CN107426111A - Hardware achievable jamming control method and device based on wred algorithm - Google Patents

Abstract

The present invention provides a hardware-realizable congestion control method and device based on the WRED algorithm. The device includes a parsing and classification module for parsing and classifying network data packets; a WRED module for performing input network data packets Random packet loss processing; the queue module is used to cache the network data packets; the dequeue scheduling module is used to reasonably schedule the network data packets in the queue to go out. Compared with the traditional congestion control method, the present invention not only guarantees the fairness of network data packet transmission, but also ensures the priority transmission of higher priority network data packets.

Description

Hardware Realizable Congestion Control Method and Device Based on WRED Algorithm

technical field

The invention relates to the field of efficient transmission of network data, in particular to a network congestion control technology.

Background technique

In recent years, the development of network technology is changing with each passing day, and the scale of the network is also expanding rapidly. Especially after entering the 1990s, the Internet based on the Internet Protocol (Internet Protocol, IP) between networks has shown explosive growth. Gradually develop into a global information infrastructure. The traffic in the Internet is growing rapidly. In addition to the traditional WWW, FTP and other data streams, there are also a large number of real-time multimedia data streams. Since different types of data streams in the network are converged in the network routers, the routing nodes of the network It caused a great burden, and the problem of network congestion was gradually exposed.

In network communication, network congestion leads to the decline of network quality of service (QoS) performance indicators such as delay and throughput, which is a key factor affecting the utilization of network resources such as network bandwidth and cache. The performance of the network is of great significance. Insufficient network node storage space, insufficient link bandwidth capacity, and weak network node processing capabilities are the direct causes of network congestion. to solve the problem of network congestion. On the contrary, some cases may further exacerbate network congestion. Network congestion has become a bottleneck restricting network development and application. How to better avoid and control network congestion has become a hot issue in network research.

The main goal of network congestion control is to control the data flow entering the network, to ensure that the communication subnet will not be flooded by the data flow sent by users, and to use network resources reasonably. There are two ways to solve network congestion. One is congestion avoidance, that is, to avoid network congestion as much as possible to keep the network running in a good state; the other is to take remedial measures after network congestion occurs to restore the network to a good state as soon as possible.

In the Internet, the occurrence of network congestion is an inherent property. Therefore, it is impractical to completely avoid network congestion, and it is more reasonable to use a combination of congestion avoidance and congestion management to solve network congestion.

Contents of the invention

Aiming at the above technical problems, this design provides a hardware-realizable congestion control method and device based on the WRED algorithm, which can effectively control network congestion.

A kind of congestion control device based on WRED algorithm, is characterized in that, comprises:

An analysis and classification module, which is connected to the WRED module, is used to analyze the input network data packet, and transmits the network data packet to the corresponding WRED module according to the IP priority field analyzed;

The WRED module is connected to the analysis and classification module and the queue module, and is used to perform random packet loss processing on the input network packets, and then transmits to the queue module;

A queue module, which is connected to the WRED module and the dequeuing scheduling module, is used to cache the input network packets;

The dequeue scheduling module is connected to the queue module and is used for dequeue scheduling processing of the network data packets in the queue.

The congestion control method based on the WRED algorithm that the hardware of the present invention can realize is characterized in that comprising the steps:

The analysis and classification module analyzes the input network data packet, and transmits the network data packet to the corresponding WRED module according to the IP priority field analyzed;

The WRED module performs random packet loss processing on the input network data packets, and then transmits them to the queue module;

The queue module caches the input network packets;

The dequeue scheduling module performs dequeue scheduling processing on the network data packets in the queue. The parsing and classification module parses the packet header of the input network data packet, and stores the IP priority field and the network data packet length field therein. Then, according to these two fields, a complete network data packet is transmitted to the corresponding WRED module.

The WRED module performs random packet loss processing on the input network data packets. The probability of packet loss is determined by the queue length fed back by the queue module and the preset queue length upper water line, queue length lower water line, queue length upper node, and queue length lower node It is decided that the probability of packet loss is 0, 1/8, 1/4, 1/2, 1 respectively.

The queue module caches the input network data packets, and monitors the length of the queue in real time, and feeds back the monitored queue length value to the WRED module for the calculation of the packet loss probability of the WRED module.

The dequeuing scheduling module is used to output the network data packets in the queue according to a certain order and weight. Higher-priority network data packets are given higher weights, and lower-priority network data packets are given lower weights, which reflects the characteristic that network data packets with higher priority are processed first . On the other hand, the present invention uses a two-layer arbitration algorithm based on round robin and weight to ensure the fairness of network data packet transmission.

The present invention has the following advantages: the present invention uses the WRED algorithm and the double-layer arbitration algorithm based on polling and weight, due to the characteristic of the WRED algorithm for the priority processing of IP data packets with higher priority, and the double-layer arbitration algorithm based on polling and weight Fairness, so that the present invention not only ensures the fairness of network data packets, but also ensures that network data packets with higher priority are processed first, and can effectively control network congestion.

Description of drawings

Further description will be given below in conjunction with the accompanying drawings and specific embodiments.

Fig. 1 is a flow chart of IP packet processing in the present invention.

Fig. 2 is a structural diagram of the analysis classification module of the present invention

Fig. 3 is the structural diagram of the WRED module of the present invention

Fig. 4 is a structural diagram of the queue module of the present invention

Fig. 5 is the structural diagram of the present invention's dequeuing and reading

Figure 6 shows the IP packet format.

FIG. 7 is a flow chart of network packet loss processing.

Fig. 8 is a circuit structure diagram of a random number generator with a packet loss probability of 1/2.

FIG. 9 is a circuit structure diagram of a random number generator with a packet loss probability of 1/4.

Fig. 10 is a circuit structure diagram of a random number generator with a packet loss probability of 1/8.

detailed description

As shown in FIG. 1, it is a flow chart of IP packet processing in the present invention. The IP data packet first enters the analysis and classification module, and the analysis and classification module analyzes the incoming network data packet, and divides all incoming network data packets into 8 categories according to the IP priority field in the network data packet header, and then classifies each network data The packet is sent to the corresponding WRED module; the function of the WRED module is to randomly discard the incoming network data packets. This module consists of 8 WRED sub-modules from WRED0 to WRED7, and each WRED sub-module corresponds to a certain priority. The difference between the 8 WRED sub-modules is that the configuration parameters of each WRED sub-module are different, that is, the upper and lower queues of each WRED sub-module, and the upper and lower nodes of the queue are different; the role of the queue module is to The network data packets processed by the WRED module are cached, and real-time queue length information is provided for the WRED module, and the queue module is also composed of 8 queue sub-modules, each queue sub-module corresponds to a WRED sub-module, and the queue length information The acquisition is obtained by counting the counter. If the queue only inputs data but does not output data, the value of the queue length counter increases by 1; if the queue only outputs data but does not input data, the value of the queue length counter decreases by 1; if the queue both inputs data and outputs data, Then the value of the queue counter remains unchanged; the effect of the dequeuing scheduling module is to output the network data packets in the queue module according to a certain scheduling algorithm, and the data scheduling algorithm used in the present invention is a double-layer arbitration algorithm based on polling and weight.

As shown in FIG. 2 , it is a structural diagram of the analysis and classification module of the present invention. This module is composed of FIFO and analysis classification module controller. The FIFO is used to cache incoming network data packets, and the classification and analysis module controller is responsible for analyzing and classifying the network data packets.

As shown in FIG. 3 , it is a structural diagram of the WRED module of the present invention. This module is composed of FIFO, WRED module controller and random number generator. The FIFO is used to cache incoming network data packets, the random number generator is used to generate random numbers, and the WRED module controller is responsible for random packet loss processing of incoming network data packets.

As shown in FIG. 4 , it is a structural diagram of the queue module of the present invention. This module is composed of FIFO and queue module controller. The FIFO is used to cache incoming network data packets, and the queue module controller is responsible for counting the length of the data in the FIFO and feeding the information back to the WRED module.

As shown in FIG. 5 , it is a structural diagram of the dequeue scheduling module of the present invention. This module is composed of FIFO and dispatching module controller. The FIFO is used to cache incoming network data packets, and the dequeue scheduling controller is responsible for reasonable scheduling of incoming network data packets.

As shown in Figure 6, it is an IP packet format. The version refers to the currently used IP version number, a total of 4 digits, 0100 means IPV4, and 0110 means IPV6; the length of the IP packet protocol header indicates the number of 32-bit word length in the protocol header; DiffServ refers to the upper layer protocol for processing the current The expected quality of service of the IP data packet; the total length refers to the byte length of the entire IP data packet; the identifier is used to identify the current IP data packet; the flag has 3 bits in total, the lowest bit controls fragmentation, and the next fragment is set to 1, otherwise Set to 0, the middle bit indicates whether the IP data packet can be fragmented, and the third bit is not applicable; the fragment offset refers to the data position related to the beginning of the source data packet; the survival time is a kind of counter, when the IP data is discarded The value of each point of the packet is reduced by 1 in turn until it is reduced to 0; the protocol refers to which upper layer protocol receives the imported data packet after the IP processing process is completed; the header checksum helps ensure the integrity of the IP protocol header; the source address Refers to the IP address of the source host; the destination address refers to the IP address of the destination host; optional fields and padding are used to support troubleshooting, measurement and security measures; the data part is the data part of the IP packet. When the network data packet is input to the classification and analysis module, a complete IP data packet is transmitted to the corresponding The WRED module. The specific processing flow is to first register the IP priority field and the total length field of the IP data packet. The IP priority field has a total of 3 bits, and the decimal number that can be represented is from 0 to 7. When the value of the IP priority of an IP data packet When it is 0, the IP data packet is transmitted to the WRED0 module, when the IP priority value of an IP data packet is 1, the IP data packet is transmitted to the WRED1 module, and so on, when an IP data packet IP When the priority value is 7, the IP data packet is transmitted to the WRED7 module. Whenever an IP data packet is transmitted, a complete IP data packet is transmitted according to the total length field.

As shown in Figure 7, when a network data packet is transmitted to the WRED module, the queue length fed back by the queue module is obtained, and then the queue length is compared with the preset queue upper and lower watermarks, and the queue upper and lower nodes to obtain The probability of packet loss. If the queue length is greater than or equal to the queue upper waterline, the packet loss probability is 1; if the queue length is less than or equal to the queue lower waterline, the packet loss probability is 0; if the queue length is between the queue upper waterline and the queue lower waterline, The packet loss probability is between 0 and 1, divided into 3 situations. If the queue length is less than the queue upper watermark but greater than the queue upper node, the packet loss probability is 1/2; if the queue length is less than or equal to the queue upper node, but greater than or equal to the queue lower node, the packet loss probability is 1/4 ; If the queue length is less than the lower node of the queue but greater than the lower line of the queue, the packet loss probability is 1/8. Packet loss processing of IP data packets, that is, when the IP data packets are transmitted from the WRED module to the queue module, while reading the IP data packets, the write signal of the queue module is pulled down, that is, the packet loss processing is realized.

As shown in Figure 8, the 1/2 probability random number generator circuit consists of 5 registers and 1 XOR logic gate. When resetting, each register is assigned a value, and then the value output by each clock can be regarded as random, and the probability of outputting 0 and 1 is the same.

As shown in Figure 9, the 1/4 probability random number generator is composed of two random number generators and an AND gate. The initial values assigned by the two random number generators are different, so it can be considered that their output values are irrelevant. After connecting an AND gate, the probability of outputting 1 is 1/4.

As shown in Figure 10, the 1/8 probability random number generator is composed of 3 random number generators and 3 AND gates. The initial values assigned by the three random number generators are different, so it can be considered that their output values are irrelevant. After connecting three AND gates, the probability of outputting 1 is 1/8.

When the IP data packet passes through the WRED module, if it is processed by packet loss, it will not be transmitted to the queue module; if it is not processed by packet loss, it will be transmitted to the queue module. When an IP data packet enters the queue module, the queue module acts as a buffer for network data packets, and at the same time monitors the length of the data in the queue in real time and feeds it back to the WRED module.

When there is a queue that is not empty, the dequeue scheduling module will serve different queues in a certain order and weight. The queue with the higher priority will transmit data first and the assigned transmission data weight will be greater. Moreover, the present invention uses a two-layer arbitration algorithm based on round robin and weight, so the fairness of network data packet transmission is guaranteed.

The above are the preferred embodiments of the present invention, and all changes made according to the technical solution of the present invention, when the functional effect produced does not exceed the scope of the technical solution of the present invention, all belong to the protection scope of the present invention.

Claims (10)

1. A congestion control device based on WRED algorithm, characterized in that, comprising: An analysis and classification module, which is connected to the WRED module, is used to analyze the input network data packet, and transmits the network data packet to the corresponding WRED module according to the IP priority field analyzed; The WRED module is connected to the analysis and classification module and the queue module, and is used to perform random packet loss processing on the input network packets, and then transmits to the queue module; A queue module, which is connected to the WRED module and the dequeuing scheduling module, is used to cache the input network packets; The dequeue scheduling module is connected to the queue module and is used for dequeue scheduling processing of the network data packets in the queue. 2. a kind of congestion control device based on WRED algorithm according to claim 1, it is characterized in that, described parsing and classification module analyzes the packet header of the network data packet of input, deposits IP priority field and network data packet wherein Length field; then according to these two fields, a complete network data packet is transmitted to the corresponding WRED module. 3. A kind of congestion control device based on WRED algorithm according to claim 1, it is characterized in that, described WRED module carries out random packet loss process to the network data packet of input, and the probability of packet loss is the queue fed back by queue module Length and the pre-set queue length upper waterline, queue length lower waterline, queue length upper node, queue length lower node, the probability of packet loss is 0, 1/8, 1/4, 1/2, 1 respectively. 4. a kind of congestion control device based on WRED algorithm according to claim 1, is characterized in that, described queue module is carried out cache processing to the network packet of input, and the length of queue is carried out real-time monitoring, will monitor the The queue length value is fed back to the WRED module for the calculation of the packet loss probability of the WRED module. 5. A congestion control device based on a WRED algorithm according to claim 1, wherein the dequeue scheduling module outputs the network data packets in the queue according to a certain order and weight. 6. a hardware-realizable congestion control method based on the WRED algorithm, characterized in that it comprises the steps: The analysis and classification module analyzes the input network data packet, and transmits the network data packet to the corresponding WRED module according to the IP priority field analyzed; The WRED module performs random packet loss processing on the input network data packets, and then transmits them to the queue module; The queue module caches the input network packets; The dequeue scheduling module performs dequeue scheduling processing on the network data packets in the queue. 7. the hardware-realizable congestion control method based on the WRED algorithm according to claim 6 is characterized in that the parsing and classification module parses the packet header of the input network packet, registers the IP priority field and the network packet wherein Length field; then according to these two fields, a complete network data packet is transmitted to the corresponding WRED module. 8. The hardware-realizable congestion control method based on the WRED algorithm according to claim 6, wherein the WRED module performs random packet loss processing on input network packets, and the probability of packet loss is the queue fed back by the queue module Length and the pre-set queue length upper waterline, queue length lower waterline, queue length upper node, queue length lower node, the probability of packet loss is 0, 1/8, 1/4, 1/2, 1 respectively. 9. the hardware-realizable congestion control method based on the WRED algorithm according to claim 6, characterized in that, the queue module buffers the input network packets, and monitors the length of the queue in real time, and monitors the detected The queue length value is fed back to the WRED module for the calculation of the packet loss probability of the WRED module. 10. The hardware-implementable congestion control method based on the WRED algorithm according to claim 6, characterized in that, the dequeue scheduling module outputs the network data packets in the queue according to a certain order and weight.