller than the threshold value of the error occurrence number threshold register 303, the process proceeds to step S305, and the entry number M is pooled in the empty entry buffer 306 as an empty entry. If it is determined in step S304 that the value of the counter C (M) is equal to or greater than the threshold value of the error occurrence threshold register 303, the empty entry management unit 301 transitions to step S306, and the failure entry number counter 304 is incremented by 1. Is added. In step S307, if the entry number M is pooled in the empty entry buffer 306, it is excluded therefrom. Here, the entry at the time of transition to steps S306 and S307 is not reused because it is not pooled in the empty entry buffer 306 as an empty entry, and the network relay device 1 operates by excluding the entry by the empty entry management unit 301. Can continue.

  If a soft error has occurred in the entry to be reused, normal data is written when the entry is reused to recover from a temporary failure. That is, in this case, for example, in the first error, the processing procedure is step S301 → step S303 → step S304 → step S305. Further, when the error is recovered at the second time or less than the threshold value of the error occurrence number threshold register 303, the processing procedure is step S301 → step S302 → step S305. If a hard error has occurred in the entry to be reused, even if normal data is written when the entry is reused, it will not recover from the failure, and eventually the value of the counter C (M) will be It is equal to or greater than the threshold value of the error occurrence number threshold register 303. That is, when the number of errors is equal to or greater than the threshold value in the error occurrence number threshold register 303, the processing procedure is step S301 → step S303 → step S304 → step S306 → step S307.

In other words, the error occurrence frequency threshold register 303 can be a threshold for separating soft errors and hard errors. In the free entry buffer 306, entry numbers where no hard error has occurred are pooled.
Further, since the packet buffer 10 is tried to be reused as many times as the threshold value of the error occurrence number threshold register 303, there is a possibility that the packet is discarded as many times as the threshold value of the error occurrence number threshold register 303. However, it is possible to prevent the communication system from being affected by setting the threshold value of the error occurrence number threshold register 303 to be equal to or less than the packet discard number assumed as the network relay device in advance.

FIG. 4 is a flowchart of the failure determination process of the packet buffer 10.
In the first step S401, the empty entry management unit 301 compares the value of the failure entry number counter 304 with the threshold value of the failure entry number threshold register 305 preset by the control unit 40, and the counter 304 is less than this threshold value. Performs step S401 again, and otherwise transitions to step S402. In step S402, the empty entry management unit 301 notifies the control unit 40 of a packet buffer failure, and the control unit 40 closes the packet buffer 10 in response to the notification.

  Accordingly, since the failure of the packet buffer 10 is managed in units of entries obtained by dividing the area, for example, when a partial failure of the packet buffer area occurs, the operation of the network relay device is continued by excluding the entry. be able to. Specifically, the empty entry management unit 301 can acquire the entry / usable entry in which no hard error has occurred with reference to the empty entry buffer 306, and the entry number acquired by the packet receiving unit 202. To be notified.

Since the packet buffer 10 managed as a shared buffer shares one large-capacity buffer with a plurality of lines, even if a part of the packet buffer 10 is excluded and the network relay device is operated, the plurality of lines are connected. As long as the virtually allocated buffer amount does not change, it is possible to prevent the communication system from being affected.
In addition, the control unit 40 has a function of outputting the contents of the error occurrence count table 302 and / or the contents of the failure entry number counter 304 as necessary in the network relay device 1, and outputs or displays each content. You may do it.

  The present invention can be applied to various relay devices such as routers and node devices that relay packets and data.

1: Network relay device 10: Packet buffer 20: Packet buffer control unit 201: Data error detection unit 202: Packet reception unit 203: Packet transmission unit 30: Entry management unit 301: Empty entry management unit 302: Error occurrence count table 303: Error occurrence threshold register 304: Failure entry number counter 305: Failure entry number threshold register 40: Control unit

Claims (10)

A relay device for relaying packets,
A packet buffer that includes a plurality of entries divided into regions, and temporarily stores received packets in each of the entries;
A free entry management unit for managing the packet buffer in units of entries;
A packet buffer control unit that extracts a packet from the packet buffer and detects a data error in units of entries;
A plurality of counters corresponding to each of the entries of the packet buffer, and an error occurrence number table for holding the number of occurrences of data errors for each entry;
An entry free entry buffer that pools the entry numbers of available free entries,

The empty entry management unit receives from the packet buffer control unit a first entry number that is an entry number in which a transmission packet is stored and a data error check result of the transmission packet,
When there is no data error, the empty entry management unit subtracts or sets the first counter corresponding to the first entry number in the error occurrence count table to 0, and sets the first entry number as empty. Pool it in the free entry buffer as an entry,
The empty entry management unit increments the first counter when there is a data error, and sets the first entry number when the value of the first counter is less than a preset error occurrence threshold. When the first entry number is pooled in the empty entry buffer as the empty entry buffer and the first counter value is equal to or larger than the error occurrence threshold, the first entry number is pooled in the empty entry buffer. Is excluded,
The empty entry management unit refers to the empty entry buffer and notifies the packet buffer control unit of any of the pooled entry numbers,
The packet buffer control unit is a relay device that stores a received packet in the entry of the notified entry number.
The relay device according to claim 1,
The packet buffer control unit
A packet receiver for storing received packets in the packet buffer;
A data error detector that stores the received packet in the packet buffer;
The empty entry management unit instructs the packet reception unit on the entry number when storing the received packet in the packet buffer,
The packet receiving unit receives the entry number for storing the received packet in the packet buffer from the empty entry management unit, and issues an instruction to store the packet to the entry number to the data error detection unit,
The relay apparatus according to claim 1, wherein the data error detection unit receives a packet storage instruction from the packet reception unit, adds a data protection code to the received packet, and stores the received packet in the packet buffer.
The relay device according to claim 1 or 2,
The packet buffer control unit
A data error detection unit for extracting a transmission packet from the packet buffer;
A packet transmission unit that receives the transmission packet from the data error detection unit and transmits the transmission packet to an output line or discards the transmission packet;
The data error detection unit extracts a transmission packet from the entry of the first entry number in the packet buffer, checks a data error using a data protection code attached to the transmission packet, and outputs a check result to the packet Send to the transmitter,
The packet transmission unit transmits the transmission packet to the output line when no data error is detected based on the check result, and discards the transmission packet without transmitting to the output line when the data error is detected. And
The packet transmission unit transmits the first entry number and the data error check result to the empty entry management unit.
The relay device according to any one of claims 1 to 3,
A failure entry number counter that holds the number of entries not used as the packet buffer;
The empty entry management unit adds the failure entry number counter when the value of the first counter is equal to or greater than the error occurrence number threshold,
The relay apparatus, wherein the empty entry management unit outputs a failure notification for closing the packet buffer when the failure entry number counter is equal to or greater than a predetermined failure entry number threshold.
The relay device according to claim 4,
A relay apparatus, further comprising a control unit that performs a blocking process of the packet buffer when a failure notification for blocking the packet buffer is received from the empty entry management unit.
The relay device according to claim 4,
The relay apparatus further comprising a control unit for setting the failure entry number threshold.
The relay device according to any one of claims 1 to 4,
The relay apparatus further comprising a control unit for setting the error occurrence frequency threshold.
The relay device according to any one of claims 1 to 7,
The relay apparatus according to claim 1, wherein the error occurrence frequency threshold is set to a threshold for separating a soft error and a hard error.
A relay device according to any one of claims 1 to 8,
The relay apparatus further comprises a function of outputting or displaying the contents of the error occurrence number table and / or the contents of the failure entry number counter.

A failure processing method in a relay device that relays packets,
The relay device is
A packet buffer that includes a plurality of entries divided into regions, and temporarily stores received packets in each of the entries;
A free entry management unit for managing the packet buffer in units of entries;
A packet buffer control unit that extracts a packet from the packet buffer and detects a data error in units of entries;
A plurality of counters corresponding to each of the entries of the packet buffer, and an error occurrence number table for holding the number of occurrences of data errors for each entry;
An entry free entry buffer that pools the entry numbers of available free entries,

The empty entry management unit receives from the packet buffer control unit a first entry number that is an entry number in which a transmission packet is stored and a data error check result of the transmission packet,
When there is no data error, the empty entry management unit subtracts or sets the first counter corresponding to the first entry number in the error occurrence count table to 0, and sets the first entry number as empty. Pool it in the free entry buffer as an entry,
The empty entry management unit increments the first counter when there is a data error, and sets the first entry number when the value of the first counter is less than a preset error occurrence threshold. When the first entry number is pooled in the empty entry buffer as the empty entry buffer and the first counter value is equal to or larger than the error occurrence threshold, the first entry number is pooled in the empty entry buffer. Is excluded,
The empty entry management unit refers to the empty entry buffer and notifies the packet buffer control unit of any of the pooled entry numbers,
The packet buffer control unit is a failure processing method for storing a received packet in the entry of the notified entry number.

Images