CN1297097C - Method for raising data transmission performance when the network is congested - Google Patents

Abstract

The invention relates to a method for improving data transmission performance when the network is congested. This method is to fragment the large data packet at the link layer when the low-speed link is congested, reduce the length of the data packet, and transfer it to the fragmentation temporary queue, and transfer the high-priority data to the high-priority Queues, so that high-priority data such as real-time business data are sent across fragments, avoiding the delay and jitter caused by waiting for large data packets to be sent in real-time business data, ensuring the quality of service for real-time business data, and minimizing It eliminates the out-of-order of fragmented packets, and at the same time ensures the priority scheduling of high-priority data such as real-time business data.

Description

A Method of Improving Data Transmission Performance in Network Congestion

technical field

The invention relates to the technical field of network communication, in particular to a method for improving data transmission performance when the network is congested.

Background technique

With the development of network communication technology, the requirements for data transmission performance in network communication are also increasing. In order to ensure the performance of data transmission in network communication, that is, ensure the reliable transmission of data to meet the communication needs of users, congestion management technologies such as PQ (Priority Queue) are usually used to manage the data transmission process. When the output interface is congested, The data flow can be classified according to the importance of the transmitted data, the level of service quality to be provided, etc., usually according to the business attributes carried by the data message to determine the priority of sending the data; and then send the classified data to different priority This method guarantees to a certain extent the reliable transmission of high-priority data services that require a higher quality of service level in the network. For example, for network communications that require a higher quality of service level The real-time business data can be sent to the high-priority queue for priority transmission, so that its service quality can be guaranteed first.

Data congestion management technology solves the problem of reliable transmission of high-priority data when network communication is congested to a certain extent. However, due to the different sizes of data packets transmitted in the network, the time required for larger data packets in the transmission process is relatively long. Small data packets are longer, and the difference in transmission time is negligible for high-speed links, but for low-speed links, the above-mentioned time difference may lead to greatly reduced data transmission performance, which cannot meet the communication needs of users . For example, for voice service data, it is generally required that the end-to-end delay should not exceed 150ms, and the single-hop delay should not exceed 20ms. The transmission delay is usually the data packet size (frame length) divided by the link bandwidth. The transmission delay of data packets is shown in Figure 1. For low-speed links below 768Kbps, the delay caused by large data packets far exceeds the end-to-end delay value usually required by voice service data. That is to say, the real-time service data packets that arrive after the large data packets enter the high-priority queue (that is, the High queue) and must wait for the large data packets to be sent before they can be dispatched. As shown in Figure 2, the real-time service data Packet 2 arrives at the moment when the big data packet 1 is scheduled and processed, but the real-time service data packet 2 must wait until the big data packet 1 is sent before it can be scheduled. Assuming that the length of the big data packet 1 is 1500 bytes, the output link bandwidth is 256Kbps, the delay of real-time service data packet 2 will reach 46ms.

In order to solve the above problems, the commonly thought method is to fragment large data packets to reduce the maximum transmission unit. Fragmentation of data packets can be performed at the network layer or link layer. If the packet is divided at the network layer Fragmentation, and data reorganization after fragmentation at the network layer of the destination end will increase the consumption of network bandwidth resources and will certainly aggravate network congestion. Therefore, fragmentation of large data packets should be performed at the link layer. Existing MP (PPP multi-link protocol) fragmentation technology is exactly a kind of link layer fragmentation technology, this technology divides PPP (point-to-point protocol) frame into several slices according to certain fragmentation strategy, in each fragmentation Encapsulate the MP header on the chip, mark the start (B), end (E), etc., and number it, then distribute all the fragments to each channel for transmission, and reassemble the fragments according to the sequence number at the receiving end. If a fragment is lost, the entire packet is discarded because it cannot be reassembled.

Therefore, if the MP fragmentation technology is used, it is necessary to ensure that the fragmented messages are sent and received in sequence. Usually, for the situation where queuing mechanisms such as PQ, CQ (Customized Queue), and WFQ (Weighted Fair Queuing) are not applied, the transmission order of fragmented packets can be guaranteed; If the existing MP fragmentation technology is still used, the transmission order of the fragmented packets may be severely disturbed, and the message cannot be reassembled at the receiving end, so that the reliable transmission of data cannot be guaranteed.

Contents of the invention

In view of the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a method for improving data transmission performance when the network is congested, so as to solve the problem that the data transmission service quality cannot be satisfied due to the transmission time delay generated by the large data packet in the link. asked questions.

The purpose of the present invention is achieved in that the method for improving data transmission performance when the network is congested includes:

a. Classify the data to be sent;

b. Transfer the data with high sending priority to the high priority queue, and transfer other data to the fragmented temporary queue and lower priority queue;

c. When performing data sending processing, the data in the high-priority queue and the fragmented temporary queue are sent first, and then the data in the lower-priority queue is sent.

The step a includes: classifying the data according to the QoS (Quality of Service) value.

The step a includes: classifying the data according to the five-tuple information of IP (Internet Protocol).

The step a includes: classifying the data according to the input interface of the data packet.

Described step b comprises: carry out fragmentation according to need from the large data packet dispatched from the lower priority queue, and after encapsulating the sequence number in the fragmentation, transfer to the fragmentation temporary queue; whether the large data packet Fragmentation and the size of the fragmentation are determined according to the allowable delay and data transmission bandwidth of the high-priority data service, and the fragmentation size is equal to the allowable delay multiplied by the data transmission bandwidth.

The data with high sending priority is: according to the business attribute carried by the data message, it is determined that the data needs to be sent preferentially when the network is congested;

The above-mentioned large data packet is: the delay caused by the size of the high-priority data in the sending process exceeds the allowable delay of this type of data.

The lower priority queue may further include one or more data sending queues of different priorities.

When using a single-channel link (i.e. a low-speed link) for data transmission, the step c further includes:

c1. Schedule the data in the high priority queue and send it out;

c2. When the data in the high-priority queue has been sent, schedule the data in the temporary queue for fragmentation and send it out;

c3. When the data in the high-priority queue and the fragmented temporary queue are all sent, the data in the lower-priority queue is scheduled and sent.

When using a multi-lane link for data transmission:

The step b also includes: adding each data to each queue after encapsulating the serial number;

Described step c further comprises:

c4. Scheduling the data in the temporary queue of fragments and sending them out;

c5. When the data in the fragmented temporary queue has been sent, schedule the data in the high-priority queue and send it out;

c6. When the data in the fragmented temporary queue and the high-priority queue are all sent, schedule the data in the lower-priority queue and send it;

Data reception and fragmented data reassembly are performed at the receiving end according to the sequence numbers of each data.

As can be seen from the above technical solution, the present invention fragments the large data packet at the link layer when the low-speed link is congested, reduces the length of the data packet, and enables high-priority data such as real-time business data to be crossed in the split It can avoid the delay and jitter caused by waiting for large data packets to be sent in real-time business data, and ensure the quality of service of real-time business data. The present invention can also minimize the out-of-sequence of fragmented messages in the data transmission process when applying the QoS queuing mechanism on the link bound with multiple channels, while at the same time ensuring the priority of high-priority data such as real-time business data scheduling. The implementation of the present invention is also relatively simple. It only needs to add a fragmented temporary queue on the basis of the original queuing strategy, and there is no other system overhead. In addition, the present invention has a wide application range and can be applied to various links including PPP, HDLC (High-level Data Link Control), FR (Frame Relay) and the like.

Description of drawings

Figure 1 is a transmission delay table of frames of different lengths in links of different bandwidths;

FIG. 2 is a schematic diagram of data packet transmission in the prior art;

Fig. 3 is the schematic diagram that adopts single channel to carry out data packet transmission in the present invention;

Fig. 4 is the schematic diagram that adopts multi-channel to carry out data packet transmission among the present invention;

Fig. 5 is a flowchart of a specific embodiment of the present invention.

Detailed ways

The core idea of the method for improving data transmission performance when the network is congested in the present invention is that when the link layer fails to send a data packet due to network congestion, the data packet is sent according to certain rules, such as IP quintuple information or QoS (Quality of Service) value, etc., enter the corresponding queue, wherein the five-tuple information of IP includes: source and destination address, source and destination port number, protocol number, and the corresponding queue includes: high priority queue, fragmentation temporary queue and lower priority queue. The real-time business enters the high-priority queue, and the large data packets to be sent are fragmented and put into the fragmented temporary queue, and other business data are put into the lower-priority queue. The real-time data packets are cross-scheduled into the fragments of the large data packets, so as to ensure that the data with high priority in the data that fails to be sent can be sent out first, and it is guaranteed that it will not cause a long delay due to the sending of other large data packets. delay.

The specific implementation of the present invention is shown in Figure 3, Figure 4 and Figure 5, and the specific implementation process is described as follows, referring to Figure 5:

Step 51: classify the data packets that are not sent out due to network congestion according to the corresponding sending priority;

Specifically, the classification may be performed according to the QoS (Quality of Service) value of each data packet, or may be classified according to the five-tuple information of the IP data packet.

Step 52: Transfer the data with high sending priority to the high priority queue, for example, put the real-time service data with higher sending priority into the high priority queue, so as to facilitate the priority sending in the next step.

Step 53: Transfer other data into the fragmented temporary queue and lower priority queue, the specific process is:

Step 531: transfer other data to a lower priority queue;

The lower priority queue can be further divided into multiple queues with different priorities according to the data service quality level, and when the data in the high priority queue and the fragmented temporary queue have been sent, they will be compared in order of priority. Scheduling of low priority queues, for example, can be divided into normal data sending queues (i.e. Normal queues) and low priority data sending queues (i.e. Low queues) as shown in Figure 3 and Figure 4, and of course can also be divided into more different Priority data transmission queue, or only one data transmission queue as a lower priority queue;

Step 532: When the data in the high-priority queue is sent and processed, dispatch the data in the lower-priority queue;

Step 533: Fragmenting the scheduled large data packet, and encapsulating the serial number in the fragment into the temporary queue of fragmentation, that is, encapsulating the tag header in the fragment, the tag header contains information determined according to the order of fragmentation shard sequence number;

The size of fragments can be controlled according to the guaranteed delay and data transmission bandwidth of high-priority data in network communication, specifically: the size of each fragment (byte) is equal to the guaranteed delay (ms) multiplied by the data transmission bandwidth (Kbps) and then divided by 8, the data transmission bandwidth is the output interface bandwidth of the data to be sent;

A large data packet means that the delay caused by the size of the high-priority data in the transmission process exceeds the allowable delay of this type of data. The fragmentation of the large data packet is to avoid the high-priority data that arrives immediately after it. The packet does not need to wait for the large data packet to be sent before sending, but only needs to wait for a small fragment message to be sent, and the remaining fragments are temporarily cached in the fragment temporary queue, thus ensuring real-time business data, etc. The delay of high-priority data meets the communication requirements;

The purpose of fragmenting the large data packets dispatched from the queue and then entering the fragmentation temporary queue is to facilitate the reception and processing of the data by the receiving end. If the classified large data packets are directly fragmented and put into corresponding queues, because of the cross-scheduling of different priority queues when leaving the queue, the fragmented messages of different data packets in different queues will be cross-mixed and sent out. It will cause serious out-of-sequence of the fragmented packets, increase the load of the receiving end to reassemble the packets, and even fail to reassemble the received fragmented packets.

Step 54: send the data in the high-priority queue and the fragmented temporary queue;

The data in the high-priority queue and the fragmented temporary queue either has a higher sending priority, or has been scheduled from a lower-priority queue to be sent, so it needs to be sent first;

The data scheduling and sending process specifically includes the following two situations:

One is to use a single-channel link for data transmission. In this case, the data in the high-priority queue can be sent first. When the data in the high-priority queue is sent and processed, the fragmented data in the fragmented temporary queue is sent. ; That is, when sending processing, always first check whether there is a data packet in the high-priority queue, if there is, it will be dispatched out of the queue, without any mark, and sent directly; when the high-priority queue is empty , to schedule other queues; the priority of fragmented temporary queues is second only to high-priority queues, but higher than other queues;

When high-priority data such as real-time service data arrives, it will be scheduled first, and fragmented packets will be scheduled prior to other data except high-priority data, thus ensuring high-priority data It is sent with the highest priority. Even if it arrives after a large data packet, it only needs to wait for a fragmented packet to be sent. Therefore, the service quality indicators such as delay and jitter of high-priority data are well guaranteed. That is to improve data transmission performance when the network is congested; at the receiving end of the link, for real-time business data and other high-priority transmission data, because there is no encapsulation marker header, it can be immediately handed over to the network layer for forwarding processing; for fragmented packets, then Put it into the cache, wait until all the fragments of the data packet are received, reassemble them according to the sequence number, and then hand them over to the network layer for processing, so that although real-time business data packets and other high-priority data are cross-transmitted in fragmented packets, Still will not cause out-of-order fragmented packets;

The specific data sending process can be seen in Figure 3. When the data packet 1 is dispatched out of the queue, it is fragmented and divided into fragmented messages 1.1, 1.2, 1.3, and 1.4. Assuming that the output link bandwidth is 256Kbps, a 10ms delay must be guaranteed. Delay, the size of each piece of the message is about 320 bytes; send the fragmented message 1.1 first, and the other fragments are cached in the fragmented temporary queue; if the fragmented message 1.1 is sent, there are When the real-time service data packet 2 arrives, the real-time service data packet 2 is scheduled to be sent; after the sending is completed, if there is no data packet in the high-priority queue, the second fragmented message 1.2 is scheduled to be sent, and so on;

The second case is to use a bundled multi-channel link to send data. The bundled multi-channel is to use multiple channels as an interface to increase the transmission bandwidth, because the data stream is blocked when sending data through a multi-channel link. Shared by multiple channels, the same data stream may be sent from different channels, and each channel may cause out-of-sequence data packets due to different bandwidth and congestion levels. Therefore, when sending data through a multi-channel link, it is necessary to determine the sequence number for the data packet , to ensure that the receiver sends the data packets to the network layer in the correct order; otherwise, the receiver may not be able to reassemble the data packets correctly;

The present invention regards a multi-channel link as a logical link, and designs a general queuing strategy for the logical link, rather than designing a queuing strategy for each channel separately. When a certain channel is congested, the data packets sent by it will be sent In order to ensure the load balance of each channel, multi-channel links usually divide large data packets into several fragments according to a certain fragmentation strategy, and send them separately by each channel; and Still adopting the method of fragmentation and temporary storage in the fragmentation temporary queue when leaving the team, the fragmentation of the data packet is cached in a temporary queue; in the case of a multi-channel link, all data packets, including real-time business data packets , all need to mark the serial number; and set the priority of the fragmented temporary queue to be the highest to ensure that each fragment of a data packet is scheduled in order to avoid disorder of fragmented messages. But different from the low-speed link solution, the fragmented temporary queue has the highest priority.

As mentioned above, high-priority data may need to wait for all fragments of a large data packet to be dispatched before being scheduled. However, for multi-channel links, the bandwidth is usually relatively large, and the delay and Jitter usually does not exceed the maximum latency allowed for high transmit priority data.

The specific process of sending data through a multi-channel link is shown in Figure 4. When the link is congested, the data packet enters the corresponding priority queue according to the queuing strategy. When the large data packet 1 is dispatched out of the queue, the multi-channel link In order to balance the load, the data packet is divided into several fragmented messages 1.1, 1.2, 1.3, and 1.4 in the channel. ) to send; other fragmented messages 1.2, 1.3, and 1.4 enter the fragmented temporary queue and wait for the next queue scheduling. Since the fragmented temporary queue has the highest priority, other queues will not be scheduled until all fragments are continuously scheduled. In the data packets, all data packets are encapsulated with tag headers and numbered, so that the load of the physical channel is approximated, which can minimize the possibility of data packet disorder; and at the same time, the multi-channel chain Route QoS queue scheduling.

Step 55: send the data in the lower priority queue;

The lower priority queues include normal data sending queues and low priority data sending queues. During data sending processing, the data in the normal data sending queues takes precedence over the low priority data sending queues;

Send processing of data in lower priority queues includes:

For normal-sized data packets to be sent directly, a normal-sized data packet means that the transmission delay of the data packet does not exceed the maximum delay allowed by network communication;

For large data packets, refer to step 533 for processing.

Claims (10)

1, a kind of method of data transmission performance when improving network congestion is characterized in that comprising:

A, data to be sent are classified;

B, will have the high data that send priority and change in the high-priority queue, other data change in interim formation of burst and the lower priority formation;

C, when carrying out data sending processing, at first the data in high-priority queue and the interim formation of burst are sent processing, then the data in the lower priority formation are sent processing.

2, the method for data transmission performance during raising network congestion according to claim 1 is characterized in that described step a comprises: data are classified according to the service quality QoS value.

3, the method for data transmission performance during raising network congestion according to claim 1, it is characterized in that described step a comprises: the five-tuple information according to Internet protocol IP is classified to data.

4, the method for data transmission performance during raising network congestion according to claim 1, it is characterized in that described step a comprises: the input interface according to packet is classified to data.

5, the method for data transmission performance during raising network congestion according to claim 1, it is characterized in that described rapid b comprises: the big packet that will dispatch out from the lower priority formation carries out burst as required, and changes in the interim formation of burst after the encapsulation sequence number in described burst.

6, the method for data transmission performance during raising network congestion according to claim 5, it is characterized in that: whether big packet needs the size of burst and burst is to determine that according to professional time delay that allows of high-priority data and data transfer bandwidth the time delay that the burst size equals to allow multiply by data transfer bandwidth.

The method of data transmission performance during 7, according to claim 5 or 6 described raising network congestions is characterized in that:

The described high data that send priority are: determine the data that needs preferentially send when the network congestion according to data message loaded service attribute;

Described big packet is: the time delay that its size produces high-priority data in process of transmitting surpasses the time delay that such data allow.

The method of data transmission performance when 8, improving network congestion according to claim 1 or 5 is characterized in that: described lower priority formation may further include the data transmit queue of one or more different priorities.

9, the method for data transmission performance during raising network congestion according to claim 1 is characterized in that, when adopting the single channel link to carry out data when sending, described step c further comprises:

Data in c1, the schedules high priority formation also send;

C2, when the data in the high-priority queue send when finishing, the data in the interim formation of scheduling burst also send;

C3, when the data in high-priority queue and the interim formation of burst all send when finishing, the data in the formation of scheduling lower priority also send processing.

10, the method for data transmission performance during raising network congestion according to claim 1 is characterized in that, when adopting the multichannel link to carry out data when sending:

Described step b also comprises: add respectively in each formation after each data is all encapsulated sequence number;

Described step c further comprises:

Data in c4, the interim formation of scheduling burst also send;

C5, when the data in the interim formation of burst send when finishing, the data in the schedules high priority formation also send;

C6, when the data in interim formation of burst and the high-priority queue all send when finishing, the data in the formation of scheduling lower priority also send processing;

Carry out the reception of data and re-assemblying of fragment data at receiving terminal according to the sequence number of each data.