JP2008199432A - Data transfer device and health check data processing method - Google Patents

Abstract

The present invention relates to a data transfer apparatus and a health check data processing method, which can reduce a health check response delay affected by a traffic amount of data transfer and easily determine an optimum timeout value of a health check monitoring timer.
A health check data-specific CRC code, which is different from a CRC code for user communication data, is added as an error check code to be added to health check data on the side where health check data is transmitted. On the other hand, an error check (7-5) is performed using a CRC code for user communication data, and if an error is detected, a double CRC check (1-1) is performed using a CRC code unique to health check data. If the result is normal, it is determined that the received data is health check data, the health check is detected (1-2), and the health check response is placed at the top of the processing queue or in a preferential order position (1- 4) and store.
[Selection] Figure 1

Description

  The present invention relates to a data transfer device and a health check data processing method, and more particularly, to a data transfer device and a health check data processing method for processing failure detection health check data transmitted and received between data transfer devices without delay.

  The health check is performed by periodically sending health check data to a partner device to confirm that the partner devices are operating normally among a plurality of devices. This is used as a technique for judging the normality of the counterpart device based on the response data and response time for the check data.

  FIG. 6 shows a conventional health check operation sequence. As shown in the figure, when the health check data 6-1 is transmitted from the partner apparatus B to the own apparatus A, the own apparatus A stores the health check data 6-1 at the end of the reception data queue 6-2. The received data that has been stored and received in the received data queue 6-1 is extracted from the received data queue 6-1 in order from the top in the FIFO (First-In First-Out) method, and processed in order. . Therefore, a processing delay due to a waiting time in the reception data queue 6-2 occurs.

  FIG. 7 shows a flow of processing of the received data. As shown in the figure, after the reception circuit establishes synchronization of the received data packet (7-1), it is determined whether the received data is addressed to the own device (7-2). The received data is transferred to the memory device via a transfer process such as (7-4). If it is not addressed to its own device, the received data is discarded (7-3).

  Error check processing such as CRC (Cyclic Redundancy Check) code error check and framing error check is performed on the received data addressed to the own device transferred to the memory device to determine whether the received data is normal or abnormal. (7-5). Received data in which an error such as a CRC code is detected is discarded (7-6). The result of the error check is usually notified by an interrupt request (IRQ) from the receiving circuit to the CPU.

  If the result of the error check is “normal (OK)” and there is already received data before receiving the received data, the received data is delivered to the packet processing unit via the FIFO received data queue and processed. (7-7).

  The packet processing unit determines whether or not there is a received data packet stored in the received data queue (7-8), and if there is a received data packet, stores the received data packet in the FIFO-type received data queue. Therefore, the received data is retained in the received data queue and waits for the completion of processing of the received data (7-9).

  When the received data stored in the received data queue runs out, processing of the received data packet is started (7-10). In this case, the larger the number of received data packets, the longer the waiting time until the received data packet is processed. As in the case of health check data, the data processing from the packet reception is completed and response data is sent to the partner device. When the time until the response data arrives at the partner device is specified, delay due to the waiting time in the reception data queue becomes a problem.

  The delay due to the waiting time in the reception data queue will be described using a queue model. The queue model can be evaluated by the following calculation formula using the following parameters. However, for simplification, the data transfer time on the transmission path is very small compared to the processing time, so it is ignored and traffic is only in one direction.

First, an average waiting time until the health check data (message) is processed together with other received data is calculated. Here, Tw is an average waiting time (expected value of waiting time until the health check data is processed), Ts is an average service time (average value of time for processing received data (message) on the processing system, and ρ is Average utilization rate (coefficient representing the degree of processing of received data), λ is the average arrival rate (number of received data (messages) arriving per unit time), μ is the average service rate (received data (messages that can be processed in unit time)) ) Average number))
Tw = (ρ / 1−ρ) * Ts, ρ = λ / μ (1)
Thus, when shown in a graph, FIG. 8 is obtained.

  As a specific example, consider a guideline of an optimum value as a timeout value of the health check monitoring timer in a processing system in which the arrival interval Ta of received data at normal time is 500 ms and the average service time Ts is 200 ms. At this time, Tw (average waiting time) = 0.2, and Tw (average waiting time) when Ta (data arrival interval) changes from 0.5 to 0.3 due to an increase in received data (message). ) Changes as follows.

  Normally, Ta (average data arrival time) is 0.5, and Ts (average service time) is 0.2. In this case, since λ (average arrival rate) = 1 / Ta = 2 and μ (average service rate) = 1 / Ts = 5, ρ (average utilization rate) = λ (average arrival rate) / μ (average service) Rate) = 0.4, and Tw (average waiting time) = (0.4 / 0.6) * 0.2 = 0.13.

  When Ta (average data arrival interval) becomes 0.3 due to an increase in traffic such as data transfer amount, λ (average arrival rate) = 1 / Ta = 3.3, μ (average service rate) = Therefore, ρ (average utilization rate) = λ (average arrival rate) / μ (average service rate) = 3.3 / 5 = 0.66, and Tw (average waiting time) = (0.66 / 0.34) * 0.2 = 0.388.

  From the above-obtained value, it is necessary to set the timeout value of the health check monitoring timer to 0.13 or more in normal time, and considering the increase in traffic, the timeout value must be 0.388 or more. Further, while considering the margin of the processing time Ts and the average waiting time Tw, the problem is how to optimally set the timeout value of the health check monitoring timer because of the necessity of early confirmation of the health check.

  As prior art documents related to the present invention, in Patent Document 1 below, when a health check signal is received, even if there is data being transmitted, the transmission is interrupted in response to a health check response request, and the health check is performed. A technique for sending a response with priority is disclosed. Patent Document 2 below discloses a technology that extends the health check reception monitoring time during reception of application data from the other party, thereby preventing a health check data monitoring timeout from occurring. .

In Patent Document 3 below, a health check signal is transmitted at a predetermined cycle, and when there is no response from the reception side, it is determined that a failure has occurred on the reception side, and the reception side has a predetermined period. A technique for determining that a failure has occurred on the transmission side when a health check signal is not received is disclosed.
Japanese Patent Laid-Open No. 03-019057 JP-A 63-104544 JP 62-175836 A

  In the data transfer device, exchange of data called health check is executed in order to confirm the normality of the data transfer function between devices or inside the device. In this case, if the transfer amount of the original user data for the communication service increases, the health check data is retained in the reception data queue until it is delivered to the analysis process, and it takes a long time to respond to the health check. In the worst case, there is a case where it is determined that the health check is abnormal (monitoring timer timeout) due to the non-return of the health check response.

  In the conventional health check, the optimum value as the timeout value of this monitoring timer varies due to the influence of the traffic volume of data transfer, etc., so as a rough standard value including a margin based on the measured value etc. There was no choice but to determine the optimum value. The present invention provides a data transfer apparatus and a health check data processing method that can reduce the delay in health check response processing affected by the amount of data transfer traffic and can easily determine the optimum timeout value of the health check monitoring timer. To do.

  In order to solve the above-mentioned problem, the present invention provides a health check data to a data transfer device that transmits health check data in a data transfer device that transfers an error check code to user communication data. As an error check code, an error check code in which an error is detected by an error check code for the user communication data in the data transfer device on the receiving side, and an identification code unique to the health check data is used as the health check data. A data transfer device that receives the health check data, and performs an error check on the received packet with the error check code for the user communication data, and then the error check The received data in which an error was detected Means for determining whether or not the received data is given the identification code unique to the check data, and the received data given the identification code unique to the health check data is received at the head of the reception data processing waiting queue or other received data Means for arranging and storing the data in order positions preferential to data.

  Further, a CRC code different from the error check code for the user communication data is used as an identification code unique to the health check data. In addition, as the identification code unique to the health check data, a CRC code having a data range different from the error check target range by the error check code for the user communication data as an error check target range is used. Further, as the identification code unique to the health check data, fixed data at a predetermined position of the health check data is used, and the health check data is identified by comparing the data at the predetermined position of the received data with the fixed data. .

  According to the present invention, in the data transfer device, the health check data can be identified by the receiving device using the error check code, and the receiving device waits for processing stored in the other received data queue. By processing with higher priority than the received data, the health check response can be returned immediately without being affected by the number of received data waiting for processing stored in the received data queue, and the amount of data transfer traffic Therefore, it is possible to easily determine the optimal timeout value of the health check monitoring timer with high accuracy.

  An error check code such as a CRC code is given to transfer data of a typical communication protocol such as Ethernet (registered trademark) or ATM (Asynchronous Transfer Mode). In the present invention, an error check code such as the CRC code is used. Since the health check data is identified by the receiving side device by performing the double check of the code, it can be easily applied to data transfer devices of these general-purpose communication protocols.

  As shown in FIG. 7, the flow at the time of data reception in a general data transfer device is as follows. After the synchronization of the received data is established, the received data addressed to the own device is transferred to the memory device via DMA or the like. An error check, such as a CRC error or a framing error, is performed to determine whether it is abnormal, and the check result is notified to the corresponding processing unit by a CPU interrupt.

  Even if the received data is health check data, if the received data exists before the reception of the health check data, the health check data is stored in the received data queue, and the health check data is received after the processing of the received data is completed. Since the check data is processed, the health check response is delayed.

  Therefore, in the present invention, health check data is distinguished from other received data such as user data in error check processing such as CRC error or framing error performed on all received data before storage in the received data queue. The extracted health check data is stored at the head of the received data queue or at a preferential order position.

  In order to realize the above function, the health check data is an error check code such as a CRC code, which is a code in which an error is detected on the reception side, and an identification code unique to the health check data is assigned to the health check data. In order to distinguish the original abnormal data and health check data from the received data determined to be abnormal data by the error check of the received data on the receiving side, Perform the process.

  As shown in FIGS. 1A and 1B, the processing consisting of steps 7-5, 7-6 and 7-7 surrounded by the broken line in FIG. 7 is performed as shown in FIG. After the abnormality determination (7-5), double CRC code check (1-1) based on the unique CRC code assigned to the health check data packet for the received data determined to be abnormal I do.

  If the result of the double CRC code check (1-1) is determined to be “normal”, the received data is recognized as health check data, and an interrupt request (IRQ) is generated to the CPU ( 1-2). When it is determined that the result of the double CRC code check (1-1) is “abnormal”, the received data is not health check data, but an interrupt request for discarding the received data as conventional abnormal received data ( IRQ) is generated (7-6).

  In this way, a double CRC code check is performed on the received data, and a CRC code that is always determined to be abnormal by the first CRC code check is assigned to the health check data to generate a trap. A function for recognizing health check data while maintaining the normality confirmation function of the original received data by giving a CRC code unique to the health check data so that it can be recognized as health check data in the second CRC code check to add.

  When it is determined that the double CRC code check (1-1) is normal and health check data is detected, an interrupt request (IRQ) is generated to the CPU (1-2), and the interrupt request (IRQ) As shown in FIG. 1 (b), prepare to store a health check response (check the health check result according to the process and make it a response) in the queue (1-3), and receive the prepared response It is arranged at the head of already received data (untransmitted data) in the data queue (1-4).

  Note that the time required for the health check process due to the generation of the interrupt request (1-2) differs depending on the process pattern. In the data transfer device, the priority of the reception processing system is the highest, and there is no particular problem even if the health check process is executed by extending the interrupt process, but when a higher level interrupt process is requested by others Can be configured to adjust the placement position of the health check response in the received data queue in accordance with the priority of the interrupt processing.

  The packet processing unit determines whether or not there is already received data stored in the received data queue (7-8). If there is received data, the received data is extracted from the head using the FIFO method and processed, and the health check response Since the data is arranged at the head of the received data queue or at a preferential order position, the data can be promptly transmitted without delaying the processing waiting time affected by the number of other received data.

  However, if it is determined as “abnormal” in the first CRC code check (7-5), the CRC error counter as error check data statistical information is not incremented, and the second CRC code check (1-1). When it is determined as “abnormal”, the CRC error counter as error check data statistical information is increased.

The following four methods can be used as a method for identifying health check data in the second CRC code check (1-1).
(1) Use the same CRC code check method as the first CRC code check (7-5), and check by changing the target range of the data to be checked.
(2) Two different types of CRC code check methods (for example, CRC-32 and CRC-16) are prepared, and different CRC code methods are applied to health check data and user data.
(3) A CRC code different from the user data is attached to the health check data and transmitted. After the abnormality is detected on the receiving side, the health check data is recalculated to check the normality, and the health check data is checked. Identify
(4) When the health check data is fixed data that does not include parameters (variables), the receiving device stores the fixed data and its arrangement position, and the receiving apparatus stores the arrangement position data in the reception data. The health check data is identified by detecting the coincidence with the stored fixed data.

  FIG. 1C shows a format of transfer data to which an error check code is added. At the time of data transfer, in order to guarantee the normality of transmission data (frame, packet, etc.), an error check code such as a CRC code or a checksum code is added to the transmission data and transmitted, and the error check code is received on the receiving side. The received data is checked based on this, and if an abnormality is detected, the received data is discarded or retransmitted.

  In the present invention, for example, a second CRC code is assigned to user data, a first CRC code is assigned to health check data, and the received data is sent to the second CRC on the receiving side. After checking the system and detecting an abnormality, the first CRC system is checked. If the check result is “normal (OK)”, it is identified as health check data.

  The overall flow of the reception process according to the present invention is shown in FIG. In FIG. 2, the same processes as those in the conventional reception process shown in FIG. 7 are denoted by the same reference numerals, and the same processes as those described in FIG. Is omitted. The overall flow of the reception process according to the present invention is obtained by replacing the conventional reception process surrounded by a broken line shown in FIG. 6 with the process shown in FIG.

  As an embodiment of the present invention, a configuration example in the case of two data transfer apparatuses as shown in FIG. 3 will be described. As shown in the figure, the data transmission / reception device 3-1 transmits a health check data packet to the data transmission / reception device 3-2, and the data transmission / reception device 3-2 sends a health check response to the health check data packet to the data transmission / reception device 3. Return to -1. Here, it is assumed that the transmission medium is Ethernet (registered trademark) as an example.

  The flow of the health check process in the above configuration example will be described with reference to FIG. First, the flow of processing during reception will be described. In the data transmitting / receiving device 3-2, whether the received data is general user data or health check data, the data normality check (CRC code check) is executed through the receiving circuit and the DMA circuit, and the received data is stored in the memory. (4-1). At that time, the reception (transmission) buffer control mechanism manages the head of the reception data by the descriptor, the reception status (normal or abnormal), and the control flag.

  When an abnormality is detected by the normality check of the received data, an interrupt (trap) is generated for the CPU, and the CRC error trap processing unit registered in the corresponding vector table entry is called (4-2). The CRC error trap processing unit reads the received data from the buffer (4-3), and performs a second CRC code check on the received data (4-4).

  If the result of the second CRC code check is normal, it is determined whether the health check is good or not, and a health check response is returned to the data transmitting / receiving device (3-1) through the transmission process described below. When the CRC code check result is abnormal, the received data is discarded. Only when the received data does not correspond to the health check, the number of occurrences of CRC check abnormality as a statistic is added. As a CRC coding method, a CRC-32 generator polynomial in Ethernet (registered trademark) can be used.

  At the time of transmission, it is not necessary to change the polynomial of the CRC code system, and CRC-32 is used. However, the calculation target is different. At the time of transmission, a health check response is placed at the head of the transmission data queue by a flag in the transmission descriptor (4-5). Two types of flags are used, the first flag is used as a DMA transfer processing flag, and the second flag is used as a final flag.

  For the descriptor whose first flag closest to the head is off, the second flag is turned off, the health check response packet is chained to the next descriptor entry, and then the flag with the second flag turned off is set. Turn it on and restart the transmit DMA (4-6).

  As a result of this processing, as shown in FIG. 5A, when two untransmitted packets are chained in this order by the descriptors 5-1 and 5-2, as shown in FIG. The health check response descriptor 5-3 is placed next to the descriptor 5-1 of the first unsent packet, interrupted before the descriptor 5-2 of the next unsent packet, and the health check response is received. (Transmission) It is arranged at the head of the data queue or at a preferential order position. After the transmission of the health check response packet is completed, the process returns to normal reception packet processing (4-7).

It is a figure which shows the identification of the health check data by this invention, and the storing to a queue. It is a figure which shows the whole flow of the reception process by this invention. It is a figure which shows the structural example of the Example of this invention. It is a figure which shows the flow of a process of the health check of the Example of this invention. It is a figure which shows chain incorporation of the health check response of the Example of this invention. It is a figure which shows the operation | movement sequence of the conventional health check. It is a figure which shows the flow of the process of the conventional reception data. It is a graph of the average waiting time until health check data is processed.

Explanation of symbols

1-1 Double CRC Check Processing 1-2 Health Check Detection Processing 1-3 Processing to Prepare Health Check Response in Queue 1-4 Processing to Place Prepared Response at Top of Unsent Data

Claims (5)

In a data transfer device for transferring an error check code to user communication data,
An error is detected by an error check using the error check code for the user communication data in the data transfer device on the reception side as an error check code to be added to the health check data in the data transfer device on the health check data transmission side. An error check code, comprising means for sending an identification code unique to the health check data to the health check data,
The data transfer device that receives the health check data performs error check on the received packet using the error check code for the user communication data, and then receives the received data in which an error is detected in the error check. Means for determining whether or not the received data has an identification code unique to the health check data, and the received data to which the identification code unique to the health check data has been assigned, And a means for storing the data arranged in an order position preferential to other received data.   The data transfer apparatus according to claim 1, wherein a CRC code different from the error check code for the user communication data is used as the identification code unique to the health check data.   2. The CRC code having a data range different from an error check target range by the error check code for the user communication data as an error check target range is used as the identification code unique to the health check data. The data transfer device described in 1.   The health check data is identified by using fixed data at a predetermined position of health check data as an identification code unique to the health check data, and comparing the data at a predetermined position of received data with the fixed data. The data transfer apparatus according to 1. In the health check data processing method between data transfer devices that transfer error check codes to user communication data and periodically send and receive health check data,
In the data transfer device on the transmission side of the health check data, an error is detected by an error check using the error check code for the user communication data in the data transfer device on the reception side as an error check code to be added to the health check data. An error check code, an identification code unique to the health check data is attached to the health check data and transmitted,
In the data transfer device that receives the health check data, after performing an error check on the received data using an error check code for the user communication data, the received data in which an error is detected in the error check is received. Whether the received data is given the identification code unique to the health check data, and the received data given the identification code unique to the health check data is A health check data processing method, wherein the health check data processing method is characterized in that it is arranged and stored in an order position having priority over received data.

Images