CN108833291B - Weighted random early detection method applied to exchange circuit - Google Patents

Abstract

The invention provides a weighted random early detection method applied to a switching circuit, which avoids the occurrence of link congestion by setting a threshold value to carry out pre-discarding in advance. The invention calculates the segment packet loss probability through the average queue length, the highest drop threshold and the lowest drop threshold, then expands the packet loss probability by 1024 times, compares the packet loss probability with a randomly generated 1024-bit pseudo random number value, and confirms whether the data packet is dropped or not, thus the data packet is dropped to become a random event, the results are different according to the packet size and the current link congestion conditions, the phenomenon of one-time mass packet loss after congestion is avoided, and the utilization rate of the link can be effectively maintained.

Description

Weighted random early detection method applied to exchange circuit

Technical Field

The invention relates to the technical field of digital communication, in particular to a weighted random early detection method applied to a switching circuit, which is mainly used for network nodes to carry out active queue management.

Background

Queue management is a research hotspot of end-to-end congestion control, and can improve congestion control and network quality of service (QoS) by optimizing a queue management algorithm based on a router technology.

The weighted Random Early detection algorithm wred (weighted Random Early detection) can effectively control the queue congestion according to different service priorities at the Early stage of the queue congestion. The algorithm calculates the length of each queue through average statistics, and then based on the average statistics, the queue manager performs packet loss with a certain probability by referring to WRED parameters, so that the problem of queue congestion is solved.

Disclosure of Invention

The technical problem to be solved by the invention is to research a WRED method supporting 1024 bits, and maintain the non-congestion state of a link by detecting the upcoming congestion condition and preemptively discarding data packets. The WRED avoids the phenomenon of one-time massive packet loss after congestion by selectively losing packets in advance rather than losing packets until the buffer is full, and can effectively maintain the utilization rate of the link.

The invention is realized according to the following technical scheme:

a weighted random early detection method applied in a switch circuit, comprising the steps of:

step S1: dividing the descriptors into different drop priorities according to the pre-coloring state of the input descriptors, and providing the data packets with different colors by adopting a plurality of threshold parameters and drop probability parameters, wherein each group of parameters comprises a weight W of an average queue length_q；

Step S2: the average queue length of the buffer queue is calculated, and every 4us of sampling time is set, and the calculation formula is as follows:

avgQ_new＝(1-W_q)*avgQ_{Old age}+q*W_q

Wherein W_qAs a weight, W_q＝2^-weightQ is the current queue length during sampling measurement, avgQ is initially 0, and avgQ_New＝avgQ_{Old age}+ q value avgQ after right shift of weight bit high order without complementing 0_{Old age}Right shifting the value of the high bit of weight without complementing 0;

step S3, calculating the packet loss probability according to the following formula:

avgQ refers to the average queue length; p_bTo the probability of packet loss, max_pTo maximum discard probability, min_thFor minimum discard queue Length, max_thIs the maximum discard queue length.

Discard rate from min_th0% to max_thMax of_pIncrease by 8 uniform steps, will (max) in computation_th-min_th) Is left shifted by 3 bits to one step delta and then determined (avgQ-min)_th) If it is determined which step length of the 8 uniform step lengths the value of (A) is, P is the nth section_bThe value is max_p*n/8；

Step S4: expanding a discarding probability table according to the set discarding probability, wherein each random discarding time, the system generates a random distribution of pseudo-random numbers u, u e (1, 800), when P is_bWhen 800 > u, discarding the packet, otherwise, entering an egress descriptor queue;

considering that a pseudo-random number needs to be generated in the calculation process, a 1024-bit pseudo-random number is generated, and max is used for facilitating comparison_pAmplifying 1024 times for rounding, and performing shift operation according to 2 times during calculation;

step S5: the calculation formula of the probability P of the router to mark the arriving data packet is as follows:

P＝P_b/(1-count*P_b)

the count represents the number of packets entering the queue from the last packet loss, the final marking probability P is a function of the average queue length avgQ and the count, and P slowly increases along with the count, so that continuous packet loss is avoided, and the discarded data packets can be uniformly distributed.

Compared with the prior art, the invention has the following beneficial effects:

the method effectively finishes the process of random discarding by generating 1024-bit pseudo-random numbers and comparing the discarded probability of the data packet which is amplified by 1024 times, and the discarded data packet is random rather than fixed, so that the completeness of data transmission is not influenced to a great extent, and the fairness in the process of data packet transmission is also maintained;

the invention carries out active queue management in the network transmission link and carries out random discarding on the data frame by the WRED algorithm before the congestion is generated, thereby solving the problem of link congestion and effectively improving the transmission efficiency.

Drawings

Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:

fig. 1 is a schematic diagram of packet loss probability;

FIG. 2 is a flow chart of the WRED method structure of the present invention.

Detailed Description

The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.

The invention relates to a weighted random early detection method applied to a switching circuit, which comprises the following steps:

step S1: dividing the descriptors into different drop priorities according to the pre-coloring state of the input descriptors, and providing the data packets with different colors by adopting a plurality of threshold parameters and drop probability parameters, wherein each group of parameters comprises a weight W of an average queue length_q；

Step S2: the average queue length of the buffer queue is calculated, and every 4us of sampling time is set, and the calculation formula is as follows:

avgQ_new＝(1-W_q)*avgQ_{Old age}+q*W_q

Wherein W_qAs a weight, W_q＝2^-weightQ is the current queue length during sampling measurement, avgQ is initially 0, and avgQ_New＝avgQ_{Old age}+ q value avgQ after right shift of weight bit high order without complementing 0_{Old age}Right shifting the value of the high bit of weight without complementing 0;

step S3, calculating the packet loss probability according to the following formula as shown in fig. 1:

avgQ refers to the average queue length; p_bTo the probability of packet loss, max_pTo maximum discard probability, min_thFor minimum discard queue Length, max_thIs the maximum discard queue length;

discard rate from min_th0% to max_thMax of_pIncrease by 8 uniform steps, will (max) in computation_th-min_th) Is left shifted by 3 bits to one step delta and then determined (avgQ-min)_th) If it is determined which step length of the 8 uniform step lengths the value of (A) is, P is the nth section_bThe value is max_p*n/8；

Step S4: root of herbaceous plantExpanding a discarding probability table according to the set discarding probability, wherein the system generates a random distribution of pseudo-random numbers u, u e (1, 800) when each random packet is discarded, and when P is_bWhen 800 > u, discarding the packet, otherwise, entering an egress descriptor queue;

considering that a pseudo-random number needs to be generated in the calculation process, a 1024-bit pseudo-random number is generated, and max is used for facilitating comparison_pAmplifying 1024 times for rounding, and performing shift operation according to 2 times during calculation;

the specific discard probability table is as follows.

max_pThe amplification is 800 times, and the amplification is divided into eight sections, which are specifically shown in the table I:

table one max_pP after 800 times enlargement_b

max_pExpand 1024 times, divide into eight segments, round off, as shown in table two:

TABLE II max_pP after 1024 times enlargement_b

Step S5: the calculation formula of the probability P of the router to mark the arriving data packet is as follows:

P＝P_b/(1-count*P_b)

the count represents the number of packets entering the queue from the last packet loss, the final marking probability P is a function of the average queue length avgQ and the count, and P slowly increases along with the count, so that continuous packet loss is avoided, and the discarded data packets can be uniformly distributed.

During the design process of the invention, the requirementsGenerating 1024-bit pseudo-random numbers to sum P_bIn the algorithm for generating pseudo random numbers, a mode of progressively shifting from a low bit to a high bit is adopted. Firstly, setting a default initial value to be zero, then generating a pseudo random number, carrying out left shift on the digit, and filling the lower digit with the newly generated pseudo random number to ensure that each digit is a randomly generated numerical value.

The descriptor pre-coloring module can be nested in an active queue management module, descriptor pre-coloring is carried out through a color marking process of an enqueue descriptor in the early stage, then the descriptor is transmitted to a WRED module to be randomly discarded, different pre-colored descriptors have different discarding thresholds, and the discarding conditions of the descriptors can be reasonably distinguished. The WRED method structure flow chart of the invention is shown in FIG. 2.

The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.

Claims (1)

1. A weighted random early detection method applied in a switch circuit, comprising the steps of:

step S1: dividing the descriptors into different drop priorities according to the pre-coloring state of the input descriptors, and providing the data packets with different colors by adopting a plurality of threshold parameters and drop probability parameters, wherein each group of parameters comprises a weight W of an average queue length_q；

Step S2: the average queue length of the buffer queue is calculated, and every 4us of sampling time is set, and the calculation formula is as follows:

avgQ_new＝(1-W_q)*avgQ_{Old age}+q*W_q

W_qAs a weight, W_q＝2^-weightQ is the current queue length at the time of the sampling measurement, avgQ is initially 0,

avgQ_new＝avgQ_{Old age}+ q right shift wevalue-avgQ after 0 is not complemented for high order bits_{Old age}Right shift the value of the high bit of weight without complementing 0,

step S3, calculating the packet loss probability according to the following formula:

avgQ refers to the average queue length; p_bTo the probability of packet loss, max_pTo maximum discard probability, min_thFor minimum discard queue Length, max_thIs the maximum discard queue length;

discard rate from min_th0% to max_thMax of_pIncrease by 8 uniform steps, will (max) in computation_th-min_th) Is left shifted by 3 bits to a uniform step delta and then determined (avgQ-min)_th) If it is determined which step length of the 8 uniform step lengths the value of (A) is, P is the nth section_bThe value is max_p*n/8；

Step S4: expanding a discarding probability table according to the set discarding probability, wherein each random discarding time, the system generates a random distribution of pseudo-random numbers u, u e (1, 800), when P is_bWhen 800 > u, discarding the packet, otherwise, entering an egress descriptor queue;

considering that a pseudo-random number needs to be generated in the calculation process, a 1024-bit pseudo-random number is generated, and max is used for facilitating comparison_pAmplifying 1024 times for rounding, and performing shift operation according to 2 times during calculation;

step S5: the calculation formula of the probability P of the router to mark the arriving data packet is as follows:

P＝P_b/(1-count*P_b)

the count represents the number of packets entering the queue from the last packet loss, the final marking probability P is a function of the average queue length avgQ and the count, and P slowly increases along with the count, so that continuous packet loss is avoided, and the discarded data packets can be uniformly distributed.

Images