WO2019153931A1

WO2019153931A1 - Data transmission control method and apparatus, and network transmission device and storage medium

Info

Publication number: WO2019153931A1
Application number: PCT/CN2018/124103
Authority: WO
Inventors: 开毅; 李吉; 朱夏; 陈建; 何健飞
Original assignee: 华为技术有限公司
Priority date: 2018-02-08
Filing date: 2018-12-27
Publication date: 2019-08-15
Also published as: CN110138678B; CN110138678A

Abstract

Disclosed are a data transmission control method and apparatus, and a network transmission device and a storage medium. The data transmission control method comprises: receiving an Internet protocol (IP) data packet to be processed; determining whether a first TCP data stream to which the IP data packet to be processed belongs satisfies a pre-set packet loss screening condition; if the first TCP data stream satisfies the packet loss screening condition, dropping the IP data packet to be processed when determining, according to a pre-set packet loss strategy, that the IP data packet to be processed is to be dropped; and if the first TCP data stream does not satisfy the packet loss screening condition, storing the IP data packet to be processed in a queue cache of a network transmission device. According to the embodiments of the present invention, by determining whether a TCP data stream to which an IP data packet belongs satisfies a packet loss screening condition, a lost packet is centralized on the TCP data stream satisfying the packet loss screening condition, such that the quantity of TCP data streams affected by packet loss is reduced, and the number of users affected by the packet loss is thus reduced.

Description

Data transmission control method and device, and network transmission device and storage medium

This application claims the priority of the Chinese Patent Application filed on February 8, 2018, the Chinese Patent Office, Application No. 201101129482.9, the application of the name of "Data Transmission Control Method and Apparatus, and Network Transmission Equipment and Storage Medium", the entire contents of which are hereby incorporated by reference. This is incorporated herein by reference.

Technical field

The present invention relates to the field of communications technologies, and in particular, to a data transmission control method and apparatus, and a network transmission apparatus and a storage medium.

Background technique

The Transmission Control Protocol (TCP) is a connection-oriented, reliable, byte stream-based transport layer communication protocol. TCP traffic dominates the Internet Protocol (IP) network. TCP guarantees the reliability of transmission by acknowledgment and retransmission, and uses congestion control algorithm to control the transmission rate, making full use of network resources while avoiding network congestion.

Network traffic has obvious burstiness, that is, the amount of traffic is not uniform over time. Network transmission equipment, such as routers, switches, optical line terminals (OLTs), and Digital Subscriber Line Access Multiplexers (DSLAMs), generally use buffers and absorb them in the form of queues. A burst of traffic. When the network traffic exceeds the processing capacity of the network device, the device will cache the data packet to prevent packet loss. If the network traffic continues to be longer than the processing capacity of the device, the length of the cache queue will continue to rise until it overflows.

When a network transmission device overflows, it will not be able to receive newly arrived packets, that is, packets will be discarded. This phenomenon is called tail drop. The main hazard of tail drop is the global synchronization between TCP flows, resulting in low device utilization. This is because, when the network device is in the tail drop state, the TCP stream passing through the device will be dropped, and the TCP streams will synchronously enter the cycle of "drop packet-down-speed-retransmission-speed increase". When the synchronous speed is increased, the network device is overloaded. When the data is slowed down synchronously, the network device utilization is low.

Active Queue Management (AQM) is a widely used technology for tail drop. The basic idea is to actively manage the queue when it starts to accumulate, instead of tail-dropping when the queue is passively overflowing. However, the packet loss behavior in the active packet loss process in AQM technology is blind. It only judges whether packet loss is performed with a certain probability. The consequence of this is that there are many TCP data streams that are lost packets, and users affected by packet loss. many.

Summary of the invention

The present application provides a network data transmission control method and apparatus, and a network transmission device and a storage medium, which can reduce the number of TCP data streams affected by packet loss and improve the overall user experience.

In a first aspect, the present application provides a data transmission control method, where the method is performed by a network transmission device, the method includes: receiving an IP data packet to be processed; and determining first TCP data to which the to-be-processed IP data packet belongs Whether the flow meets the preset packet loss screening condition; if the first TCP data flow satisfies the packet loss screening condition, when the IP data packet is to be discarded according to the preset packet loss policy, the And processing the IP data packet to be discarded; if the first TCP data flow does not satisfy the packet loss screening condition, storing the to-be-processed IP data packet in a queue cache of the network transmission device.

In the present application, the TCP data stream to which the received IP data packet belongs is filtered by setting a packet loss filtering condition, and further determining whether to drop the IP data packet when the TCP data flow to which the IP data packet belongs satisfies the packet loss filtering condition. . It can be seen that, in the implementation manner of the first aspect of the present application, when the active packet loss is required, the IP data packet to be discarded is concentrated on the TCP data stream that meets the packet loss screening condition, and the number of TCP data streams affected by the packet loss is reduced. This reduces the number of users affected by packet loss and improves the overall user experience.

With reference to the first aspect, in a possible implementation manner of the first aspect, the packet loss screening condition includes: the flow state information of the TCP data stream meets a preset condition; and the determining the first TCP to which the IP data packet belongs Whether the data stream meets the preset packet loss screening condition includes: determining whether the flow state information of the first TCP data stream meets the preset condition; wherein, the flow state information of the first TCP data stream includes the A rate characterization value of the first TCP data stream, the preset condition comprising a rate characterization value greater than a set value, the rate characterization value of the first TCP data stream being used to characterize a transmission rate of the first TCP data stream.

In the present application, by setting the packet loss screening condition to the TCP data stream, the rate representation value is greater than the set value, and the TCP data stream that is lost is selected on the large stream, which reduces the packet loss triggering TCP retransmission timeout (Retransmission Timeout). , RTO), thereby reducing the occurrence of window down, and reducing the impact of packet loss on the transmission performance of the device network.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the rate characterization value of the first TCP data stream is stored in the queue cache within a preset duration The amount of data of the IP packet belonging to the first TCP data stream.

With reference to the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the data quantity of the IP data packet belonging to the first TCP data stream belongs to the first The number of bytes of the IP data packet of the TCP data stream and/or the number of IP data packets; the rate representation value is greater than the set value. The number of bytes of the IP data packet is greater than the set number of bytes and/or the number of IP data packets. The number is greater than the set number.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the flow state information of the first TCP data stream further includes that the first TCP data stream is selected last time. The preset condition further includes that the selected time difference is greater than the first set duration; wherein the first TCP data stream is selected to mean that the first TCP data stream satisfies the packet loss screening condition, the first The selected time difference of the TCP data stream is the difference between the reception time of the IP packet to be processed and the time when the first TCP data stream is last selected.

In the present application, by setting the first set duration as the packet loss protection period, the same TCP data stream is prevented from being selected multiple times in one packet protection period, thereby avoiding excessive packet loss due to the same data stream. The situation in which the retransmission packet of the data stream is discarded causes the triggering of the TCP RTO.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the flow state information of the first TCP data stream further includes that the first TCP data stream is continuously selected The preset condition further includes that the duration of the TCP data stream being continuously selected is less than the second set duration; wherein the first TCP data stream is selected to indicate that the first TCP data stream satisfies the packet loss screening condition The duration of the first TCP data stream being continuously selected, the earliest time of the receiving time of the to-be-processed IP data packet and the first TCP data stream being consecutively selected before the receiving time of the to-be-processed IP data packet The length of time between times selected.

It should be noted that, for each TCP data stream, for example, for the first TCP data stream, each time an IP data packet belonging to the first TCP data stream is received, it is necessary to determine the IP data received at this time. Whether the first TCP data stream to which the packet belongs meets the packet loss filtering condition. If the packet loss filtering condition is met, the first TCP data stream is selected. The first TCP data stream is continuously selected, indicating that the plurality of IP data packets belonging to the first TCP data stream are received before the receiving time of the IP data packet to be processed, and the TCP data to which the multiple IP data packets belong The stream (i.e., the first TCP data stream) is continuously determined to satisfy the packet loss filtering condition. In addition, if only one IP data packet belonging to the first TCP data stream is received before the receiving time of the to-be-processed IP data packet, the first TCP data stream is continuously selected, that is, the IP data is received. When the packet is used, it is determined that the first TCP data stream to which the one IP data packet belongs satisfies the packet loss filtering condition.

It can be understood that, when the first TCP data stream is continuously selected, including the IP packet belonging to the first TCP data stream that was received last time before the receiving time of the IP packet to be processed, the latest receiving The first TCP data stream to which the IP packet belongs belongs to the packet filtering condition.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the method further includes: determining, according to a first preset period, a cache queue length in the queue cache; The buffer queue length adjusts the set value, wherein the set value is negatively correlated with the buffer queue length.

In this application, the periodic adjustment of the screening condition can be implemented according to the cache queue length in the queue buffer of the network transmission device, so that the filtering condition is more consistent with the actual storage state of the cache in the device, so that the control effect of the data transmission is more in line with the actual network state. .

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the determining whether the first TCP data flow to which the to-be-processed IP data packet belongs meets a preset packet loss And the filtering condition includes: querying the flow state information table according to the unique identifier of the first TCP data flow to which the to-be-processed IP data packet belongs, and obtaining flow state information of the first TCP data flow, where the flow state information table is used for And storing, by the unique identifier of the TCP data stream, flow state information of each TCP data stream; determining, according to the flow state information of the first TCP data flow, whether the first TCP data flow satisfies the packet loss screening condition.

In this application, the flow state information table for recording the flow state information of the TCP data flow may be configured, so that when the network transmission device receives the IP data packet to be processed, the IP address to be processed can be quickly determined by means of table lookup. The flow state information of the first TCP data stream to which the data packet belongs improves the processing efficiency of the IP data packet.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, if the flow state information of the first TCP data stream includes the first TCP data stream is selected last time After determining that the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, the method further includes: using the first TCP data stream in the flow state information table The last selected time is updated to determine whether the first TCP data stream to which the to-be-processed IP data packet belongs meets the packet loss screening condition or is updated to the reception time of the to-be-processed IP data packet.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the packet loss filtering condition includes: the unique identifier of the TCP data stream is one of the preset data stream identifiers, The pre-configured data flow identifier includes a unique identifier of at least one of the TCP data streams stored in the queue cache.

In this application, in order to reduce the storage and resource occupation of the network transmission device, the data flow identifier may be directly used as a packet blanking condition, and the TCP data stream that needs to be lost is concentrated into the TCP data corresponding to the preset data stream identifier. On the flow, reduce the number of TCP data streams affected by packet loss.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the method further includes: rotating the selected pre-configured ha according to a preset time interval and a preset selection rule Setting a hash index of the number in the hash table and marking the currently selected hash index; the packet filtering condition includes one of the hash indexes of the current data tag of the TCP data stream, wherein the TCP data The hash value of the stream is a value obtained by hashing the unique identifier of the TCP data stream through a hash function corresponding to the hash table.

In the present application, by pre-configuring the hash table, when the IP data packet is received, the hash value of the TCP data stream to which the IP data packet belongs may be compared with the currently selected hash index, and the fast judgment is performed. Whether the TCP data stream to which the IP packet belongs meets the packet loss filtering condition.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the method further includes: determining, according to a second preset period, a cache queue length in the queue cache; The buffer queue length adjusts the set number, wherein the set number is positively correlated with the buffer queue length.

In conjunction with the first aspect or the above embodiment of the first aspect, in a possible implementation of the first aspect, the selection rule comprises a sequential loop selection.

With reference to the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the unique identifier of the TCP data stream is a quintuple of the TCP data stream, or the TCP data stream according to a preset algorithm. The quintuple is calculated to obtain a unique identifier.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the determining whether the first TCP data flow to which the to-be-processed IP data packet belongs meets a preset packet loss Before the screening condition, the method further includes: determining, according to the quintuple of the to-be-processed IP data packet, the first TCP data stream to which the IP data packet belongs.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the packet loss policy includes a random early detection (RED) algorithm.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, when determining that the to-be-processed IP data packet is not to be discarded, the method further includes: the to-be-processed IP packets are stored in the queue cache.

In a second aspect, the present application provides a data transmission control apparatus, where the apparatus includes: a data receiving module, configured to receive an Internet Protocol IP data packet to be processed; and a data processing module, configured to determine that the to-be-processed IP data packet belongs to Whether the first TCP data stream meets the preset packet loss filtering condition, and if the first TCP data stream satisfies the packet loss filtering condition, determining to discard the to-be-processed IP data according to the preset packet loss policy And deleting the to-be-processed IP data packet, if the first TCP data flow does not satisfy the packet loss screening condition, storing the to-be-processed IP data packet in a queue cache of the network transmission device .

With reference to the second aspect, in a possible implementation manner of the second aspect, the packet loss filtering condition includes: the flow state information of the TCP data stream meets a preset condition; and the data processing module determines the to-be-processed IP data. When the first TCP data stream to which the packet belongs meets the preset packet filtering condition, the method is specifically configured to: determine whether the flow state information of the first TCP data stream meets the preset condition; wherein, the first TCP The flow state information of the data stream includes a rate characterization value of the first TCP data stream, the preset condition includes a rate characterization value greater than a set value, and the rate characterization value of the first TCP data stream is used to characterize the first The transmission rate of a TCP data stream.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the rate characterization value of the first TCP data stream is stored in the queue cache within a preset duration The amount of data of the IP packet belonging to the first TCP data stream.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the flow state information of the first TCP data stream further includes that the first TCP data stream is selected last time The preset condition further includes that the selected time difference is greater than the first set duration, wherein the first TCP data stream is selected to mean that the first TCP data stream satisfies the packet loss screening condition, the first The selected time difference of the TCP data stream is the difference between the reception time of the IP packet to be processed and the time when the first TCP data stream is last selected.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the flow state information of the first TCP data stream further includes that the first TCP data stream is continuously selected The preset condition further includes that the duration of the TCP data stream being continuously selected is less than the second set duration; wherein the first TCP data stream is selected to indicate that the first TCP data stream satisfies the packet loss screening condition The duration of the first TCP data stream being continuously selected, the earliest time of the receiving time of the to-be-processed IP data packet and the first TCP data stream being consecutively selected before the receiving time of the to-be-processed IP data packet The length of time between times selected.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the device further includes: a first condition adjustment module, configured to determine the queue according to a first preset period The length of the cache queue in the cache; the set value is adjusted according to the length of the cache queue, wherein the set value is negatively correlated with the length of the cache queue.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the data processing module determines whether the first TCP data flow to which the to-be-processed IP data packet belongs is satisfied When the packet filtering condition is set, the method is specifically configured to: query the flow state information table according to the unique identifier of the first TCP data flow to which the to-be-processed IP data packet belongs, to obtain flow state information of the first TCP data flow, where The flow state information table is configured to store, by using a unique identifier of the TCP data stream, flow state information of each TCP data stream; and determining, according to the flow state information of the first TCP data flow, whether the first TCP data flow is satisfied The packet loss screening condition.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the device further includes: a flow state information table update module, configured to be in the first TCP data stream When the flow state information includes the time when the first TCP data stream is selected last time, after determining that the first TCP data flow to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, the flow state is The latest selected time of the first TCP data stream in the information table is updated to determine whether the TCP data stream to which the to-be-processed IP data packet belongs meets the packet loss screening condition or is updated to the to-be-processed IP data. The receiving time of the packet.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the packet loss filtering condition includes: the unique identifier of the TCP data stream is one of the preset data stream identifiers, The pre-configured data flow identifier includes a unique identifier of at least one of the TCP data streams stored in the queue cache.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the device further includes: a hash index rounding module, configured according to a preset time interval and a preset Selecting a rule, rotating a hash index of a set number in the pre-configured hash table, and marking the currently selected hash index; the packet filtering condition includes a hash of the TCP data stream as a hash of the current tag One of the indexes, wherein the hash value of the TCP data stream is a value obtained by hashing the unique identifier of the TCP data stream by using a hash function corresponding to the hash table.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the device further includes: a second condition adjustment module, configured to determine the queue according to a second preset period The length of the cache queue in the cache; the set number is adjusted according to the length of the cache queue, wherein the set number is positively correlated with the length of the cache queue.

With reference to the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the selecting rule includes a sequential loop selection.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the unique identifier of the TCP data stream is a quintuple of the TCP data stream, or the TCP data stream is according to a preset algorithm. The quintuple is calculated to obtain a unique identifier.

With the second aspect or the foregoing implementation manner of the second aspect, in a possible implementation manner of the second aspect, the device further includes: an associated TCP data flow determining module, configured to determine the to-be-processed IP data packet Before the first TCP data stream belongs to the preset packet loss screening condition, the first TCP data stream to which the IP data packet belongs is determined according to the quintuple of the to-be-processed IP data packet.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the packet loss policy includes a RED algorithm.

With the first aspect or the foregoing implementation manner of the first aspect, in a possible implementation manner of the first aspect, the data processing module is further configured to: when determining that the to-be-processed IP data packet is not to be discarded, The processing IP packet is stored in the queue cache.

In a third aspect, the present application provides a network transmission device, the device comprising a memory and a processor; the memory for storing computer program code; the processor for reading the computer program code to run with the computer A computer program corresponding to the program code for performing the data transmission control method in any one of the first aspect or the first aspect.

In a fourth aspect, the present application provides a computer readable storage medium having stored therein computer instructions that, when executed on a computer, cause the computer to perform the first aspect or the first aspect A data transmission control method in any of the embodiments.

In a fifth aspect, the application provides a computer program product, comprising: readable computer instructions, when the computer product is run on a computer, causing the computer to perform any of the first aspect or the first aspect described above A data transmission control method in an embodiment.

In a sixth aspect, the present application provides a computer program that, when run on a computer, causes the computer to perform the data transfer control method of any of the above first aspect or the first aspect.

DRAWINGS

1 is a schematic diagram of a network transmission architecture to which an embodiment of the invention applies;

FIG. 2 is a schematic flowchart diagram of a data transmission control method according to an embodiment of the present invention; FIG.

A schematic diagram of a hash table is shown in FIG.

FIG. 4 is a schematic diagram showing determining an active packet loss rate according to an average length of a cache queue;

FIG. 5 is a schematic diagram showing a comparison between the network transmission effect of the embodiment of the present invention and the existing packet loss mode;

FIG. 6 is a schematic diagram of a frame of a data transmission control apparatus according to an embodiment of the invention; FIG.

FIG. 7 is a schematic diagram of a frame of a data transmission control apparatus according to another embodiment of the present invention; FIG.

FIG. 8 is a block diagram showing a data transmission control apparatus according to still another embodiment of the present invention; FIG.

FIG. 9 is a schematic block diagram of a network transmission device according to an embodiment of the present invention.

Detailed ways

Due to the unpredictability of network traffic, the burstiness and duration of traffic are not predictable, and the cache of network transmission equipment is limited. Therefore, all network transmission equipment on the end-to-end transmission path is at risk of packet loss. In particular, it is a device that caches smaller switches and access network devices.

The embodiments of the present invention provide a data transmission control method, device, device, and storage medium, which are applicable to an application scenario in which IP data packets are lost due to excessive network traffic in all IP networks.

FIG. 1 is a schematic diagram of a typical network transmission architecture of a metropolitan area network according to an embodiment of the present invention. As shown in FIG. 1 , the network transmission architecture may include a core router (CR), a broadband remote access server (BRAS), an LSW, an OLT, and the like. Each network transmission device may be provided with a queue buffer for buffering IP data packets received by the device. When the network traffic bursts, the network transmission device stores the received IP data packets in a queue to the queue cache. , such as the queue buffer of each network transmission device shown in FIG. 1 and the cache queue (the queue of stored IP data packets) in each queue cache, as shown for the CR device, as shown schematically in FIG. Cache a and cache queue b. The data transmission control method of the embodiment of the present invention is applicable to any of the CR, BRAS, LSW, and OLT shown in FIG. 1, that is, CR, BRAS, LSW, and OLT may all be implemented by the present invention. The execution body of the data transmission control method of the example.

The data transmission control method of the embodiment of the present invention is further described below by taking the LSW in FIG. 1 as an example. LSW is generally not equipped with a large queue cache due to cost considerations, so packet loss is likely to occur when network traffic is suddenly congested.

In order to reduce the number of affected TCP data streams when the packet is lost, the embodiment of the present invention pre-configures the packet loss filtering condition on the LSW. When receiving the IP data packet, the LWS first determines whether the TCP data stream to which the IP data packet belongs is first. The packet loss filtering condition is met. When the TCP data stream to which the IP data packet belongs meets the packet loss filtering condition, the IP packet is further discarded according to the preset packet loss policy. With this method, when the LSW needs to drop packets, the LSG concentrates the discarded IP data packets on the TCP data stream that satisfies the above-mentioned packet loss screening condition, and implements a targeted selective packet loss, thereby reducing the TCP affected by the packet loss. The number of data streams reduces the number of users affected by packet loss and improves the overall user experience of network users.

FIG. 2 is a schematic flowchart diagram of a data transmission control method according to an embodiment of the present invention, where the method may be specifically performed by a network transmission device. As shown in FIG. 2, the data transmission control method may mainly include the following steps:

Step S11: Receive an IP data packet.

For convenience of description, the IP data packet received in step S11 is hereinafter referred to as an IP data packet to be processed.

Step S12: Determine whether the TCP data stream to which the IP data packet to be processed belongs meets the preset packet loss screening condition.

For convenience of description, the TCP data stream to which the IP packet to be processed belongs is referred to as the first TCP data stream.

In an optional embodiment of the present invention, before determining whether the first TCP data stream to which the to-be-processed IP data packet belongs meets the preset packet loss filtering condition, the method further includes:

The first TCP data stream to which the to-be-processed IP data packet belongs is determined according to the quintuple of the IP data packet to be processed.

In order to reduce the number of TCP data streams affected by packet loss, and thereby reduce the number of affected users, the network transmission device first determines the IP to be processed when receiving the IP data packets to be processed sent by other network transmission devices or user terminal devices. The first TCP data stream to which the data packet belongs is determined according to whether the first TCP data stream satisfies the packet loss screening condition to determine whether the to-be-processed IP data packet is an alternate IP data packet to be discarded, that is, the network transmission device determines the need IP packets to be discarded when packet loss occurs.

The quintuple of the IP data packet refers to the source IP address, source port, destination IP address, destination port, and transport layer protocol of the IP data packet. The quintuple can distinguish different sessions, and the corresponding session is unique, that is, one session. Corresponding to a TCP link and a TCP data stream, therefore, the quintuple can uniquely identify the TCP data stream to which the IP data packet belongs, and the same IP data packet of the quintuple belongs to the same TCP data stream.

The foregoing packet loss screening condition is a data stream for determining whether the TCP data stream is an alternative and needs to be discarded by the IP data packet for the TCP data stream. Based on the packet loss screening condition, the IP data packets discarded by the network transmission device are concentrated only on the TCP data stream that satisfies the packet loss screening condition.

In an optional embodiment of the present invention, the packet loss screening condition includes that the flow state information of the TCP data stream meets a preset condition. Correspondingly, determining whether the first TCP data flow to which the to-be-processed IP data packet belongs meets the preset packet loss screening condition, specifically:

Determining whether the flow state information of the first TCP data stream satisfies the foregoing preset condition.

The flow state information of the first TCP data stream includes a rate characterization value of the first TCP data stream, and the preset condition includes that the rate characterization value is greater than a set value, and the rate characterization value of the first TCP data stream is used to represent the first TCP data. The transmission rate of the stream.

It should be noted that the “to-be-processed” in the above-mentioned IP data packet to be processed is only to distinguish the received IP data packet from the IP data packet involved in describing the packet loss screening condition or the preset condition. Similarly, the “first” in the foregoing first TCP data stream is only used to compare the TCP data stream to which the received IP data packet to be processed belongs to the TCP data involved in describing the packet loss filtering condition or the preset condition. The flow is separated and no other has a specific order or relationship definition.

In the embodiment of the present invention, for any TCP data stream, the transmission rate of the TCP data stream may be the transmission rate of the TCP data stream or the receiving rate of the TCP data stream. The rate characterization value of the TCP data stream may be the transmission rate itself of the TCP data stream, for example, may be the reception rate or transmission rate of the TCP data stream. For any TCP data stream, the rate representation value of the TCP data stream may also be the amount of data of the IP data packet belonging to the TCP data stream received or transmitted within a preset duration.

In an optional embodiment of the present invention, for any TCP data stream, the data volume of the IP data packet that belongs to the TCP data stream received within the preset duration may be received by the TCP data stream that is received within the preset duration. The data amount of the IP data packet indicates that the data amount of the IP data packet belonging to the TCP data stream sent within the preset time length can be used for the data of the IP data packet belonging to the TCP data stream stored in the queue buffer within a preset time period. The quantity indicates that, at this time, the traffic of the TCP data stream is the amount of data of the IP data packet belonging to the TCP data stream stored in the queue buffer within a preset duration.

For the first TCP data stream, the rate representative value of the first TCP data stream may be the data amount of the IP data packet belonging to the first TCP data stream stored in the queue buffer within a preset duration.

In an optional embodiment of the present invention, for any TCP data stream, the data volume of the IP data packet belonging to the TCP data stream may be the number of bytes of the IP data packet belonging to the TCP data stream and/or the IP data packet. The number, correspondingly, the rate of the TCP data stream is greater than the set value. The number of bytes of the IP packet belonging to the TCP stream is greater than the set number of bytes and/or the IP packet belonging to the TCP stream. The number is greater than the set number.

When the network transmission device sends an IP packet, it adjusts the transmission rate by adjusting the size of the transmission window according to the state of the network. When the network is in a light load state, the transmission window will be continuously expanded, the transmission rate will be increased, and the network capacity will be tested. When the network is judged to be in a congested state (generally, packet loss is used as a signal for network congestion), the transmission window will be narrowed (referred to as window down). ), reduce the transmission rate to alleviate network congestion.

In the current TCP window-down mechanism, as long as the TCP RTO is not triggered, the network transmission device will only perform a window down (retransmission packet is discarded or the entire transmission window), regardless of whether multiple packets are dropped or one packet is dropped. The dropped packet will trigger the TCP RTO, and the TCP send window will drop to the initial value, which will have a greater impact on the transmission performance of the TCP data stream. When a packet loss occurs, the sender device that has dropped the packet can retransmit the discarded packet based on the Selective Acknowledgments (SACK) feature of TCP. The SACK feature refers to the TCP receiving device to describe the receiving information in more detail, including which packets, that is, IP packets have been received and which packets have not been received, so that the sending device can only discard the packets. The packet is retransmitted.

According to the window-down mechanism and the SACK feature of the TCP, for the same TCP data stream, in a transmission window, whether a packet is dropped or multiple packets are dropped, the TCP data is not triggered under the premise of not triggering the TCP RTO. The performance of the stream has almost no effect. However, in order to avoid triggering the TCP RTO, you need to control the number of lost packets in each data stream. The large number of packets that can be tolerated by the large stream is larger, and the packet loss is concentrated on the large stream, which can better control the packet loss. influences. In the embodiment of the present invention, the large stream refers to a TCP data stream whose transmission rate is greater than a set rate in a TCP data stream transmitted by the network transmission device, where the transmission rate may be a reception rate or a transmission rate, and TCP The receiving rate of the data stream may be represented by the amount of data of the IP data packet that belongs to the TCP data stream received within the preset duration, and the sending rate may be the TCP data that belongs to the queue buffer within a preset duration. The amount of data of the stream indicates that, for example, for a TCP data stream, if the amount of data of the IP data packet belonging to the TCP data stream stored in the queue buffer within a preset duration is greater than the set data amount, the The TCP data stream is a large stream.

Therefore, in the embodiment of the present invention, the preset condition may be set such that the rate representation value of the TCP data stream is greater than the set value. When receiving the IP data packet to be processed, the network transmission device may determine, according to the traffic of the first TCP data flow to which the IP data packet to be processed belongs and the set value, the first TCP data flow to which the to-be-processed IP data packet belongs. Whether the preset conditions are met.

As a specific example, for example, the preset duration is configured to be 2 seconds, and the rate representation value of the TCP data stream is represented by the number of IP data packets belonging to the TCP data stream stored in the queue buffer within 2 seconds, and the number is set. Is 8. Assume that there are 21 IP data packets stored in the queue buffer within the current 2 seconds. Five of the 21 data packets belong to TCP data stream A, seven belong to TCP data stream B, and nine belong to TCP data stream C. Then, the rate characterization value of the TCP data stream A is 5, the rate characterization value of the TCP data stream B is 7, and the rate characterization value of the TCP data stream C is 9, and the pending IP belonging to the TCP data stream C is currently received. In the case of the data packet, since the traffic of the TCP data stream C to which the to-be-processed IP data packet belongs is 9, which is greater than the set number 8, the TCP data stream C to which the to-be-processed IP data packet belongs satisfies the preset condition.

In an optional embodiment of the present invention, the flow state information may further include a time when the first TCP data stream is last selected, that is, a time when the first TCP data stream is last determined to meet the packet loss screening condition, and corresponding The preset condition further includes that the selected time difference is greater than the first set duration, wherein the first TCP data stream is selected to indicate that the first TCP data stream satisfies the packet loss screening condition, and the selected time difference of the first TCP data stream is to be processed. The difference between the reception time of the IP packet and the time when the first TCP stream was last selected.

In the embodiment of the present invention, when a TCP data stream is lost, the same TCP time can be avoided by recording the time when the TCP data stream is last selected, and setting the packet protection period, that is, the first setting duration. The data stream is selected for multiple times in a packet loss protection period, thereby avoiding the situation that the TCP RTO is triggered due to the excessive number of lost packets of the same data stream.

It should be noted that, for the same TCP data stream, each time an IP data packet belonging to the TCP data stream is received, it is determined whether the TCP data stream satisfies the packet loss filtering condition, and if it is satisfied, the TCP data stream is described. The time is selected, if not, the TCP stream is not selected at this time. The time when the TCP data stream is selected may be the time for determining whether the TCP data stream satisfies the packet loss filtering condition, or the time when the TCP data stream is selected to receive the IP data packet of the TCP data stream.

As a specific example, it is assumed that the network transmission device receives an IP data packet s, the reception time of the packet s is t, and the packet s belongs to the TCP data stream a. Before the time t, the network transmission device has successively received three IP data packets belonging to the TCP data stream a. For the convenience of description, the three IP data packets are respectively recorded as the packet 1, the packet 2, and the packet 3 according to the receiving time. When receiving the packet 1, it is determined that the TCP data stream to which the packet 1 belongs (ie, the TCP data stream a) does not satisfy the packet loss filtering condition, and then the TCP data stream a is not selected, and then when the packet 2 is received, it is determined. The TCP data stream to which the packet 2 belongs (ie, the TCP data stream a) satisfies the packet loss filtering condition, and then the TCP data stream is selected, and when the packet 3 is received, the TCP data stream to which the packet 3 belongs is determined (ie, the TCP data stream a) When the packet data filtering condition is met, the TCP data stream is also selected at this time, and when the network transmitting device receives the packet s, the TCP data stream a is selected last time, and when the packet 3 is received, it is determined that the packet 3 belongs. The TCP data stream satisfies the packet loss filtering condition. The time when the TCP data stream a is selected last time may be the time for determining that the TCP data stream to which the packet 3 belongs meets the packet filtering condition, or the receiving time of the packet 3. In this specific example, for the packet s, the selected time difference of the TCP data stream a to which the packet s belongs is between the reception time of the packet s and the time when the TCP data stream a was last selected before the reception time of the packet s. The difference between the reception time of the packet s and the time at which the TCP data stream to which the packet 3 belongs satisfies the packet loss screening condition, or the difference between the reception time of the packet s and the reception time of the packet 3 .

In an optional embodiment of the present invention, the flow state information of the first TCP data stream may further include a duration in which the first TCP data stream is continuously selected. Correspondingly, the preset condition further includes that the duration of the TCP data stream is continuously selected is less than The second set time.

As can be seen from the foregoing description, the first TCP data stream is selected to mean that the first TCP data stream satisfies the packet loss screening condition, and the duration of the first TCP data stream is continuously selected, which is the receiving time of the IP data packet to be processed and the first TCP data. The length of time between the earliest selected time that the stream is continuously selected before the reception time of the IP packet to be processed. For example, for the specific example of the TCP data stream a in the foregoing, for the packet s, the TCP data stream a is continuously selected before the reception time of the packet s, which means that for the TCP data stream a, the TCP data is continuously received. When packet 2 and packet 3 of stream a are selected, the two TCP data streams are selected. At this time, the duration of the first TCP data stream is continuously selected, and the reception time of the packet s and the reception of the TCP data stream a in the packet s are received. The difference between the time when the first time is selected consecutively before the time is selected (when the packet 2 is received, the TCP data stream a to which the packet 2 belongs is determined to satisfy the packet loss filtering condition), specifically, the receiving time and the determination of the packet s The difference between the time when the TCP data stream to which the packet 2 belongs satisfies the packet loss screening condition, or the difference between the reception time of the packet s and the reception time of the packet 2.

It can be understood that the first TCP data stream is selected one by one in the consecutively selected operations before the receiving time of the IP packet to be processed, which means that the first TCP is received before the IP data to be processed is received. The IP data packet of the data stream determines that the first TCP data stream is selected (ie, the packet loss filtering condition is satisfied).

In the embodiment of the present invention, by setting the packet loss rotation period, that is, the second setting duration, it is ensured that the duration of consecutive packet loss of the same TCP data stream is not greater than the second setting duration, that is, one TCP data stream is continuous. When the length of the packet loss is not less than the second set duration, the packet loss needs to be rotated from the TCP data stream to another TCP data stream to reduce the possibility that the TCP RTO is triggered.

In practical applications, in order to determine the length of time that an IP packet of a TCP data stream is being discarded and continuously lost, the IP data packet from the TCP data stream may be received when the IP data packet of the TCP data stream is discarded. The discarding time starts, and the duration of the continuous loss of the TCP data stream (the length of time that the IP data packets belonging to the TCP data stream are continuously discarded) is counted.

In an optional embodiment of the present invention, in order to improve the processing efficiency of the IP data packet, the determining whether the first TCP data stream to which the IP data packet to be processed belongs to meet the preset packet filtering condition may include:

Querying the flow state information table according to the unique identifier of the first TCP data flow to which the IP data packet to be processed belongs, and obtaining flow state information of the TCP data flow to which the IP data packet belongs; wherein the flow state information table is used to uniquely use the TCP data flow Identifying, as an index, storing flow state information of each TCP data stream;

And determining, according to the flow state information of the first TCP data stream, whether the first TCP data flow satisfies the packet loss screening condition.

In the embodiment of the present invention, when the packet loss screening condition includes the flow state information of the TCP data flow meeting the preset condition, the flow state information table may be set to record each TCP data flow in the cache queue of the network transmission device. The latest flow state information, so that when the network transmission device receives the to-be-processed IP data packet, the flow state information of the first TCP data flow to which the to-be-processed IP data packet belongs can be directly queried in the flow state information table. Then, it is quickly determined whether the first TCP data stream to which the to-be-processed IP data packet belongs meets the packet loss screening condition.

It can be understood that the flow state information of the TCP data stream stored in the flow state information table is also different for the specific information included in the flow state information. For example, if the flow state information includes a rate characterization value of the TCP data stream, the flow state information stored in the flow state information table by using the unique identifier of the TCP data stream as an index includes a rate characterization value of the TCP data stream; The status information also includes the time when the TCP data stream was last selected, and the stream status information stored in the flow status information table also includes the time when the TCP data stream was last selected.

In actual applications, the flow state information of each TCP data stream received within the preset duration may be counted according to the preset duration, and the statistics may be updated into the flow state information table to implement the convection state information. Update of the flow state information of the TCP data stream stored in the table.

Similarly, when the flow state information of each TCP data stream received within the preset duration is counted, the flow state information to be counted is different according to the specific information included in the flow state information. For example, for any TCP data stream, if the rate characterization value of the TCP data stream is the amount of data of the IP data packet belonging to the TCP data stream stored in the queue buffer within a preset duration, the required statistical stream state information is The amount of data of the IP data packet belonging to the TCP data stream stored in the queue buffer within a preset duration; if the rate of the TCP data stream is an IP packet that is received within the preset duration and belongs to the TCP data stream The amount of data, the flow state information that needs to be counted is the amount of data of the IP packet belonging to the TCP data stream received within the preset duration.

It can be understood that if the unique identifier of a TCP data stream stored in the queue buffer does not exist in the current state information table within the current preset duration, the unique identifier of the TCP data stream can be directly indexed. The flow state information corresponding to the TCP data stream may be stored in the flow state information table; if the unique identifier of a TCP data flow already exists in the flow state information table before the current preset duration is set, the The stream state information of the TCP stream that is newly counted in the current preset duration replaces the stream state information of the TCP stream that is counted by the previously set duration, and the flow state information of the TCP stream is updated.

As a specific example, a schematic diagram of a flow state information table is shown in Table 1. In this example, the flow state information includes a rate representation value of the TCP data stream (characterized by the number of IP data packets of the TCP data stream stored in the queue buffer in a preset time) and a time when the TCP data stream was last selected ( The time when the TCP data stream was last selected), where the stream sequence number represents the unique identifier of the TCP stream.

Table 1

流序号Flow number	流大小(包个数)Stream size (number of packets)	上次被选中时间Last selected time
11	m_1 M_1		t_1T_1
22	m_2M_2	t_2T_2
……......	……......	……......
NN	m_Nm_N	t_Nt_N

Based on the flow state information table shown in Table 1, when the network transmission device receives an IP data packet, it can find the corresponding flow state information in the table according to the flow sequence number of the TCP data flow to which the IP data packet belongs. For example, if the received stream number of the TCP data stream to which an IP data packet belongs is 1, the rate of the TCP data stream is represented by m_1, and the last time selected (most recently selected) is t_1.

In an optional embodiment of the present invention, if the unique identifier of the first TCP data stream to which the IP data packet to be received belongs does not exist in the flow state information table, the data transmission control method may further include: determining the IP data packet to be processed. The rate characterization value of the associated first TCP data stream is zero.

That is, if the unique identifier of the first TCP data stream to which the IP data packet to be processed belongs (such as the stream sequence number in Table 1) is not found in the flow state information table, the number of the pending IP data packet belongs to A TCP data stream does not exist in the queue cache. The currently received IP data packet is the first data packet of the first TCP data stream to which the IP data packet belongs in the current state of the network transmission device. The traffic of the TCP data stream to which the IP packet to be processed belongs belongs to zero.

It can be understood that if the unique identifier of the first TCP data stream to which the IP data packet to be processed belongs does not exist in the flow state information table, it indicates that the first TCP data stream to which the to-be-processed IP data packet belongs does not satisfy the packet loss screening. condition.

In an optional embodiment of the present invention, if the first TCP data stream state information further includes a time when the first TCP data stream is last selected, determining that the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset. After the packet loss screening condition, the data transmission control method of the embodiment of the present invention further includes:

Updating the last selected time of the first TCP data stream in the flow state information table, specifically, updating the last selected time of the first TCP data stream in the flow state information table to determine that the pending IP data packet belongs to The first TCP data stream satisfies the time of discarding the screening condition or is updated to the receiving time of the IP packet to be processed.

It can be understood that, when determining whether the received first TCP data stream to which the to-be-processed IP data packet belongs meets the packet loss filtering condition, if the set packet filtering condition is different, determining that the to-be-processed IP data packet belongs to After the first TCP data stream satisfies the packet loss screening condition, the stream state information of the first TCP data stream to be updated is different. For example, the flow state information in the packet loss filtering condition includes the time when the TCP data stream is last selected, and after the TCP data stream is selected, the latest selected time of the first TCP data stream needs to be updated to be determined. Processing the time when the first TCP data stream to which the IP data packet belongs meets the discarding filter condition or is updated to the receiving time of the IP packet to be processed. For example, if the flow state information in the packet loss filtering condition includes the duration in which the TCP data stream is continuously selected, after determining that the first TCP data flow to which the to-be-processed IP data packet belongs meets the packet loss filtering condition, the statistics need to be started. The length of time during which a TCP data stream is continuously selected or the length of time that the already counted first TCP data stream is continuously selected.

In an optional embodiment of the present invention, the data transmission control method may further include:

Calculating the length of the cache queue in the queue cache according to the first preset period;

The set value in the preset condition is adjusted according to the length of the buffer queue, wherein the set value is negatively correlated with the buffer queue length.

In the embodiment of the present invention, the network transmission device can implement periodic adjustment of the set value in the first screening condition according to the buffer queue length in the queue cache, so that the packet loss filtering condition is more consistent with the actual storage state of the cache in the device. The length of the cache queue refers to the length of the IP data packet stored in the queue cache. The length of the cache queue may be represented by the number of bytes of the IP data packet stored in the queue cache, or may be in the queue cache. Characterized by the number of stored IP packets.

It can be understood that, when the set value is adjusted according to the length of the cache queue, the corresponding adjustment rule may be pre-configured, for example, the correspondence between the length of the cache queue and the set value is configured. After the cache queue length is calculated, the configuration may be configured according to the configuration. This correspondence adjusts the set value. Currently, in the actual application, after calculating the length of the buffer queue, after performing other processing based on the calculated buffer queue length, the set value is adjusted according to the processing result. For example, in a specific embodiment, adjusting the set value in the preset condition according to the length of the cache queue may include:

Calculating the active packet loss rate of the network transmission device according to the length of the cache queue;

The set value is adjusted according to the active packet loss rate, wherein the active packet loss rate has a negative correlation with the set value.

That is, the set value can be adjusted according to the current active packet loss rate of the network transmission device. When the active packet loss rate is high, the set value is lowered to make the number of selected TCP data streams more and more. It conforms to the actual application scenario of packet loss.

It can be understood that, in actual applications, in addition to adjusting the set value according to the length of the cache queue, the first set duration and/or the second set duration in the preset condition may be adjusted according to the buffer queue length. Specifically, the cache queue length can be configured to have a negative correlation change with the first set duration and/or the second set duration.

In the actual application scenario, because the storage space and computing power of the network transmission device are limited, and the amount of data transmitted in the network is large, the network transmission device may select the TCP data stream based on the preset condition. It is difficult. For example, maintaining a flow state information table requires a certain amount of storage space, and the complexity of maintaining the flow state information in the table may be unbearable for some network transmission devices. Therefore, in order to avoid the problem, in the embodiment of the present invention, the screening of the TCP data stream may be implemented based on the unique identifier of the TCP data stream.

The unique identifier of the TCP data stream may be a quintuple corresponding to the TCP data stream, or a unique stream identifier calculated by calculating a quintuple corresponding to the TCP data stream according to a preset algorithm, or may be another unique identifier. The identifier of a TCP data stream.

In an optional embodiment of the present invention, the foregoing packet loss filtering condition may include: the unique identifier of the TCP data stream is one of the preset data stream identifiers, where the pre-configured data stream identifier includes the TCP data stream stored in the queue cache. The unique identifier of at least one of the TCP data streams.

That is, the data flow identifier of the TCP data stream may be pre-set in the network transmission device, and the possible identifier may be matched by matching the unique identifier of the first TCP data stream to which the received IP data packet belongs to the preset data stream identifier. TCP data stream that needs to be dropped. The number of the pre-configured data stream identifiers may be selected according to requirements. If the unique identifier of the first TCP data stream to which the received IP data packet belongs is one of the preset data stream identifiers, the first A TCP data stream satisfies the packet loss filtering condition.

It can be understood that the foregoing preset data stream identifier may be a unique identifier of one or more TCP data streams in the TCP data stream to which the IP data packet stored in the cache belongs. Further, the preset data stream identifier may be selected according to the size of the data stream in the TCP data stream. For example, the identifier of the TCP stream of the large stream is preferably used as a preset data stream identifier. In addition, the preset data stream identifier may be Periodic adjustments are made based on changes in the data stored in the cache.

In an optional embodiment of the present invention, the data transmission control method of the embodiment of the present invention may further include:

According to the preset time interval and the preset selection rule, the hash index of the set number in the pre-configured hash table is rotated, and the currently selected hash index is marked;

At this time, the corresponding packet loss filtering condition includes: the hash value of the TCP data stream is one of the hash indexes of the current label, wherein the hash value of the TCP data stream is a hash function corresponding to the hash table, The unique identifier of the TCP data stream is hashed to obtain the value.

In the embodiment of the present invention, the TCP data stream that may need to drop the data packet may be selected by pre-configuring the hash table and periodically selecting the hash index in the hash table according to a preset time interval. Therefore, when the IP data packet to be processed is received, the first TCP may be quickly determined according to whether the hash value of the first TCP data stream to which the IP data packet to be processed belongs is one of the currently selected hash indexes. Whether the data stream satisfies the packet loss filtering condition. In this manner, when packet loss is required, the discarded IP data packet is concentrated on the TCP data stream corresponding to each selected hash index, and other unselected ha The TCP data stream corresponding to the index is not lost.

It can be understood that the hash function corresponding to the above hash table refers to a hash function used when initializing the hash table, and the specific selection of the hash function can be determined according to actual needs. For example, the hash table may be used to create a hash table. The hash function may directly modulate the keyword pair with a number that is not greater than the length of the hash table. In the embodiment of the present invention, the keyword is a TCP data stream. Uniquely identifies. When the unique identifier of the TCP data stream is the quintuple corresponding to the data stream, the quintuple of the received IP data packet may be directly hashed according to the hash function, and the calculated hash value is IP. The hash value of the TCP data stream to which the data packet belongs is determined by determining whether the hash value belongs to one of the marked hash indexes, and quickly determines whether the TCP data stream to which the IP data packet belongs satisfies the packet loss filtering condition.

The time interval, the number of settings, and the selection rule can be set and adjusted according to actual application requirements. In an alternative embodiment of the invention, the selection rules include sequential cycle selection.

Calculating the length of the cache queue in the queue cache according to the second preset period;

The number of the above settings is adjusted according to the length of the buffer queue, wherein the set number is positively correlated with the length of the buffer queue.

In the embodiment of the present invention, the number of hash indexes selected in each queue may be adjusted according to the length of the cache queue in the queue buffer of the network transmission device. The longer the queue queue length is, the larger the number of settings is, and the selected number may be selected. The greater the number of dropped TCP streams, the greater the probability that the received IP packets are selected to be discarded, so that the data transmission control is more in line with the actual storage state in the queue buffer of the network transmission device. The specific adjustment policy for adjusting the number of settings according to the length of the cache queue may be set according to actual application requirements. For example, in one specific example, adjusting the number of settings according to the length of the cache queue may include:

According to the active packet loss rate, the number of settings is adjusted, wherein the active packet loss rate is positively correlated with the set number.

That is, according to the active packet loss rate, the number of settings is adjusted. The higher the active packet loss rate, the larger the value of the set number, and the more hash indexes are selected each time.

As a specific example, a schematic diagram of a hash table is shown in FIG. In this example, the hash function corresponding to the hash table directly modulates the number p of the keyword pair not greater than the hash table table length (ie, the number of indexes), and the hash table includes N hash indexes, the table 1, 2, ..., N schematically show N hash values, that is, hash indexes. When the hash index in the hash table is sequentially cycled, the rotation may be performed according to a preset time interval, and the number of settings may be selected according to actual application requirements. For example, in the ith round selection, the hash index is The TCP data stream of 1, 2 is selected. When the i+1th round is selected, the TCP data stream with the hash index of 3, 4, and 5 is selected, and when a round selection reaches the last Nth hash index. Then, you can go back to the first hash index to achieve the loopable selection. After receiving the modulo IP data packet, the unique index of the first TCP data stream to which the IP data packet to be processed belongs is modulo, if the modulo result is one of the currently selected hash indexes, Quickly determine that the TCP data stream of the IP data packet satisfies the packet loss filtering condition.

Step S131: If the first TCP data stream meets the packet loss screening condition, the IP data packet to be processed is discarded when it is determined that the IP data packet to be processed is to be discarded according to the preset packet loss policy.

Step S132: If the first TCP data stream does not meet the packet loss filtering condition, the IP data packet to be processed is stored in a queue cache of the network transmission device.

In the embodiment of the present invention, different packet loss policies may be configured according to actual application requirements to determine whether IP data packets satisfying the packet loss screening condition need to be discarded.

In an optional embodiment of the present invention, the packet loss policy may be directly configured as a RED algorithm.

At this time, it is possible to directly determine, according to the RED algorithm, whether IP packets of the TCP data stream that meet the packet loss screening condition need to be discarded. Specifically, when the RED algorithm is used to determine whether the IP packet needs to be discarded, the active packet loss rate of the network transmission device needs to be determined. If the active packet loss rate is not less than the first packet loss rate threshold, the IP data packet is discarded. If the packet loss rate is greater than the second packet loss threshold and is less than the first packet loss threshold, the IP packet is randomly generated. If the random number is not less than the pre-configured packet threshold, the IP packet is discarded. If the packet rate is not greater than the second packet loss threshold, the IP packet is not discarded. The first packet loss rate threshold is greater than the second packet loss rate threshold, and the first packet loss rate threshold and the second packet loss rate threshold may be set as needed.

In the embodiment of the present invention, the corresponding packet loss threshold may be pre-configured for the active packet loss rate that is greater than the second packet loss threshold and smaller than the first packet loss threshold. For example, the packet loss threshold is not less than the second packet loss threshold. The packet loss threshold corresponding to the active packet loss rate of the first packet loss rate threshold may be configured as X, and the active packet loss rate of the device at this time is met when the TCP data stream of the received IP packet satisfies the packet loss screening condition. If the second packet loss rate threshold is greater than the first packet loss rate threshold, a random number Y is generated for the IP data packet, and if Y is not less than X, the IP data packet is determined to be discarded. In an example, the first packet loss rate threshold may be set to 100%, the second packet loss rate threshold may be set to 0%, and the packet loss threshold may be configured to be 60, and when the active packet loss rate is 100%, The IP data packet that meets the packet loss filtering condition is directly discarded. If the active packet loss rate is greater than or equal to 60% and less than 100%, a random number between 1 and 100 is randomly generated for the IP data packet, and the generated random number is assumed to be 50. Since 50 is smaller than the packet loss threshold 60, the IP data packet is not discarded. If the generated random number is 70, since 70 is greater than the packet loss threshold 60, the IP data packet is discarded.

The RED packet loss method measures the congestion level according to the length of the cache queue of the network transmission device. The longer the queue length, the more serious the congestion degree and the higher the active packet loss rate. FIG. 4 shows a specific example of determining the active packet loss rate according to the average queue length in the RED. In the figure, the abscissa ave_qlen represents the buffer queue length of the device's queue buffer, and the ordinate ρ represents the active packet loss rate. As shown in FIG. 4, when the buffer queue length is less than the threshold min_thr, the active packet loss rate is 0 when no packet loss occurs, and the active packet loss rate is linear according to the length of the buffer queue when the buffer queue length is between the threshold min_thr and the threshold max_thr. The maximum active packet loss rate is max_p. When the cache queue length exceeds the threshold max_thr, the active packet loss rate is 1. When receiving an IP packet, if the active packet loss rate of the device is 1, the filtered IP packet is directly discarded. If the active packet loss rate is 0, the IP packet is stored in the queue cache. If the packet loss rate is greater than 0 and less than 1, the random number is used to determine whether to discard.

In the data transmission control method of the embodiment of the present invention, the active data packet is discarded, and the IP data packet of the TCP data stream that meets the packet loss filtering condition is used as an alternative IP packet to be discarded, and then The packet policy (such as the RED algorithm) determines whether to discard the candidate IP packets to be discarded that are filtered by the packet loss filtering condition. The method of the embodiment of the present invention achieves that the packet loss is concentrated on part of the TCP data stream, which is reduced. The number of users affected by packet loss ensures the transmission of other TCP data streams without further deteriorating the user experience, which improves the experience of other users and thus improves the user experience statistically.

In a specific example, FIG. 5 is a schematic diagram showing a data transmission control method, a random drop method based on AQM technology, and a throughput change of a network transmission device based on a tail drop method without using AQM, according to an embodiment of the present invention, Wherein, the abscissa represents time, the ordinate represents port utilization of the device, the curve S1 in the figure is a device port utilization curve according to the data transmission control method of the embodiment of the present invention, and the curve S2 is a random discarding method based on the AQM technology. The device port utilization curve, and curve S3 is a device port utilization curve based on the tail drop method that does not use AQM. As can be seen from the curves S1, S2, and S3, the method of the embodiment of the present invention can significantly improve the port utilization rate when the network is lightly loaded, because congestion occurs in the network transmission, and window down occurs. When the number of affected TCP data streams is reduced, the number of windowed TCP data streams is small. Under the same congestion level, the port utilization rate of the embodiment of the present invention is significantly improved.

Corresponding to the data transmission control method shown in FIG. 2, FIG. 6 is a schematic diagram of a frame of a data transmission control apparatus 100 according to an embodiment of the present invention. The data transmission control apparatus 100 may be implemented as a processor or directly. Deployed in a network transport device. As shown in FIG. 6, the network transmission control apparatus 100 may include a data receiving module 110 and a data processing module 120. among them:

The data receiving module 110 is configured to receive an IP data packet to be processed.

The data processing module 120 is configured to determine whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, and if the first TCP data stream satisfies the packet loss filtering condition, the data packet according to the preset packet loss When the policy determines that the IP packet to be processed is to be discarded, the IP packet to be processed is discarded. If the first TCP data stream does not meet the packet filtering condition, the IP packet to be processed is stored in the queue buffer of the network transmission device.

In the embodiment of the present invention, when receiving the IP data packet to be processed, the network transmission control apparatus 100 first determines whether the first TCP data stream to which the received IP data packet belongs belongs to the packet loss screening condition, and only meets the lost condition. The packet filtering condition is further determined according to the packet loss policy whether the IP packet to be processed needs to be discarded. Through the embodiment of the present invention, the network transmission control apparatus 100 concentrates the IP data packets that need to be discarded into the TCP data flow that meets the packet loss screening condition, thereby reducing the number of TCP data streams affected by the lost packet and reducing the packet loss. The number of users affected increases the overall user experience.

In an optional embodiment of the present invention, the packet loss screening condition includes that the flow state information of the TCP data stream meets a preset condition. Correspondingly, the data processing module 120 is specifically configured to: when determining whether the first TCP data stream to which the to-be-processed IP data packet belongs meets the preset packet loss filtering condition, specifically:

Determining whether the flow state information of the first TCP data stream meets a preset condition; wherein the flow state information of the first TCP data flow includes a rate representation value of the first TCP data flow, and the preset condition includes that the rate representation value is greater than a set value, The rate characterization value of the first TCP data stream is used to characterize the transmission rate of the first TCP data stream.

In an optional embodiment of the present invention, the rate characterization value of the first TCP data stream is the amount of data of the IP data packet belonging to the first TCP data stream stored in the queue buffer within a preset duration.

In an optional embodiment of the present invention, the flow state information of the first TCP data stream further includes a time when the first TCP data stream is last selected, and the preset condition further includes that the selected time difference is greater than the first set duration, where the first TCP When the data stream is selected, the first TCP data stream satisfies the packet loss screening condition, and the selected time difference of the first TCP data stream is the difference between the receiving time of the IP packet to be processed and the time when the first TCP data stream is last selected.

In an optional embodiment of the present invention, the flow state information of the first TCP data stream further includes a duration in which the IP data packet of the first TCP data stream is continuously selected, and the preset condition further includes that the IP data packet of the TCP data stream is continuously selected. The length of time is less than the second set duration, wherein the first TCP data stream is selected to mean that the first TCP data stream satisfies the packet loss screening condition, and the duration of the first TCP data stream is continuously selected. The length of time between the reception time of the to-be-processed IP data packet and the earliest selected time of the first TCP data stream that is consecutively selected before the reception time of the to-be-processed IP data packet.

In an optional embodiment of the present invention, the data transmission control apparatus 100 may further include a first condition adjustment module.

The first condition adjustment module is configured to determine a buffer queue length in the queue cache according to the first preset period; and adjust the set value according to the buffer queue length, wherein the set value is negatively correlated with the buffer queue length.

In an optional embodiment of the present invention, the data processing module 120 is specifically configured to: when determining whether the first TCP data stream to which the IP data packet to be processed belongs meets the preset packet loss screening condition, specifically:

Querying the flow state information table according to the unique identifier of the first TCP data stream to which the IP data packet to be processed belongs, and obtaining flow state information of the first TCP data flow, where the flow state information table is used to store the unique identifier of the TCP data flow as an index. The flow state information of the TCP data stream; determining, according to the flow state information of the first TCP data flow, whether the first TCP data flow satisfies the packet loss filtering condition.

In an optional embodiment of the present invention, the data transmission control apparatus 100 may further include a flow state information table update module 121, as shown in FIG.

The flow state information table update module 121 is configured to: when the flow state information of the first TCP data flow includes the time when the first TCP data flow is last selected, determine that the first TCP data flow to which the to-be-processed IP data packet belongs meets the pre- After the packet filtering condition is set, the latest selected time of the first TCP data stream in the flow state information table is updated to determine whether the first TCP data flow to which the pending IP data packet belongs meets the discarding screening condition or is updated to The receiving time of the pending IP packet.

In an optional embodiment of the present invention, the packet loss filtering condition includes: the unique identifier of the TCP data stream is one of the preset data stream identifiers, and the pre-configured data stream identifier includes at least one of the TCP data streams stored in the queue cache. A unique identifier for the TCP data stream.

In an optional embodiment of the present invention, the data transmission control apparatus 100 may further include a hash index rounding module 122, as shown in FIG.

The hash index rounding module 122 is configured to: rotate the selected hash index in the pre-configured hash table according to the preset time interval and the preset selection rule, and mark the currently selected hash index; The packet loss filtering condition includes: the hash value of the TCP data stream is one of the currently marked hash indexes, wherein the hash value of the TCP data stream is a hash function corresponding to the hash table, and the TCP data stream is used. The unique identifier is hashed to get the value.

In an optional embodiment of the present invention, the data transmission control apparatus 100 further includes a second condition adjustment module.

The second condition adjustment module is configured to determine a buffer queue length in the queue cache according to the second preset period, and adjust the set number according to the buffer queue length, wherein the set number is positively correlated with the buffer queue length.

In an alternative embodiment of the invention, the selection rules include sequential cycle selection.

In an optional embodiment of the present invention, the unique identifier of the TCP data stream is a quintuple of the TCP data stream, or a unique identifier calculated by calculating a quintuple of the TCP data stream according to a preset algorithm.

In an optional embodiment of the present invention, the data transmission control apparatus 100 further includes an associated TCP data flow determining module.

The TCP data flow determining module is configured to determine, according to the quintuple of the IP data packet, the TCP data flow to which the IP data packet belongs before determining whether the TCP data flow to which the IP data packet belongs meets the preset packet loss filtering condition.

In an optional embodiment of the present invention, the packet loss policy includes a RED algorithm.

In an optional embodiment of the present invention, the data processing module 120 is further configured to store the IP data packet into the queue cache when it is determined that the IP data packet is not discarded.

It can be understood that the data transmission control apparatus 100 according to the present invention can correspond to the execution body in the data transmission control method according to the embodiment of the present invention, and the operations and/or functions of the respective modules in the data transmission control apparatus 100 are respectively implemented. The corresponding processes in the data transmission control method of the embodiment of the present invention are implemented. For brevity, details are not described herein again.

FIG. 9 is a schematic block diagram of a network transmission device 200 in accordance with an embodiment of the present invention. As shown in FIG. 9, the network transmission device 200 includes a processor 201, a memory 202 for storing executable program code, and a communication interface 203. The processor 201 operates by reading executable program code stored in the memory 202. A program corresponding to the executable program code for performing the data transfer control method of any of the above embodiments of the present invention. The communication interface 203 is used for communication with an external device. The network transmission device 200 may further include a bus 204 for connecting the processor 201, the memory 202, and the communication interface 203, so that the processor 201, the memory 202, and the communication interface 203 pass through the bus 204. Communicate with each other.

The network transmission device 200 according to an embodiment of the present invention may correspond to an execution subject in a data transmission control method according to an embodiment of the present invention, and operations and/or functions of respective modules in the network transmission device 200 are respectively implemented to implement the present invention. Corresponding processes of the method in the embodiment are not described herein for brevity.

The embodiment of the invention further provides a computer readable storage medium, wherein the readable storage medium stores computer instructions, when the computer instructions are run on a computer, causing the computer to execute the data in any of the above embodiments of the present invention. Transmission control method.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with embodiments of the present invention are generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be from a website site, computer, server or data center Transfer to another website site, computer, server, or data center by wire (eg, coaxial cable, fiber optic, digital subscriber line (DSL), or wireless (eg, infrared, wireless, microwave, etc.). The computer readable storage medium can be any available media that can be accessed by a computer or a data storage device such as a server, data center, or the like that includes one or more available media. The usable medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium such as a Solid State Disk (SSD) or the like.

Claims

A data transmission control method, the method being performed by a network transmission device, wherein the method comprises:

Receiving an IP protocol IP packet to be processed;

Determining whether the first TCP data flow to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition;

And if the first TCP data stream meets the packet loss screening condition, when the IP data packet is to be discarded according to the preset packet loss policy, the to-be-processed IP data packet is discarded;

And if the first TCP data stream does not satisfy the packet loss screening condition, storing the to-be-processed IP data packet in a queue cache of the network transmission device.
The method of claim 1 wherein

Determining whether the first TCP data flow to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, including:

Determining whether the flow state information of the first TCP data stream meets a preset condition;

The flow state information of the first TCP data stream includes a rate characterization value of the first TCP data stream, and the preset condition includes a rate characterization value greater than a set value, and a rate characterization of the first TCP data stream The value is used to characterize the transmission rate of the first TCP data stream.
The method according to claim 2, wherein the rate characterization value of the first TCP data stream is an IP data packet belonging to the first TCP data stream stored in the queue buffer within a preset duration The amount of data.
The method according to claim 2 or 3, wherein the flow state information of the first TCP data stream further includes a time when the first TCP data stream is last selected, and the preset condition further includes selecting The time difference is greater than the first set time length;

The first TCP data stream is selected to be that the first TCP data stream meets the packet loss screening condition, and the selected time difference of the first TCP data stream is the receiving time of the to-be-processed IP data packet. The difference in time when the first TCP data stream was last selected.
A method according to any one of claims 2 or 3, wherein

The flow state information of the first TCP data stream further includes a duration of the first TCP data stream being continuously selected, and the preset condition further includes that the duration of the TCP data stream being continuously selected is less than the second set duration;

The first TCP data stream is selected to be that the first TCP data stream satisfies the packet loss screening condition, and the first TCP data stream is continuously selected for the duration of the pending IP data packet. The length of time between the reception time and the earliest selected time that the first TCP data stream is continuously selected before the reception time of the IP packet to be processed.
The method according to any one of claims 2 to 5, wherein the method further comprises:

Determining a cache queue length in the queue cache according to a first preset period;

And adjusting the set value according to the buffer queue length, wherein the set value is negatively correlated with the buffer queue length.
The method according to any one of claims 2 to 6, wherein the determining whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition includes:

And querying the flow state information table according to the unique identifier of the first TCP data flow to which the to-be-processed IP data packet belongs, to obtain flow state information of the first TCP data flow, where the flow state information table is used for the TCP data flow. Uniquely identified as an index to store flow state information for each TCP data stream;

Determining, according to the flow state information of the first TCP data stream, whether the first TCP data flow satisfies the packet loss screening condition.
The method according to claim 7, wherein if the flow state information of the first TCP data stream includes a time when the first TCP data stream is last selected, determining that the to-be-processed IP data packet belongs to After the first TCP data stream meets the preset packet loss screening condition, the method further includes:

Updating a last selected time of the first TCP data stream in the flow state information table to a time for determining that the first TCP data flow to which the to-be-processed IP data packet belongs meets the discard filter condition or updating to The receiving time of the IP packet to be processed.
The method according to claim 1, wherein the packet loss filtering condition comprises: the unique identifier of the TCP data stream is one of preset data stream identifiers, and the pre-configured data stream identifier includes the queue cache. A unique identifier of at least one TCP data stream in the stored TCP data stream.
The method of claim 1 further comprising:

According to the preset time interval and the preset selection rule, the hash index of the set number in the pre-configured hash table is rotated, and the currently selected hash index is marked;

The packet loss filtering condition includes: the hash value of the TCP data stream is one of the hash indexes of the current label, wherein the hash value of the TCP data stream is a hash function corresponding to the hash table, and is used for TCP. The unique identifier of the data stream is hashed to get the value.
The method of claim 10, wherein the method further comprises:

Determining a cache queue length in the queue cache according to a second preset period;

And adjusting the set number according to the length of the buffer queue, wherein the set number is positively correlated with the length of the buffer queue.
The method of claim 10 wherein said selection rules comprise sequential loop selection.
The method according to any one of claims 7 to 12, wherein the unique identifier of the TCP data stream is a quintuple of the TCP data stream, or the quintuple of the TCP data stream is calculated according to a preset algorithm. Unique identifier.
The method according to any one of claims 1 to 13, wherein before the determining whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, the method further includes:

Determining, according to the quintuple of the to-be-processed IP data packet, the first TCP data stream to which the IP data packet belongs.
The method according to any one of claims 1 to 14, wherein the packet loss policy comprises a random early detection RED algorithm.
A data transmission control device, characterized in that the device comprises:

a data receiving module, configured to receive an Internet Protocol IP data packet to be processed;

a data processing module, configured to determine whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, and if the first TCP data stream meets the packet loss screening condition, When the preset packet loss policy determines that the to-be-processed IP data packet is to be discarded, the to-be-processed IP data packet is discarded, and if the first TCP data flow does not satisfy the packet loss screening condition, the to-be-processed The processing IP data packet is stored in a queue buffer of the network transmission device.
The device according to claim 16, wherein the data processing module is configured to: when determining whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition, specifically:

Determining whether the flow state information of the first TCP data stream meets a preset condition;

The flow state information of the first TCP data stream includes a rate characterization value of the first TCP data stream, and the preset condition includes a rate characterization value greater than a set value, and a rate characterization of the first TCP data stream The value is used to characterize the transmission rate of the first TCP data stream.
The apparatus according to claim 17, wherein the rate characterization value of the first TCP data stream is an IP data packet belonging to the first TCP data stream stored in the queue buffer within a preset duration The amount of data.
The device according to claim 17 or 18, wherein the flow state information of the first TCP data stream further includes a time when the first TCP data stream is last selected, and the preset condition further includes selecting The time difference is greater than the first set time length;

The first TCP data stream is selected to be that the first TCP data stream meets the packet loss screening condition, and the selected time difference of the first TCP data stream is the receiving time of the to-be-processed IP data packet. The difference in time when the first TCP data stream was last selected.
Device according to claim 17 or 18, characterized in that

The flow state information of the first TCP data stream further includes a duration of the first TCP data stream being continuously selected, and the preset condition further includes that the duration of the TCP data stream being continuously selected is less than the second set duration;

The first TCP data stream is selected to be that the first TCP data stream satisfies the packet loss screening condition, and the first TCP data stream is continuously selected for the duration of the pending IP data packet. The length of time between the reception time and the earliest selected time that the first TCP data stream is continuously selected before the reception time of the IP packet to be processed.
The device according to any one of claims 17 to 20, wherein the data processing module determines whether the first TCP data stream to which the to-be-processed IP data packet belongs meets a preset packet loss screening condition Specifically for:

And querying the flow state information table according to the unique identifier of the first TCP data flow to which the to-be-processed IP data packet belongs, to obtain flow state information of the first TCP data flow, where the flow state information table is used for the TCP data flow. Uniquely identified as an index to store flow state information for each TCP data stream;

Determining, according to the flow state information of the first TCP data stream, whether the first TCP data flow satisfies the packet loss screening condition.
The apparatus according to claim 16, wherein the packet loss filtering condition comprises: the unique identifier of the TCP data stream is one of preset data stream identifiers, and the pre-configured data stream identifier includes the queue cache. A unique identifier of at least one TCP data stream in the stored TCP data stream.
The device according to claim 16, wherein the device further comprises:

The hash index rounding module is configured to: rotate the selected hash index in the pre-configured hash table according to the preset time interval and the preset selection rule, and mark the currently selected hash index;

The packet loss filtering condition includes: the hash value of the TCP data stream is one of the hash indexes of the current label, wherein the hash value of the TCP data stream is a hash function corresponding to the hash table, and the TCP data is used. The unique identifier of the stream is hashed to get the value.
A network transmission device, characterized in that the device comprises a memory and a processor;

The memory is for storing executable computer program code;

The processor is configured to read the computer program code to execute a computer program corresponding to the computer program code for performing the data transfer control method according to any one of claims 1 to 15.
A computer readable storage medium, wherein the readable storage medium stores computer instructions that, when executed on a computer, cause the computer to perform the method of any one of claims 1 to Data transmission control method.