CN100459529C - Time delay characteristic measuring method in data communication network - Google Patents

Abstract

A method for measuring character of time delay in data communication network includes generating at least one measurement message by sending end and using microcode / hardware to label the first sending time seal on said message then sending them out, separating out said message and labeling the first receiving time seal on them as well as labeling the second sending time seal on separated out message then sending said message back to sending end by receiving end, separating out said message and labeling the second receiving time seal on said message by sending end, calculating out time delay character between sending and receiving ends according to four labeled time seals.

Description

The Measurement Method of Delay Characteristic in Data Communication Network

technical field

The invention relates to data communication network technology, in particular to a method for measuring time delay characteristics in a data communication network.

Background technique

In a data communication network where data, voice, and video are integrated into one network, the data transmission of voice and video puts forward higher and higher requirements for network bearer, and the delay characteristic is an important indicator to measure the performance of the data bearer network. The delay characteristics refer to characteristics related to data transmission time in the network, including indicators such as latency (Latency) and latency jitter (Jitter). Latency refers to the time required for data packets to be transmitted in the network; delay jitter refers to the change in the time it takes for multiple data packets to transmit in the same segment of the network.

The delay characteristic is of great significance to measure the performance of the existing data bearer network. For example, in the next-generation voice service network, the two important indicators of delay and delay jitter can be provided in real time, which can provide reliable data for diagnosing network faults. In addition, the delay characteristic can also provide a fast and accurate basis for the performance analysis of user services. According to the online delay characteristic measurement results, network performance bottlenecks can be found, thereby optimizing the operating network, improving network maintainability and manageability, and ultimately improving customer satisfaction. After the delay characteristic is obtained, it can be applied to enhance NTP message processing, thereby improving the accuracy of data network time by 2 to 3 orders of magnitude.

Based on the above applications, how to measure the delay characteristics of data packets on the data communication network online has become a problem to be solved. In actual network operation, since the data packet sending and receiving devices may be thousands of kilometers or more apart geographically, it is unrealistic to complete the above measurement with a test instrument in the laboratory. For this reason, some other practical methods are also used in the prior art to measure the delay characteristics.

Method 1. The global positioning system (GPS) or other equipment that can unify the time of the entire network provides a clock with an accuracy of the order of μs, so that the absolute time error of the two test instruments in different geographical locations is maintained at the order of μs. By recording the data The sending time and receiving time of the packet are used to obtain the delay characteristics of the data packet transmission on the data communication network. This method is a relatively common method for measuring delay characteristics at present.

This method has the following disadvantages:

(1) It is necessary to use a GPS system that can provide a measurement clock, as well as a message generator and receiver used in conjunction with the GPS system. Because the GPS system is expensive, the cost of this measurement method is high, and the requirements for measuring instruments are relatively strict.

(2) It is necessary to reserve a special measurement port in the device under test to send measurement data through the measurement port, and the measurement port cannot be used to send actual business data packets, so on the one hand, the utilization rate of the device is reduced; on the other hand , the delay characteristic of the service flow cannot be measured in the actual service environment, which belongs to offline measurement.

Method 2: By issuing a general diagnostic command "ping" to the network device, the software records the sending and receiving time of the data packet, so as to obtain the delay characteristic.

Because the time recording operation of this method is realized by means of software, it is easily affected by factors such as software execution efficiency, task scheduling, and heavy-duty tasks in the system, resulting in large errors in measurement. Generally, the time delay characteristics obtained by this method are almost unusable in the voice bearer network where the measurement accuracy is required to be on the order of ms.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a method for measuring delay characteristics in a data communication network, which can control measurement errors and make the measurement accuracy meet the requirements of practical applications.

In order to achieve the above object, the technical solution of the present invention is specifically realized in the following way:

A method for measuring delay characteristics in a data communication network, the method comprising the following steps:

a. When sending a data packet to the receiving end, the sending end generates at least one measurement message, marks the generated measurement messages one by one with the first sending timestamp through its own microcode/hardware, and sends the above-mentioned measurement message; The receiving end separates the measurement message from the received data packet, and uses its own microcode/hardware to mark the first reception timestamp for the separated measurement message one by one;

b. The receiving end uses its own microcode/hardware to mark the second sending time stamps for the separated measurement messages one by one, and sends the above-mentioned measurement messages to the sending end through the data communication network; Separate the measurement message, and use its own microcode/hardware to mark the second reception timestamp for the separated measurement message one by one;

c1. The sending end and/or the receiving end generate a self-loop measurement message whose destination address is itself, mark the self-loop measurement message with a self-loop sending timestamp through microcode/hardware, and send the self-loop measurement message;

c2. The sending end and/or the receiving end separate the self-loop measurement message from the received data packet, and use microcode/hardware to mark the self-loop reception time stamp for the separated self-loop measurement message;

c3. Calculate the additional delay according to the self-loop receiving time stamp and self-loop sending time stamp of the sending end and/or receiving end;

c. Calculate the delay in the delay characteristic between the sending end and the receiving end according to the first sending time stamp, first receiving time stamp, second sending time stamp and second receiving time stamp marked in the measurement message ; Subtract the additional time delay from the time delay to obtain the time delay after compensation and correction.

The measurement message generated in step a has a measurement message identifier;

Then, the method for separating the measurement message is: according to the measurement message identifier, the flow classification matching method is used to separate the measurement message.

The flow classification matching method is realized by using a three-state content addressable memory.

The delay characteristic includes a time difference, and step c also includes a method for calculating the time difference in the delay characteristic, which is: adding the sum of the first receiving timestamp of each measurement message to the second sending timestamp, and subtracting the first The sum of the sending time stamp plus the second receiving time stamp is divided by two; then the calculation results of all measurement messages are averaged to obtain the time difference between the sending end and the receiving end.

The delay characteristic includes a delay, and the method for calculating the delay in the delay characteristic between the sending end and the receiving end in step c is: add the first receiving timestamp of each measurement message to the second receiving time stamp The sum of the time stamps, subtract the sum of the first sending time stamp plus the second sending time stamp, and divide the obtained difference by two to get the delay of each measurement message; then average the time delay of all measurement messages , to get the time delay of the sender and receiver.

The method for calculating the additional time delay described in step c3 is:

Subtract the self-loop sending timestamp from the self-loop receiving timestamp of the sender to obtain additional delay;

Or, subtract the self-loop sending timestamp from the self-loop receiving timestamp of the receiving end to obtain additional delay;

Alternatively, the first additional delay obtained by subtracting the self-loop sending timestamp from the self-loop receiving timestamp of the sending end, and the second additional delay obtained by subtracting the self-loop sending timestamp from the self-loop receiving timestamp of the receiving end are averaged , to obtain an additional delay.

The delay characteristics also include delay jitter, and in step a, the sending end generates two or more measurement messages within a period of time;

Then step c also includes a method for calculating the delay jitter in the delay characteristic, which is: select the maximum delay and the minimum delay from the delays of all measurement messages generated within this period, and reduce the maximum delay by Remove the minimum delay to obtain the delay jitter of the sending end and the receiving end.

Visible by above-mentioned technical scheme, the measurement method of time delay characteristic in this data communication network of the present invention emulates measurement message by upper layer software, and carries out measurement message by microcode or hardware (microcode/hardware) of network equipment Timestamp marks to obtain the transmission time of the measurement message in the network, and finally calculate the delay characteristics of the measured network.

The method generates a measurement message in a real business environment, and the time delay parameter obtained by the measurement can reflect the business flow delay situation in the real environment, which belongs to the online measurement. In addition, this method utilizes the microcode/hardware of the network device to record the time without the need for additional measuring instruments and auxiliary equipment with special functions, so the implementation is simple and the accuracy of time recording is much higher than that of the upper-layer software.

Furthermore, this method can calculate the extra delay through the self-loop at the local end, and compensate and correct the measured delay characteristics, so as to obtain higher measurement accuracy, and a message delay of 10 μs can be obtained on the 100M Ethernet precision.

Description of drawings

Fig. 1 is the flow process that realizes delay characteristic measurement in a preferred embodiment of the present invention;

FIG. 2 is a flowchart of error compensation in a preferred embodiment of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and examples.

FIG. 1 shows a flow of measuring delay characteristics of data packets in a certain segment of network in a preferred embodiment of the present invention. Assume that the originating device of the network is A, and the receiving device is B, and the originating device A and the receiving device B are respectively any network devices at both ends of the network, such as a router. The router can be implemented by using a network processor. The network processor is a programmable data frame processing unit on the data communication device. Usually, dozens to hundreds of processing units operate in parallel, and can complete table lookup, frame processing, and reporting. Operations such as file forwarding and discarding, and has hardware and microcode functions. The method specifically includes the following steps:

Step 101, the upper layer software of the originating device A simulates the actual service flow to generate at least one measurement message, marks the first transmission time stamp t1 one by one on the generated measurement message through the microcode/hardware, and marks the measurement message with the time stamp The packet is sent to receiving device B.

In this step, the upper-layer software can generate a measurement message with any source address, destination address, message length, priority, protocol number and port number. This simulation method is commonly used in the prior art and will not be described here. In addition, when the upper layer software generates the measurement message, it will mark the measurement message with a specific measurement message identifier.

The microcode (Picocode/Microcode) is a proprietary binary code executed by the network processor. Define a time stamp action in the microcode, so that the microcode can mark the time stamp on the measurement message sent by the upper layer software. In an actual system, the above functions can also be accomplished by hardware such as digital logic circuits.

Step 102, after the measurement message is sent to the receiving end device B via the data communication network, the receiving end device B separates the above measurement message from many data packets through its own microcode/hardware, and writes the measurement message on the separated measurement message Mark the first reception timestamp t2 one by one. The data communication network includes Internet, national backbone network, provincial backbone network, metropolitan area network, access network, enterprise network and various private networks.

In this step, the microcode/hardware separates the measurement message according to the measurement message identifier. The above separation process adopts the existing flow classification matching method, for example, it can be realized by Tri-State Content Addressable Memory (TCAM), etc., which will not be described here.

Step 103: The receiving device B marks the received measurement messages with the second transmission time stamp t3 one by one through its own microcode/hardware, and then sends the measurement messages to the sending device A through the egress port.

At the output port of the receiving device B, the processing time of a measurement message from the timestamp to the last bit of the last byte is less than 20 microseconds, and the above-mentioned processing time is basically constant, so the time of this method is The delay measurement accuracy can reach 0.1ms, which meets the accuracy requirements of the voice network.

Step 104: After the measurement message is sent back to the originating device A via the data communication network, the microcode/hardware of the originating device A separates the measurement message through flow classification and matching, stamps the second receiving time stamp t4 on it one by one, and then sends the measurement report The file is sent to the upper layer software of the originating device A.

Step 105, the upper layer software of the originating device A extracts all time stamps recorded in the measurement message, and calculates the delay characteristics of the network between the transmitting and receiving end devices.

In the above process, the upper-layer software can calculate time delay characteristics such as time difference, time delay and time delay jitter based on all recorded time stamps.

1. Usually, the time of network devices is inconsistent. The time difference refers to the time difference marked by the sending device A and the receiving device B at the same time. For example, at a certain moment, the time of the sending device A is 11:00, and the time of the receiving device B is 9:00. The time difference between the two is 2 hours.

The method of calculating the time difference is as follows: the upper layer software judges whether the round-trip paths recorded in the measurement messages are consistent one by one. If the round-trip paths are consistent, calculate the time difference z recorded in each measurement message according to the formula (1), and then compare all the measurement messages to The calculation results are averaged to obtain the time difference between the sending and receiving devices.

z＝(t2+t3-t1-t4)/2 (1)

On an IP network, the round-trip path of the packet can be obtained through ICMP packet path tracing; on an MPLS network, when an MPLS LSP tunnel is established, the round-trip path of the packet can be obtained by using the signaling path recording method of MPLS TE.

When the back and forth paths of some measurement messages are inconsistent, the time difference of all measurement messages is calculated according to the formula (1) and then averaged, which also has a certain reference value for obtaining the delay characteristics between sending and receiving devices.

2. Calculate the time delay x of each measurement message according to the formula (2), and then average the time delays of all the measurement messages to obtain the time delay of the sending and receiving device.

x＝t2-t1-z＝t2-t1-(t2+t3-t1-t4)/2 (2)

3. Calculate the delay jitter. Delay jitter (jitter) refers to the delay difference between a packet with the largest delay and a packet with the smallest delay within a period of time. Since the delay jitter is not affected by the delay of the message, that is, it is not affected by the time difference between the two ends and the classification processing of the message flow, so its measurement accuracy is 1-2 orders of magnitude higher than that of the delay.

In the process shown in FIG. 1 , the time stamp marked by the microcode/hardware on the measurement message is provided by the hardware clock. The hardware clock uses a local crystal oscillator to generate a signal source, which can be realized by using a 64-bit counter with an accuracy of 10ppm. During the delay characteristic measurement process, when the clock frequency of the source device A and the receiver device B are inconsistent, the time adjustment factor can be notified between A and B devices through the existing protocol of the data communication network, so that the clock frequency of the two devices be consistent.

In high-speed Ethernet, a packet enters the ingress port of the device from the first bit of the first byte, performs flow classification matching, and stamps the time. The entire process is completed by microcode/hardware within 20 microseconds. the following. Since the actual service packets are not time-stamped, the above-mentioned processing time is additionally introduced, which will increase measurement errors.

In a bearer voice network that does not require high measurement accuracy, the measurement methods described in the above embodiments can already meet the accuracy requirements. Furthermore, in order to improve the measurement accuracy of the delay characteristics and meet the needs of other network applications, the local end self-loop can also be performed during the measurement process, so as to calculate the microcode/hardware introduced in the marking timestamp and flow classification The additional time delay Tz is used to compensate and correct the final measurement results. The specific process is shown in Figure 2:

Step 201, the upper layer software of the originating device A generates a self-loop measurement message whose destination address is its own address, and marks the self-loop transmission time stamp t5 on the above-mentioned self-loop measurement message through microcode/hardware, and sends the self-loop measurement message. Measurement messages.

Step 202, after the self-loop measurement message returns to the originating device A through the data communication network, the microcode/hardware of the originating device A separates the self-loop measurement message through flow classification matching, stamps the self-loop receiving time stamp t6 on it, and then sends the self-loop The ring measurement message is sent to the upper layer software of the originating device A.

Step 203 , the upper layer software calculates the extra time delay Tz according to the self-loop sending time stamp t5 and the self-loop receiving time stamp t6 recorded in the self-loop measurement message. Among them, Tz=t6-t5.

Step 204, the upper-layer software compensates and corrects the delay characteristic according to the calculated extra delay Tz.

For example, for the time delay x, the time delay after compensation and correction is Tx=x-Tz.

In the above process, the additional delay Tz can also be calculated by the receiving device B through the local self-loop; or, the first additional delay Tz1 can be calculated by the transmitting device A, and the second additional delay Tz2 can be calculated by the receiving device B , and then take the average value of the first extra time delay Tz1 and the second extra time delay Tz2 to obtain the extra time delay Tz.

The actual test shows that the time delay characteristic obtained after compensation and correction can reach the precision of 10μs order.

In addition, the upper-layer software can also send measurement messages multiple times as needed and record the time, and then use statistical methods to obtain more accurate delay characteristics of the corresponding network.

It can be seen from the above-mentioned embodiments that the method for measuring the delay characteristic in the data communication network of the present invention uses the microcode/hardware of the network device to perform timestamp marking on the measurement message generated by the upper layer software, thereby calculating the time delay of the network. extension characteristics. The method is simple to implement, does not affect actual operation services, and can well obtain the measurement accuracy required by the network.

Claims (7)

1, the method for measurement of delay character in a kind of data communication network is characterized in that this method may further comprise the steps:

A, transmitting terminal generate at least one measured message when sending packet to receiving terminal, be the measured message that generated mark first transmitting time stamp one by one by self microcode/hardware, and above-mentioned measured message is sent; Receiving terminal is isolated measured message from the packet that receives, and with self microcode/hardware be isolated measured message one by one mark first time of reception stab;

B, receiving terminal are isolated measured message mark second transmitting time stamp one by one with the microcode/hardware of self, and by data communication network above-mentioned measured message are sent to transmitting terminal; Transmitting terminal is isolated described measured message from the packet that receives, and with self microcode/hardware be isolated measured message one by one mark second time of reception stab;

It is the measured message of ring certainly of self that c1, transmitting terminal and/or receiving terminal generate destination address, by microcode/hardware be described oneself encircle the measured message mark and send timestamp from environment-development, and will encircle measured message certainly and send;

C2, transmitting terminal and/or receiving terminal are isolated described from encircling measured message from the packet that receives, and with microcode/hardware be isolated oneself encircle the measured message mark and receive timestamp from articulating;

C3, receiving timestamp and sending timestamp to calculate extra time delay according to transmitting terminal and/or receiving terminal from environment-development from articulating;

C, first transmitting time stamp according to measured message institute mark, first time of reception stamp, second transmitting time stamp and second time of reception stab, and calculate the time delay in the delay character between transmitting terminal and the receiving terminal; Described time delay is deducted extra time delay, the time delay behind the acquisition compensation correction.

2, method according to claim 1 is characterized in that, the measured message that generates among the step a has the measured message sign;

Then the separating and measuring method of message is: according to the measured message sign, use the traffic classification matching process to isolate measured message.

3, method according to claim 2 is characterized in that, described traffic classification matching process adopts three-state content addressing memory to realize.

4, method according to claim 1, it is characterized in that, described delay character comprises the time difference, then step c also comprises the method for the time difference in the calculation delay characteristic, for: first time of reception stamp of each measured message is added second transmitting time stamp sum, deduct first transmitting time stamp and add second time of reception stamp sum, the difference of acquisition is divided by two; Again the result of calculation of all measured message is averaged, obtain the time difference of transmitting terminal and receiving terminal.

5, method according to claim 1, it is characterized in that, the method of the time delay in the delay character between described calculating transmitting terminal of described step c and the receiving terminal is: first time of reception stamp of each measured message is added second time of reception stamp sum, deduct first transmitting time stamp and add second transmitting time stamp sum, the difference that obtains obtains the time delay of each measured message divided by two; Again the time delay of all measured message is averaged, obtain the time delay of transmitting terminal and receiving terminal.

6, method according to claim 5 is characterized in that, the method for the extra time delay of the described calculating of step c3 is:

Send timestamp with deducting from environment-development of transmitting terminal, obtain extra time delay from articulating receipts timestamp;

Perhaps, send timestamp, obtain extra time delay deducting from environment-development of receiving terminal from articulating receipts timestamp;

Perhaps, send first extra time delay of timestamp acquisition and deducting from environment-development of receiving terminal to send the second extra time delay of timestamp acquisition to average deducting from environment-development of transmitting terminal, obtain extra time delay from articulating receipts timestamp from articulating receipts timestamp.

7, method according to claim 5 is characterized in that, described delay character also comprises delay variation, and transmitting terminal generates two or more measured message among the step a in a period of time;

Then step c also comprises the method for the delay variation in the calculation delay characteristic, for: maximum delay and minimal time delay are selected in the time Yanzhong of all measured message that generate in the period from this section, and maximum delay deducted minimal time delay, obtain the delay variation of transmitting terminal and receiving terminal.

Images