WO2020087523A1

WO2020087523A1 - Network communication method and apparatus, and electronic device

Info

Publication number: WO2020087523A1
Application number: PCT/CN2018/113789
Authority: WO
Inventors: 马可; 欧阳晋
Original assignee: 阿里巴巴集团控股有限公司
Priority date: 2018-11-02
Filing date: 2018-11-02
Publication date: 2020-05-07

Abstract

A network communication method and apparatus, an electronic device and a storage device. The method comprises: classifying, before data packets to be sent enter sending queues, the data packets by means of a filter according to a network flow classification rule and setting corresponding network flow class identifiers, the network flow classification rule being used for matching the features of the data packets (S201); and selecting, according to the network flow class identifiers, sending queues to forward the data packets, wherein the filter is used for setting corresponding network flow class identifiers according to a matching relationship between the data packets and the network flow classification rule (S202). The method uses network flow class identifiers for multi-queue sending to implement flow control, thereby increasing the number of network flow classes sent by means of multiple queues.

Description

Network communication method, device and electronic equipment

Technical field

The present application relates to the field of network communication, and in particular to a network communication method, device, electronic device, and storage device. The present application also relates to a second network communication method and device.

Background technique

Linux is an open source software. Because of its flexibility and scalability, Linux is in line with the development trend of cloud computing as a centralized service platform that provides scalable shared computing capabilities. It is the basic software widely used in the cloud computing industry. The computing power provided by cloud computing, including processors, memory, storage, and network interfaces, needs to process massive amounts of data, so the flow control scheme and throughput of network packets are particularly important.

The network flow control scheme of the Linux platform is generally based on TC (Traffic Control, or flow controller) for kernel flow control. The TC module supports the classification of network packets according to the socket type identification, and the available bandwidth of a specific network flow is limited by the token bucket technology To achieve the purpose of flow control. The TC module uses a global spin lock (spin_lock), which allows only one processor core to process one network packet at any time. To improve the concurrency and throughput of packet processing, the existing TC technology supports multi-queue network flow control The solution, specifically, implements the network packet-based socket priority through the Linux mq_prio module to distribute network packets to multiple network card sending and receiving queues, thereby allowing multiple processor cores to process network packets located on multiple queues, respectively. Among them, the socket priority refers to the TOS (Type of Service) priority defined in the IP protocol, a total of 8 levels, which improves the concurrency of packet processing and improves the packet throughput. The packet throughput is PPS ( Packet, Second, or the number of network packets processed per second).

The existing Linux TC network flow control technology, although supporting multiple queues, can only use socket priority to classify network flows. Since the protocol limit supports only 8 levels, this greatly limits the ability of network flow classification. Because the network flow cannot be categorized flexibly according to the category identification of the network flow, the network packet cannot be distributed to multiple sending and receiving queues according to the network flow, resulting in a bottleneck in network packet forwarding performance in practical applications. For example, when the existing Linux TC technology is applied to multiple network application bandwidth isolation scenarios, there is a performance bottleneck in network processing throughput. On commercial multi-core processors (such as Intel Xeon CPU E5-2630 2.30GHz), the throughput cannot exceed 600K PPS, that is, the number of network data packets processed per second does not exceed 600,000 packets. Therefore, in the case of 10Gbps (or higher transmission rate) network cards and network packet sizes less than 2K bytes, only part of the available bandwidth can be used. For example, a network packet with a packet length of 512 bytes can only use 2G bandwidth, and the utilization rate is 20%, and expanding the number of processor cores cannot solve this performance bottleneck.

Summary of the invention

This application provides a method for multi-queue network flow control, to solve the existing problems in the existing Linux TC network flow control technology, because the network flow cannot be flexibly classified according to the network flow type identification, resulting in practical application. There is a bottleneck in network packet forwarding performance.

A network communication method provided by this application includes:

Before the data packet to be sent enters the sending queue, the data packet is classified according to a network flow classification rule through a filter and a corresponding network flow class identifier is set, and the network flow classification rule is used to match the characteristics of the data packet;

Select a sending queue to forward the data packet according to the network flow category identifier;

Wherein, the filter is used to set a corresponding network flow category identifier based on the matching relationship between the data packet and the network flow classification rule.

Optionally, the sending queue is a serial queue with consecutive serial numbers associated with queuing rules; wherein, the category queue is identified by a master serial number and a slave serial number, and the master serial number of the category queue is used for association The network device; the queuing rule is associated with the filter, and is identified by a master sequence number and a slave sequence number, and the master sequence number of the queuing rule takes a value from the slave sequence number of the category queue.

Optionally, the characteristics of the data packet include at least one of the following: IP quintuple information and socket information.

Optionally, the filter is a filter based on Linux flow control;

Correspondingly, the passing filter classifies the data packet according to the network flow classification rule and sets the corresponding network flow class identifier, including:

Classify the data packet according to the network flow classification rules according to the IP header characteristics of the data packet through a filter based on Linux flow control;

Through a filter based on Linux flow control, according to the matching relationship between the IP header characteristics of the data packet and the network flow classification rules, the network category identifier of the data packet is set;

The IP header characteristic of the data packet includes at least any of the following information: source IP, destination IP, source socket port, destination socket port, and protocol type.

Optionally, the selecting a sending queue to forward the data packet according to the network flow category identifier includes:

Through the sending queue scheduling module introduced in the Linux kernel, the sending queue is selected according to the network flow category identifier of the filter associated with the queuing rule.

Optional, including:

An extended command of a Linux control flow command is provided for associating the network flow category identifier with the queuing rule.

Optionally, the selecting a sending queue according to the network flow category identifier of the filter associated with the queuing rule includes:

Determine the queuing rule corresponding to the network flow type identifier according to the network flow type identifier of the data packet;

According to the main sequence number of the queuing rule, a hash table of the main sequence number of the queuing rule to the sending queue number is searched to determine the sending queue number.

Optional, also includes:

The dequeuing order of the data packets is determined according to the priority of the network flow corresponding to the queuing rule associated with the selected send queue, and the queuing rule is used to sort the data packets of the send queue.

Optional, also includes:

Each sending queue sends the data packets in sequence according to the dequeuing order of the data packets entering the queue.

This application also provides a network communication method, including:

Classify network packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is a queuing rule associated with the receiving end Filter

Based on the matching relationship between the data packet and the network flow classification rule, setting the network flow category identifier of the data packet;

According to the network flow type identifier of the data packet, the data packet is forwarded according to the queuing rule of the receiving end.

Optionally, the network flow classification rules include information used to match the characteristics of the packet header; the characteristics of the packet header include at least any of the following information: source IP address, destination IP address, and the corresponding end of the packet receiving end Socket and protocol.

Optionally, the forwarding the data packet according to the queuing rule of the receiving end according to the network flow type identifier of the data packet includes:

Redirect the data packet to an intermediate function module through a filter associated with the queuing rule of the receiving end;

According to the network flow type identifier, different sending queues of the intermediate function module are selected to forward the data packet.

Optional, also includes:

Use the receiving thread reuse technique to reuse the receiving thread context.

Optionally, the receiving thread is a thread for processing the receiving of the data packet at the receiving end; correspondingly, the reuse of the receiving thread context using the receiving thread reuse technology includes:

Directly use the receiving thread to forward the data packet to an intermediate function module according to the queuing rule of the receiving end; the intermediate function module is used to receive the redirected data packet and the network flow corresponding to the queuing rule The priority is sent back to the network device before being redirected.

In addition, the present application also provides a network communication device, including:

The identification unit is used to classify the data packets according to the network flow classification rules and set the corresponding network flow class identification through the filter before the data packets to be sent into the sending queue, and the network flow classification rules are used to match the data Characteristics of the package;

A forwarding unit, configured to select a sending queue to forward the data packet according to the network flow type identifier;

The network flow classification unit is used to classify the network data packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is associated A filter of the queuing rules to the receiving end;

An identification unit, configured to set a network flow type identification of the data packet based on the matching relationship between the data packet and the network flow classification rule;

The forwarding unit is configured to forward the data packet according to the queuing rule of the receiving end according to the network flow type identifier of the data packet.

In addition, this application also provides an electronic device, including:

Memory, and processor;

The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions:

In addition, this application also provides an electronic device, including:

Memory, and processor;

In addition, the present application also provides a storage device that stores instructions that can be loaded by the processor and perform the following steps:

Compared with the prior art, this application has the following advantages:

The method, apparatus, electronic device and storage device for network communication provided by the present application determine the sending queue by using the network flow type identifier. Since the available space of the network flow type identifier is much larger than the number of socket priorities, the pass rate is greatly improved. The number of network flow classifications sent by multiple queues, thus solving the performance bottleneck problem of packet processing in the existing network flow control technology because it does not support flexible classification of network flows. Further, the network flow control module is associated with multiple queues to implement support for network flow control based on network flow category identification.

BRIEF DESCRIPTION

1 is a schematic diagram of an actual deployment environment of a network communication method provided by an embodiment of the present application;

2 is a processing flowchart of a network communication method provided by an embodiment of the present application;

3 is a schematic diagram of sending queue grouping of a network communication method according to an embodiment of the present application;

4 is a schematic diagram of data packet flow of a network communication method according to an embodiment of the present application;

5 is a schematic diagram of a network environment actually deployed in a second network communication method provided by this application;

6 is a processing flowchart of a second embodiment of a network communication method provided by this application;

7 is a schematic diagram of a data packet flow direction of a second embodiment of a network communication method provided by this application;

8 is a schematic diagram of an embodiment of a network communication device provided by this application;

9 is a schematic diagram of an embodiment of a second network communication device provided by this application;

10 is a schematic diagram of an embodiment of an electronic device provided by this application.

detailed description

In the following description, many specific details are set forth in order to fully understand the application. However, this application can be implemented in many other ways than those described here. Those skilled in the art can make similar promotion without violating the connotation of this application, so this application is not limited by the specific implementation disclosed below.

This application provides a network communication method, device, electronic equipment, and storage equipment. The present application also provides a second network communication method, device, electronic device, and storage device. Detailed descriptions will be made one by one in the following embodiments.

The first embodiment of the present application provides a network communication method. The following describes an embodiment of a network communication method provided by the present application with reference to FIGS. 1 to 4.

In network communication, an analogous water flow is used to describe the transmission of data packets in the network. A series of data packets sent from a specific source to a specific destination is called a network flow (or network data flow, or simply stream). For example, a data packet with the same IP quintuple information transmitted unidirectionally between a source IP address and a destination IP address is defined as a network flow. The so-called IP quintuple information includes: source IP address, destination IP address, source port, destination port, and protocol number. Different identifiers can be used to mark different network flows. The identifier used to mark network flows is called a network flow category identifier (also called a category identifier, or classid).

Taking different forwarding strategies for different network flows to provide differentiated services for different services and improve bandwidth utilization is an important means of cloud computing and the Internet industry to save network resources. Therefore, the flow control scheme and throughput of network flows appear especially important. The Linux platform commonly used in cloud computing and the Internet industry is based on Linux TC (Traffic Control, or flow controller) for kernel flow control. The forwarding process of network data packets in Linux is received from the entrance (input network card), find the route, determine that it is sent to the local machine, and forward it to the upper layer protocol processing, such as TCP, if it is determined that forwarding is required, then exit (output NIC) forwarded out, the Linux native TC function network flow control is controlled at the exit, because the TC module uses a global spin lock (spin_lock), only one processor core is allowed to process one network packet at any time, in order to improve data The concurrency and throughput of packet processing, the existing TC technology provides a multi-queue network flow control solution, supports the classification of network packets according to the socket category identifier, and uses the token bucket technology to limit the maximum and minimum bandwidth available for a specific network flow So as to achieve the purpose of flow control.

The network communication method provided in the embodiments of the present application is generally deployed at the sending end of a data packet. The sending end may be a network device, for example, may be a network interface board (Network Interface Card, or NIC), and a network interface board It is called a network card; the sending end may also be a virtual network device similar to the network device, and sending data packets through the virtual network device is similar to sending data packets through the physical port of the network device. For example, the intermediate function module IFB (Intermediate Functinal Block). The network communication method supports flow control of the network flow marked by the network flow type identification based on the network flow type identification. Specifically, it distributes the data packets to multiple sending queues according to the network flow category identifier, so as to categorize the network flow flexibly. Because the available space of the network stream category identifier is large, the number of category identifiers is much larger than the number of socket priorities. Greatly improve the ability of network traffic classification, so that the network PPS can increase linearly with the number of processor cores, for example, it can reach 14.4M / s on the configured 24-core processor, thereby improving network bandwidth utilization and data center servers Resource utilization.

The following uses the embodiment of the network communication method based on Linux as an example to illustrate the network communication method provided by the present application.

Before explaining this embodiment, the basic concepts of Linux TC flow control: queuing rules, categories, and filters. Each network card is associated with a queuing rule. When the kernel forwards the data packet from the network card, it will add the data packet to the queue according to the corresponding queuing rule of the network card. To achieve powerful flow control, complex queues are required, that is, queues with class queuing rules. Classifiers need to be used to filter packets into different classes, and then according to different queuing rules, according to Send packets in sortable queues in different orders to achieve flow control.

FIG. 1 shows a network environment actually deployed by the network communication method, including: the network application 101 constructs a network data packet and sends it to the TCP / IP layer 102 through the socket mechanism; the TCP / IP layer 102 adds the network data packet After the TCP / IP protocol header information is sent to the net_clscGroup packet marking module 103; the net_clscGroup packet marking module 103 adds a category identifier to the network packet with the TCP / IP protocol header information added and sends it to the mq_cls sending queue scheduling module 104; The mq_cls send queue scheduling module 104 selects the send queue to forward the network data packet with added TCP / IP protocol header information to the outgoing interface of the network card through the network card driver 105, for example, the ethernet interface. Among them, the sending queue can be selected in the mq_cls sending queue scheduling module 104 to control the inbound and outbound queues of the network data packets containing TCP / IP protocol header information to achieve network flow control. The flow control of the network flow can be controlled at the exit, including shaping, scheduling, and dropping. The Linux kernel forwards the data packet from the network card to control the flow at the exit. Among them, the so-called shaping, when the traffic is limited, its transmission rate is controlled below a certain value, the limit value can be much smaller than the effective bandwidth, which can smooth out burst data traffic and make the network more stable; the so-called scheduling, through scheduling For data packet transmission, the bandwidth can be allocated according to priority within the bandwidth range; if the so-called dropping exceeds a certain set bandwidth, the data packet is discarded.

The net_clscgroup is a subsystem for classifying and marking data packets. The net_clscgroup marking of data packets can be used as the network flow category identifier, and the Linux flow control TC (traffic control) can recognize the data after net_clscgroup marking package.

FIG. 2 is a processing flowchart of a network communication method provided by an embodiment of the present application.

The network communication method shown in FIG. 2 includes steps S201 to S202.

Step S201: Before the data packet to be sent enters the sending queue, classify the data packet according to a network flow classification rule through a filter and set a corresponding network flow class identifier. The network flow classification rule is used to match the feature.

The network flow classification rule is used to match the characteristics of the data packet; wherein, the characteristics of the data packet include at least one of the following: IP quintuple information and socket information. For example, if specific IP quintuple information is set as the network flow classification rule, the data packet matching the specific IP quintuple information is classified into a specific network flow.

This step is to set a corresponding network flow category identifier on the data packets matching the network flow classification rules described above.

In the embodiment of the present application, the data packet is classified according to a network flow classification rule through a filter and a corresponding network flow type identifier is set. For example, the network flow classification rule of the filter contains information used to match the characteristics of the data packet to the network flow of the destination IP address 192.168.0.1 and destination port number 80, and the destination IP address 192.168.0.1 and destination port number 80 The category ID is 10: 1; the filter will filter out the packets matching the destination IP address 192.168.0.1 and the destination port number 80, and set the network flow category ID 10 for the packets matching the network flow classification rules described above :1.

In the embodiment of the present application, the filter may be a filter based on Linux flow control;

Among them, the so-called protocol types include four-layer protocol and / or seven-layer protocol, such as UDP, HTTP, etc.

In the embodiment of the present application, the filter may be a Linux cgroup filter. The so-called cgroup filter is a filter provided by the subsystem net_clscgroup in Linux cgroup (Linux control group, or Linux control group). The so-called Linux cgroup refers to the various subsystems provided by the Linux kernel for resource management, where net_cls cgroup is a subsystem that classifies and marks data packets, and the net_cls cgroup's marking of data packets can be used as the network flow category identification, which Linux can recognize Packets marked by each specific cgroup filter configured in the net_cls cgroup subsystem. For example, create and configure a specific net_cls cgroup, and then use the TC filter command to classify the data packets marked by the specific net_cls cgroup into the corresponding category of the network flow class identifier.

Step S102: Select a sending queue to forward the data packet according to the network flow type identifier.

The sending queue is a serial queue with consecutive serial numbers associated with queuing rules; wherein, the category queue is identified using a master serial number and a slave serial number, and the master serial number of the category queue is used to associate network devices; The queuing rule is associated with the filter, and is identified by a master sequence number and a slave sequence number. The master sequence number of the queuing rule takes a value from the slave sequence number of the category queue. Specifically, in the embodiment of the present application, the data packets are managed by the queuing rules through the class in Linunx TC, and the data packets are inserted into the sending queue. Since the sending queue is grouped, it is called a class queue.

The queuing rules in this embodiment are class queuing rules (ie classful queueing disciplines), and classless queuing rules pass through the queues in a first-in, first-out manner for data packets, except for caching packets that cannot be processed by the network interface No other processing is performed on packets entering the queue. The so-called class queuing rules include CBQ, HTB and Prior Qdisc, in which CBQ (class based queueing, or class based queuing) implements a rich connection sharing class structure, which has both bandwidth limiting and bandwidth priority Level management capability, bandwidth limitation is done by calculating the idle time of the connection. The standard for calculating idle time is the frequency of data packet dequeue events and the bandwidth of the lower layer connection (data link layer); HTB (hierarchy token bucket, or layered order (Bucket), to achieve bandwidth limitation through TBF (token bucket filter, or token bucket filter), allowing certain categories to break through the upper limit of bandwidth and occupy the bandwidth of other categories; rules such as Prio Qdisc will be high priority The low-priority data packets are sent after the data packets are sent, that is, the data packets of different priorities are dequeued in sequence. The queuing rules can contain some categories, and different categories can contain deeper queuing rules. Through these subdivision queuing rules, you can queue packets for the incoming queue, and set the network data by setting the dequeuing order of various types of packets The priority of the traffic. The so-called filter is used to classify data packets and determine the queuing rules for data packets to enter the queue. Whenever a data packet enters a sub-category category, it needs to be classified. When using filter classification, the kernel calls all the filters attached to this class (class until a decision is returned. If no decision is returned, further processing is performed, and the processing method is related to the queuing rules.

Specifically to this embodiment, the queuing rules, classes, and filters are all identified by an ID. The ID is composed of a master serial number and a slave serial number, and the two numbers are separated by a colon. Among them, the main sequence number of the queuing rule is called a handle, which uses an expression like "1:". Use the serial number as the namespace for the class. Classes in the same queuing rule share the master sequence number of the queuing rule, but each class has its own slave sequence number, called the class identifier (classid). The class identifier is only related to the parent queuing rules, and not to the parent class.

This step is to select a sending queue to forward the data packet marked by the network flow type identifier according to the network flow type identifier set in step S201.

In the embodiment of the present application, the functional modules of Linux TC are expanded to include the following two aspects:

On the one hand, the Linux kernel introduces a send queue scheduling module. The sending queue scheduling module introduced by the Linux kernel selects the sending queue according to the network flow category identifier of the filter associated with the queuing rule.

On the other hand, it provides extended commands for Linux control flow commands. The extended command is used to associate the network flow category identifier with the queuing rule.

In the embodiment of the present application, the network flow classification rule corresponding to the filter is used to perform matching processing on the data packet, and the filter is attached to the class. Network devices that forward packets (such as network cards) need to create root queue queuing rules, and establish classes and subclasses under the root queue queuing rules. Each class has only one parent class, and each class can have multiple subclasses. These classes Form a tree, called TC tree. When the data packets are matched with the network flow classification rules, they are layered according to the TC tree. According to the processing logic of each layer, from the root to the leaf nodes, they are really queued up to the real queue to forward the data packets through the network device. For example, the TC tree included in the schematic diagram of the transmit queue grouping shown in FIG. 3, the TC tree includes:

Root queue queuing rule (Root QDISC) 301, the main sequence number of the root queue queuing rule is 8001:

The root queue queuing rule 301 includes two classes: class3202 and class303, the class ID of class302 is 8001: ab51, and the class ID of class303 is 8001: ab52;

The leaf queuing rule HTB304 of the class 302;

The leaf queuing rule HTB305 of the class 303.

A specific method of the embodiment of the present application includes: associating the filter corresponding to the network flow category identifier to the queuing rule through the extended command of the Linux control flow command; and then transmitting the network flow category identifier of the filter associated with the queuing rule through the sending queue scheduling module Select the send queue. The introduction of the sending queue scheduling module and the command expansion are described below.

The queue allocation method based on network flow class identification needs to define a flexible and reusable interface. The mq_cls send queue scheduling module is used to implement the socket class identification setting. Use the SO_PRIORITY option to set the class ID in the socket-> sk_classid field for the classified data packets The class ID is the network flow class identifier, and supports the selection of the sending end queue according to the class ID. The specific mq_cls module includes the following:

mq_cls supports associating a TC qdisc (queueing discrimination) to a continuous set of sending queues. The major sequence number of the class ID of the queuing rule is used as the class ID of mq_cls to select a certain A group of send queues. Extend Linux's existing TC qdisc command and add configuration commands for the mq_cls send queue scheduling module. The expanded new command format is:

tc qdisc (add | change | del) dev (dev) root [handle major:] mqcls [numtc tcs] [classes, class0, class1…] [queues, count1 @ offset1count2 @ offset2…]

Among them, the meaning of each command parameter is as follows:

qdisc indicates commands related to queue rules;

(add | change | del) Commands to add, modify and delete queue rules;

dev specifies the keywords of the device;

(dev) device name;

root operation root queue rules;

[handle major:] [queue rule identifies the main sequence number of the queue rule:];

mqcls queue rule identification name;

[numtc tcs] specifies the number of subcategories of mqcls;

[classes class0 class1 ...] Specify the subclass serial number (from serial number) of mqcls. Class0 refers to the serial number of the first subcategory;

[queuescount1 @ offset1count2 @ offset2…] specifies the queue group corresponding to the sub-category of mqcls, and count1 specifies the number of queues in the queue group corresponding to the first sub-category. Offset1 specifies the zero-based offset of the first queue in the queue group corresponding to the first subcategory.

An example is as follows: the transmission queue grouping diagram shown in FIG. 3 is used to explain the transmission queue grouping of the network device eth0. Use TXQ0 and TXQ1 as a group of transmission queues 306, and TXQ2 and TXQ23 as a group of transmission queues 307;

class 302 inserts the data packet with the network flow category identification as ab51 into the sending queue 306 according to the HTB queuing rules;

The class 303 inserts the data packet with the network flow class identifier of ab52 into the sending queue 307 according to the HTB queuing rules.

The HTB (Hieararchical Token Bucket) qdisc is mounted on the TXQ0 and TXQ1 configuration queues as follows:

Command 1: tc qdisc add dev deveth0 root mqcls numtc 1 classes ab51 queues 2 @ 0

Command 2: tc qdisc add dev eth0 parent 8001: ab51 handle ab51: 0 htb

The command 1 divides the transmission queues TXQ0 and TXQ1 of the network device eth0 into one packet, and the command 2 mounts the HTB queuing rule to this packet.

The configuration commands for mounting HTB qdisc to TXQ2 to TXQ23 are as follows:

Command 3: tc qdisc change dev eth0 root mqcls numtc 2 classes classes ab51 ab52 queues 2 @ 02222 @ 2

Command 4: tc qdisc add dev eth0 parent 8001: ab52 handle ab52: 0htb

The command 3 divides the transmission queues TXQ2 and TXQ23 of the network device eth0 into one packet, and the command 4 mounts the HTB queuing rule to this packet.

It should be noted that in this embodiment of the application, the rules for the value of the queuing rule handle (qdisc) and the class ID (class ID) of Linux TC are used. The queuing rules associated with the sending queue group must use the corresponding class The minor serial number of the class handle is used as the major serial number of its root handle. For example, the value of the handle parameter in Command 2 is "ab51: 0", which depends on the value of the class parameter in Command 2 as "ab51". In this way, the cgroup filter associated with the class of Linux TC (filter classified according to the class ID in net_cls cgroup) can classify the data packet as the network flow class identifier of the class class ID and match all Set the packet of the above-mentioned packet.

In the embodiment of the present application, the following process specifically selects the sending queue according to the network flow category identifier of the filter associated with the queuing rule, including:

Determine a queuing rule corresponding to the network flow category identifier;

In actual implementation, in order to realize fast search of the sending queue, a hash table of the handle of the queuing rule added to the sending queue number in the Linux kernel is added to the hash table of the sending queue number, and the sending queue number is quickly searched through the hash table. Specifically, the netdev_pick_tx () function is modified so that the major number in skb's net_cls is used to select txq. The Linux kernel adds a hash table from net_cls major number to txq number, which can be used to quickly find txq, so as to pass Major number selects txq and then its corresponding qdisc.

In addition, in the embodiment of the present application, in the process of selecting the sending queue to forward the data packet, the following process is further included:

According to the priority of the network flow corresponding to the queuing rule associated with the selected sending queue, the dequeuing order of the data packets is determined, and the queuing rule is used to sort the data packets of the sending queue. Each sending queue sends the data packets in sequence according to the dequeuing order of the data packets entering the queue. It implements the use of network flow class identification to determine the sending queue, and associates the network flow control module to multiple queues to achieve flow control support based on network class identification (network class ID).

FIG. 4 shows the flow of forwarding processing of data packets at the sending end in the above network communication method. include:

The network application 401 constructs a data packet;

Sending the data packet to the transport layer / network layer TCP / IP 403 through the network interface layer socket 402;

TCP / IP 403 adds a protocol header to the data packet, such as a port number, an IP address, and a checksum, and sends the data packet with the added protocol header to TXQ Selection 404, which is a sending queue selection module;

TXQ Selection 404 actually deploys mq_cls, which is used to select the transmission queue of network equipment (such as a network card, or network interface card, abbreviated as NIC), and send the data packet to Traffic Control 405 for flow control, and the flow control is shown in the figure The queuing rule is HTB;

The data packet processed by Traffic Control 405 for flow control processing enters the network card through NIC Driver 406, and arrives at the actual sending queue out NIC 407.

Based on the above embodiment of the network communication method, the second embodiment of the present application provides a second network communication method.

The second embodiment of the second network communication method provided by the present application will be described below with reference to FIGS. 5 to 7. Among them, FIG. 5 is a schematic diagram of a network environment in which the network communication method provided by the present application is actually deployed; FIG. 6 is a processing flowchart of a second network communication method provided by the present application; FIG. 7 is a second method provided by the present application Schematic diagram of the forwarding flow of the data packet at the receiving end in the network communication method.

Since this embodiment is based on the embodiment of the network communication method provided in the first embodiment of the present application, its description is brief, and the relevant part may refer to the description of the first embodiment of the present application.

The network communication method shown in FIG. 6 includes steps S601 to S603.

Step S601: Classify the network data packets received by the receiving end according to the network flow classification rules of the filter. The network flow classification rules are used to match the characteristics of the data packet. The filter is associated with the received The filter of the queue rule at the end.

The network communication method provided in the second embodiment of the present application is generally deployed at the receiving end of a data packet transmitted on the network to achieve flow control at the receiving end. The receiving end is a network device, for example, it may be a network interface card (Network Interface Card, or NIC), and the network interface board is also called a network card. Specifically, after the data packet received by the receiving end from the network side passes through the receiving queue of the network card, it is redirected to the transmission path of the IFB (Intermediate Functional Block, or intermediate function module), and flow control is performed when the data packet is forwarded by the IFB.

The so-called IFB simulates network card forwarding data packets and provides flow control functions. Since the flow control module TC is associated with the IFB, the IFB can use the TC to provide flow control functions. IFB does not change the transmission direction of data packets. That is to say, after the data packets of the network card receive queue are redirected to IFB, after being processed by the flow control function of IFB, they are forwarded back to the network card before being redirected to continue the receiving process. For example, whether to send a data packet from a network card or receive a data packet from a network card, redirect to the IFB after the flow control function is processed, and then forward it back to the network card before being redirected by processing in the IFB's dev_queue_xmit function.

The network communication method provided in this embodiment aims to complete the flow control of the receiving end through the network flow type identification. In order to be able to determine the network flow category identification of the data packets arriving at the IFB, a cls_sk filter module (also called cls_sk filter) is introduced. The cls_sk filter module is associated with the queuing rules of the receiving end, provides a cls_sk filter, determines the socket port number of each data packet received by the receiving end, and sets the network flow of the data packet according to the characteristics of the data packet Category identification.

FIG. 5 shows a network environment actually deployed by the network communication method provided by this embodiment, including: data packets are put into a network card receiving queue;

The data packet reaches the receiving end queuing rule 502 through the network card driver 501, and the receiving end queuing rule is also called ingress QDISC;

Query the socket port number of the data packet through the cls_sk filter 503 associated with the receiving end queuing rule 502, determine the network flow type corresponding to the data packet according to the characteristics of the data packet, and set the corresponding network flow to the data packet Category identification

The cls_sk filter 503 redirects the data packet marked with the network flow category identifier to the sending path of the IFB 504, the sending path of the IFB 504 is associated with the mq_cls sending queue scheduling module for selecting the sending queue;

Through the queuing rules of the sending path of IFB504, according to the network flow type identifier of the data packet, the data packet is controlled to be forwarded back to TCP / IP 505 to process the network protocol stack, so as to achieve the purpose of flow control. The TCP / IP Refers to the network protocol stack module;

TCP / IP 505 sends the processed data packet to network application 506 for further processing.

This step is to classify the network data packets received by the receiving end.

In the embodiment of the present application, the data packets received from the network side are classified according to the network flow classification rules of the filter. Specifically, the cls_sk filter is used to query the socket port number of the data packet, and the network flow type corresponding to the data packet is determined according to the characteristics of the data packet. The network flow classification rules are used to match the characteristics of the data packet, and specifically include information used to match the characteristics of the packet header; the characteristics of the data packet header include at least any of the following information: source IP address and destination IP address 3. The socket and protocol corresponding to the receiving end of the data packet. The filter is a filter associated with the queuing rule of the receiving end.

Step S602: Set a network flow category identifier of the data packet based on the matching relationship between the data packet and the network flow classification rule.

This step is to identify the network flow category of the data packet device.

In the embodiment of the present application, after classifying the data packet by using a cls_sk filter, the data packet matching the network flow classification rule of the cls_sk filter is set with a network flow class identifier.

Step S603: According to the network flow type identifier of the data packet, forward the data packet according to the queuing rule of the receiving end.

In this step, the data packet is forwarded to the network protocol stack of the kernel for processing according to the queuing rule of the receiving end, so as to realize the flow control of the receiving end.

In the embodiment of the present application, specifically performing the following operations to realize forwarding the data packet according to the network flow type identifier of the data packet according to the queuing rule of the receiving end includes:

According to the network flow type identifier, the sending queue of the intermediate function module is selected to forward the data packet.

Specifically, the intermediate function module refers to IFB. For example, through the mq_cls send queue scheduling module associated with IFB, select the send queue of IFB, and perform flow control according to IFB's queuing rules, such as HTB.

The embodiment of the present application further includes: using the receiving thread reuse technology to reuse the receiving thread context. Preferably, it is supported to control whether to enable the reuse function of the receiving thread through a switch. The receiving thread is a thread for processing reception of a data packet at the receiving end.

The content in the context of a specific thread is used by other modules, which can be implemented through mechanisms such as memory sharing and message queues, or it can directly reuse the specific thread to complete at least part of the work required by other modules. In the embodiment of the present application, the so-called receiving thread reuse technology refers to using the receiving thread to process other data packet processing tasks than the reception of the data packet, so that the content saved in the context of the receiving end can be directly used for the Other packet processing work. Specifically, the receiving thread is reused in the intermediate function module IFB. For example, directly using the receiving thread to forward the data packet to an intermediate function module (IFB) according to the queuing rules of the receiving end, without using the tasklet mechanism, can avoid the processor core competition that the tasklet may bring. The intermediate function module is configured to receive the redirected data packet, and send the redirected data packet back to the network device before being redirected according to the priority of the network flow corresponding to the queuing rule. When the redirected data packet is redirected by the receiving end, the intermediate function module processes the redirected data packet and sends it back to the receiving end before being redirected.

In addition, the reuse of the receiving thread can also use the receiving thread to directly execute the network protocol stack, which can improve the concurrency of the flow control processing at the receiving end.

7 shows the flow of the receiving end after receiving the data packet from the network side. The data packet is received through a network device (network card), the received data packet is processed through the network communication method, and the processed data packet is sent to a network application for further use. For the flow of data packets in these processes, see FIG. 7 ,include:

NIC 701 is a network card, and the network card has n + 1 receiving queues RXQ0, RXQ1, RXQ2, ..., RXQn; the data packets enter the receiving queue of NIC 701;

The data packet reaches the NIC Driver 702 through the receiving queue of NIC 701, and the NIC Driver is a network card driver;

The data packet arrives at Ingress QDISC 703 by NIC Driver 702. Ingress QDISC is the queuing rule of the receiving end, also known as the receiving end flow control scheduler;

The data packet reaches the cls_sk filter associated with the queuing rule of the receiving end by Ingress QDISC 703, that is, cls_sk / mirred action 704 in the figure, and cls_sk determines the network flow type of the data packet and sets the network flow type identifier for the data packet; The mirred action shown refers to the cls_sk filter performing the redirect action, and the data packet will be redirected to the intermediate function module IFB 706;

Traffic Control 705 is the flow control module associated with IFB 706. According to the flow control module associated with IFB 706, select the transmission queue on the transmission path of IFB 706 (see TXQ Selection in the figure. The transmission queue is a reused reception queue. There are n transmission queues) ), And after performing flow control according to the queuing rules associated with IFB 706 (such as the illustrated HTB), IFB 706 sends back to the receiving queue of the network card before redirection to continue receiving processing, and continuing to receive processing including sending to TCP / IP 707 performs network protocol stack processing, and the illustrated TCP / IP indicates that the network protocol stack is an IP protocol stack;

TCP / IP 707 sends the processed data packets to the network application 708 through the socket layer for further processing or use.

Corresponding to an embodiment of a network communication method provided by the first embodiment of the present application, the present application also provides a first network communication device.

Referring to FIG. 8, it shows a schematic diagram of a first network communication device according to the present application. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple. For related parts, please refer to the corresponding descriptions of the corresponding method embodiments.

This application provides the first network communication device, including:

The identification unit 801 is used to classify the data packet according to the network flow classification rule and set the corresponding network flow class identification through the filter before the data packet to be sent into the sending queue. The network flow classification rule is used to match the Characteristics of the data packet;

A forwarding unit 802, configured to select a sending queue to forward the data packet according to the network flow type identifier;

Optionally, the filter is a filter based on Linux flow control;

Correspondingly, the identification unit 801 is specifically used for:

Optionally, the forwarding unit 802 is specifically used to:

Optionally, the network communication device further includes a command unit for:

An extended command of a Linux control flow command is provided, and the extended command is used to associate the network flow category identifier with the queuing rule.

Optionally, the forwarding unit 802 is specifically used to:

Optionally, the network communication device further includes a dequeuing order determination unit, which is used to:

Corresponding to the embodiment of the second network communication method provided by this application, this application also provides a second network communication device.

Referring to FIG. 9, it shows a schematic diagram of a second network communication device according to the present application. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple. For related parts, please refer to the corresponding descriptions of the corresponding method embodiments.

This application provides a second network communication device, including:

The network flow classification unit 901 is used to classify the network data packets received by the receiving end according to the network flow classification rules of the filter. The network flow classification rules are used to match the characteristics of the data packet. The filter is A filter associated with the queuing rule of the receiving end;

The identification unit 902 is configured to set the network flow type identification of the data packet based on the matching relationship between the data packet and the network flow classification rule;

The forwarding unit 903 is configured to forward the data packet according to the queuing rule of the receiving end according to the network flow type identifier of the data packet.

Optionally, the forwarding unit 903 is specifically used to:

Optionally, the network communication device further includes a receiving thread reuse unit, which is used to:

Use the receiving thread reuse technique to reuse the receiving thread context.

Optionally, the receiving thread is a thread for processing the receiving of data packets at the receiving end; correspondingly, the receiving thread reuse unit is specifically used for:

The present application also provides an embodiment of an electronic device for implementing the network communication method provided in the first embodiment of the present application. Referring to FIG. 10, which shows an electronic device provided by the present embodiment Schematic.

The embodiment of the electronic device provided in this application is described relatively simply, and for related parts, please refer to the corresponding description in the first embodiment of this application.

This application provides an electronic device, including:

Memory 1001, and processor 1002;

The memory 1001 is used to store computer executable instructions, and the processor 1002 is used to execute the computer executable instructions:

Optionally, the filter is a filter based on Linux flow control;

Correspondingly, the processor 1002 is also used to execute the following computer executable instructions:

Optionally, the processor 1002 is also used to execute the following computer-executable instructions:

The present application also provides a second embodiment of an electronic device that is used to implement the network communication method provided by the second embodiment of the present application, and a schematic diagram of the electronic device is similar to FIG. 10.

The embodiment of the electronic device provided in this application is described relatively simply, and for related parts, please refer to the corresponding description in the second embodiment of this application.

This application provides a second type of electronic equipment, including:

Memory, and processor;

Optionally, the processor is also used to execute the following computer-executable instructions:

Use the receiving thread reuse technique to reuse the receiving thread context.

Optionally, the receiving thread is a thread for processing the reception of data packets at the receiving end; correspondingly, the processor is also used to execute the following computer-executable instructions:

The present application also provides an embodiment of a storage device.

The present application provides a storage device, including: a memory and a processor, wherein the memory stores instructions, and the instructions can be loaded by the processor and perform the following steps:

This application also provides a second embodiment of the storage device.

The second storage device provided by the present application includes: a memory and a processor, wherein the memory stores instructions, and the instructions can be loaded by the processor and perform the following steps:

In a typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

The memory may include non-permanent memory, random access memory (RAM) and / or non-volatile memory in computer-readable media, such as read only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

1. Computer-readable media, including permanent and non-permanent, removable and non-removable media, can implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, Magnetic tape cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media can be used to store information that can be accessed by computing devices. As defined in this article, computer-readable media does not include non-transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

2. Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

Although this application is disclosed as above with preferred embodiments, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. The scope of protection shall be subject to the scope defined in the claims of this application.

Claims

A method of network communication, characterized in that it includes:

Before the data packet to be sent enters the sending queue, the data packet is classified according to a network flow classification rule through a filter and a corresponding network flow class identifier is set, and the network flow classification rule is used to match the characteristics of the data packet;

Select a sending queue to forward the data packet according to the network flow category identifier;

Wherein, the filter is used to set a corresponding network flow category identifier based on the matching relationship between the data packet and the network flow classification rule.
The method according to claim 1, wherein the sending queue is a serial queue with consecutive serial numbers associated with queuing rules; wherein, the category queue is identified by a master serial number and a slave serial number, and The main serial number of the category queue is used to associate with the network device; the queuing rule is associated with the filter, and is identified by the main sequence number and the slave sequence number. From the serial number.
The method according to claim 1, wherein the characteristics of the data packet include at least one of the following: IP quintuple information and socket information.
The method according to claim 3, wherein the filter is a filter based on Linux flow control;

Correspondingly, the passing filter classifies the data packet according to the network flow classification rule and sets the corresponding network flow class identifier, including:

Classify the data packet according to the network flow classification rules according to the IP header characteristics of the data packet through a filter based on Linux flow control;

Through a filter based on Linux flow control, according to the matching relationship between the IP header characteristics of the data packet and the network flow classification rules, the network category identifier of the data packet is set;

The IP header characteristic of the data packet includes at least any of the following information: source IP, destination IP, source socket port, destination socket port, and protocol type.
The method according to claim 2, wherein the selecting a sending queue to forward the data packet according to the network flow type identifier includes:

Through the sending queue scheduling module introduced in the Linux kernel, the sending queue is selected according to the network flow category identifier of the filter associated with the queuing rule.
The method of claim 2, further comprising:

An extended command of the Linux control flow command is provided for associating the network flow category identifier with the queuing rule.
The method according to claim 5, wherein the selection of the sending queue according to the network flow category identifier of the filter associated with the queuing rule includes:

Determine the queuing rule corresponding to the network flow type identifier according to the network flow type identifier of the data packet;

According to the main sequence number of the queuing rule, a hash table of the main sequence number of the queuing rule to the sending queue number is searched to determine the sending queue number.
The method according to claim 1, further comprising:

The dequeuing order of the data packets is determined according to the priority of the network flow corresponding to the queuing rule associated with the selected send queue, and the queuing rule is used to sort the data packets of the send queue.
The method according to claim 8, wherein the selecting a sending queue to forward the data packet according to the network flow type identifier comprises:

Each sending queue sends the data packets in sequence according to the dequeuing order of the data packets entering the queue.
A method of network communication, characterized in that it includes:

Classify network packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is a queuing rule associated with the receiving end Filter

Based on the matching relationship between the data packet and the network flow classification rule, setting the network flow category identifier of the data packet;

According to the network flow type identifier of the data packet, the data packet is forwarded according to the queuing rule of the receiving end.
The method according to claim 10, wherein the network flow classification rule contains information used to match the characteristics of the packet header; the characteristics of the packet header include at least any of the following information: source IP address, Destination IP address, socket and protocol corresponding to the receiving end of the data packet.
The method according to claim 10, wherein the forwarding the data packet according to the queuing rule of the receiving end according to the network flow type identifier of the data packet includes:

Redirect the data packet to an intermediate function module through a filter associated with the queuing rule of the receiving end;

According to the network flow type identifier, different sending queues of the intermediate function module are selected to forward the data packet.
The method according to claim 10, further comprising:

Use the receiving thread reuse technique to reuse the receiving thread context.
The method according to claim 13, wherein the receiving thread is a thread for processing the receiving of the data packet at the receiving end; correspondingly, the reuse of the receiving thread context using the receiving thread reuse technology includes:

Directly use the receiving thread to forward the data packet to an intermediate function module according to the queuing rule of the receiving end; the intermediate function module is used to receive the redirected data packet and the network flow corresponding to the queuing rule The priority is sent back to the network device before being redirected.
A network communication device, characterized in that it includes:

The identification unit is used to classify the data packets according to the network flow classification rules and set the corresponding network flow class identification through the filter before the data packets to be sent into the sending queue, and the network flow classification rules are used to match the data Characteristics of the package;

A forwarding unit, configured to select a sending queue to forward the data packet according to the network flow type identifier;

Wherein, the filter is used to set a corresponding network flow category identifier based on the matching relationship between the data packet and the network flow classification rule.
A network communication device, characterized in that it includes:

The network flow classification unit is used to classify the network data packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is associated A filter of the queuing rules to the receiving end;

An identification unit, configured to set a network flow type identification of the data packet based on the matching relationship between the data packet and the network flow classification rule;

The forwarding unit is configured to forward the data packet according to the queuing rule of the receiving end according to the network flow type identifier of the data packet.
An electronic device, characterized in that it includes:

Memory, and processor;

The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions:

Before the data packet to be sent enters the sending queue, the data packet is classified according to a network flow classification rule through a filter and a corresponding network flow class identifier is set, and the network flow classification rule is used to match the characteristics of the data packet;

Select a sending queue to forward the data packet according to the network flow category identifier;

Wherein, the filter is used to set a corresponding network flow category identifier based on the matching relationship between the data packet and the network flow classification rule.
An electronic device, characterized in that it includes:

Memory, and processor;

The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions:

Classify network packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is a queuing rule associated with the receiving end Filter

Based on the matching relationship between the data packet and the network flow classification rule, setting the network flow category identifier of the data packet;

According to the network flow type identifier of the data packet, the data packet is forwarded according to the queuing rule of the receiving end.
A storage device is characterized by storing instructions, which can be loaded by a processor and perform the following steps:

Before the data packet to be sent enters the sending queue, the data packet is classified according to a network flow classification rule through a filter and a corresponding network flow class identifier is set, and the network flow classification rule is used to match the characteristics of the data packet;

Select a sending queue to forward the data packet according to the network flow category identifier;

Wherein, the filter is used to set a corresponding network flow category identifier based on the matching relationship between the data packet and the network flow classification rule.
A storage device is characterized by storing instructions, which can be loaded by a processor and perform the following steps:

Classify network packets received by the receiving end according to the network flow classification rules of the filter, the network flow classification rules are used to match the characteristics of the data packet, and the filter is a queuing rule associated with the receiving end Filter

Based on the matching relationship between the data packet and the network flow classification rule, setting the network flow category identifier of the data packet;

According to the network flow type identifier of the data packet, the data packet is forwarded according to the queuing rule of the receiving end.