CN114666681B - A stateful in-band network telemetry method and system - Google Patents

Abstract

The present invention discloses a stateful in-band network telemetry method and system (Stateful In-band Network Telemetry, SF-INT), including: defining an instruction header of SF-INT, which includes a new field unique to SF-INT; network nodes use their own registers to store INT state information in a distributed manner, and process and forward it in combination with the arrived INT header information, and the size of the metadata stack of its normal forwarding is unchanged; a network monitoring system (NMS) calculates multiple INT headers received to obtain a complete network telemetry result; the network node can instantly and seamlessly switch to the current INT protocol in an emergency, and the NMS can correctly identify this situation and make corresponding calculations. Compared with the stateless current INT protocol, the stateful INT of the present invention can not only meet the network telemetry requirements, but also save Internet bandwidth by significantly reducing the INT metadata carried by the service packet.

Description

A stateful in-band network telemetry method and system

Technical Field

The present invention relates to the field of network communication technology, and in particular to a stateful in-band network telemetry (INT) method and system, referred to as SF-INT.

Background technique

In-band Network Telemetry (INT) is a new network measurement technology. Traditional network measurement methods mostly focus on the network performance between terminals, and their measurement results are often based on packet loss and delay information; INT, on the other hand, focuses on the real-time network status of each network node. The current INT protocol embeds an INT command header containing telemetry instructions in the header of the service message. Each network node adds the corresponding network status information (that is, metadata) to the INT header according to the telemetry instructions, so that network measurement is no longer a "black box". After receiving the INT header, the Network Monitoring System (NMS) can dynamically perceive the details of the network quality and improve its analysis capabilities.

Although INT can provide status information of each network device on the telemetry path, the problem of excessive measurement overhead needs to be solved urgently. In the normal working mode of INT, each time a data packet passes through a network node, the node will add its own metadata. Therefore, the number of bits occupied by the INT header will increase with the increase in the number of nodes passed by the data packet. Excessive INT header overhead will lead to a serious decrease in bandwidth utilization. In addition, since the telemetry path is dynamically changing, it is difficult to estimate and reserve the space required for the INT header, which will cause the network status information of some nodes to be discarded due to insufficient space, or form a large number of data fragments, thereby affecting network performance. Another "postcard" working mode only inserts an INT instruction header containing telemetry instructions into the header of the business data packet. Each network node executes the telemetry instructions in the data packet header and sends out the metadata representing its current network status in the form of an independent data packet. Although this fixes the overhead of the INT header, the number of telemetry data packets caused will be several times that of normal business packets, which will seriously affect the accuracy of network measurement. In addition, the processing burden of network nodes is also greatly increased.

In order to reduce the overhead of the INT header, some studies compress the INT header according to the characteristics of the network status that need to be collected. For example, for the buffer occupancy ratio on the packet output port, the occupancy ratio of each node is no longer recorded and only the maximum value is retained; for the resident processing time on the node, the time of each node is no longer recorded, only the cumulative time sum is recorded or the longest time on a single node is retained. Unfortunately, such compression processing sacrifices the state details of the node and loses the global visualization and versatility of network telemetry. Some research works also use sampling strategies (based on probability prediction, etc.) to only allow a part of the business data packets to carry INT metadata to reduce the average overhead of the INT header. However, this cannot completely solve the problem of excessive INT overhead, and the payload of the data packet carrying INT metadata is still difficult to estimate.

Nowadays, the bandwidth of network physical interfaces is constantly increasing, reaching 10G, 100G and even higher, and the interval between the arrival of data packets has become very short. Coupled with the packet detection feature of INT, the high-speed and continuous generation of INT reports will form a data lake in the NMS, and the network status carried in a large number of INT reports is repeated or has very little difference, resulting in information redundancy. How to handle massive INT headers and efficiently record changes in network status are new challenges facing INT.

Summary of the invention

The present invention provides a stateful in-band network telemetry method and system, which efficiently completes the network telemetry task based on the state register of the network node and the SF-INT forwarding rule of the state information; in the standard mode of the network telemetry, the INT header length is kept as a set constant and no additional data packet is generated; in the emergency mode, seamless switching with the current INT can be achieved; the "sampling" attribute of the state register enables the NMS to avoid saving INT reports with extremely short intervals and redundant information, and better solves the problem of massive data.

The present invention adopts the following technical solution:

In one aspect, a stateful in-band network telemetry method includes:

INT source node processing steps: the INT source node receives the data packet that needs to perform network telemetry, inserts the INT header after parsing, and sends it to the next node after repackaging; the INT header includes the INT valid header; the INT valid header includes the INT instruction header and the metadata stack; the INT instruction header includes the C field, the U field, the ST field and the packetID field; the C field is set to whether to use SF-INT; the U field is set to emergency mode or standard mode; the ST field is set to X, which is used to indicate to the subsequent INT forwarding node that the INT header carries the metadata segment that needs to be XORed; the packetID field is set to the data packet ID; the metadata stack is used to insert the metadata segment of the INT source node itself;

INT forwarding node processing steps: when the U field is set to the standard mode, after the INT forwarding node receives the data packet from the previous node, it parses the valid INT header carried in the arrived data packet; compares the arrived INT valid header with the INT valid header in the register, and obtains the INT valid header to be sent and the INT valid header to be stored in the register based on the SF-INT forwarding rule; the INT valid header to be sent is encapsulated and sent to the next node; the data packet to be stored is stored in the register;

INT end node processing steps: After receiving the data packet from the previous node, the INT end node parses the INT header in the data packet, directly adds its own metadata segment to the metadata stack, generates an INT report, and encapsulates the INT report in an independent message and sends it to the NMS;

NMS processing steps: After receiving the INT report from SF-INT, NMS combines the multiple INT reports with the same packetID received before and after to obtain the raw metadata segment of each node.

Preferably, the INT forwarding node processing step further includes:

When the U field is set to emergency mode, or when the node detects a network event or receives a related network event notification, it directly switches to emergency mode to achieve instant seamless switching; it is determined whether there is an INT valid header stored in the register, and if so, the INT valid header stored in the register is encapsulated into an INT report and sent to the NMS in an independent data packet; thereafter, the node no longer uses the register to store historical network status information, but directly adds its own metadata segment to the metadata stack carried by the arriving data packet.

Preferably, before the INT source node processing step, a telemetry control step is also included, which is as follows:

Based on the configuration data, the terminal APP or the SDN northbound APP sends a message to the INT source node; the message includes the data packet that needs to be telemetered, the telemetry instruction, the maximum number of telemetry nodes, and whether to enable SF-INT.

Preferably, the method further comprises:

When an INT source node, an INT forwarding node, or an INT end node inserts or XORs its own metadata segment in the metadata stack, the value in the "RemainingHop Cnt" field in the INT instruction header is reduced by 1.

Preferably, the value of the ST field can also be set to H, L and E; H indicates high priority; L indicates low priority; E indicates that the metadata stack does not contain the metadata segment.

Preferably, the arriving INT valid header is compared with the INT valid header in the register, and the INT valid header to be sent and the INT valid header to be stored in the register are obtained according to the SF-INT forwarding rule. The specific rules include:

If the ST field of the arriving INT valid header is X and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register, the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L;

If the ST field of the INT valid header that arrives is X, and the ST field of the INT valid header in the register is L, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the INT valid header that arrives and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the INT valid header that arrives, and the corresponding ST field is set to H;

If the ST field of the arriving INT valid header is X, and there is no INT valid header stored in the register, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the arriving INT valid header and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the corresponding ST field is set to H;

If the ST field of the arriving INT valid header is H and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header;

If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include the same data packet ID, the INT valid header to be sent is set to the INT valid header in the register; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L;

If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include different packet IDs, the INT valid header to be sent is set to the arriving INT valid header; and the INT valid header in the register remains unchanged;

If the ST field of the arriving INT valid header is H and there is no INT valid header stored in the register, the INT valid header to be sent is set to the arriving INT valid header and the register remains empty;

If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header;

If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is L, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header;

If the ST field of the arriving INT valid header is L and the register does not store the INT valid header, the INT valid header to be sent is set to the arriving INT valid header, and the register remains empty.

Preferably, the NMS processing step specifically includes:

If the current data packet ID is equal to 255, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, discard the bottom metadata segment, read the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID;

If the current data packet ID is not equal to 255, and the ST field in the data packet is equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, then read the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID;

If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment, sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID;

If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, the processing also includes: reading non-bottom bare data segments, and arranging all bare data segments, and ending the processing of the current data packet ID.

Preferably, the method for obtaining the number of nodes contained in the bottom metadata segment of the metadata stack is as follows:

N3＝N1-(N2-1)＝N1+1-N2

Among them, N1 indicates that the metadata stack contains the network status information of N1 nodes; N1=MAX-R, MAX is the maximum number of nodes allowed by telemetry, and R represents the value in the "RemainingHop Cnt" field of the INT instruction header; N2 represents the number of metadata segments contained in the metadata stack, N2=(L2-12)/L0, where L0 represents the length of the metadata segment, and L2 is the length of the INT effective header.

Preferably, the NMS processing step further includes:

If the U field is in emergency mode, relevant processing of the emergency message is performed.

In another aspect, a stateful in-band network telemetry system includes:

A telemetry control module, used to send configuration data to the INT source node; the configuration data includes a data packet to be telemetered, a telemetry instruction, a maximum number of telemetry nodes, and whether to enable SF-INT;

A telemetry data packet processing module, including a source node processing submodule, a forwarding node processing submodule and an end node processing submodule;

The source node processing submodule is used to receive data packets that need to perform network telemetry, insert an INT header after parsing, and re-encapsulate and send it to the next node; the INT header includes an INT valid header, and the INT valid header includes an INT instruction header and a metadata stack; the INT instruction header includes a C field, a U field, an ST field and a packetID field; the C field is set to indicate whether SF-INT is adopted; the U field is set to emergency mode or standard mode; the ST field is set to X, which is used to indicate to subsequent INT forwarding nodes that the INT header carries a metadata segment that needs to be XORed; the packetID field is set to the data packet ID; the metadata stack is used to insert the metadata segment of the INT source node itself;

The forwarding node processing submodule is used for, when the U field is set to adopt the standard mode, after the INT forwarding node receives the data packet from the previous node, parsing out the valid INT header carried in the arrived data packet; comparing the arrived INT valid header with the INT valid header in the register, and obtaining the INT valid header to be sent and the INT valid header to be stored in the register based on the SF-INT forwarding rule; encapsulating the INT valid header to be sent and sending it to the next node; and storing the data packet to be stored in the register;

The end node processing submodule is used to parse the INT header in the data packet after receiving the data packet from the previous node, directly add its own metadata segment in the metadata stack, generate an INT report, and encapsulate the INT report in an independent message and send it to the NMS;

The NMS collection and calculation module is used to process the INT report received from SF-INT in combination with multiple INT reports with the same packetID received before and after to obtain the raw metadata segment of each node.

The beneficial effects of the present invention are as follows:

(1) In the existing INT, the space occupied by the INT header increases with the number of nodes on the telemetry path; however, in the SF-INT of the present invention, the overhead of the INT header is fixed regardless of how many nodes it passes through. Therefore, SF-INT is not sensitive to the number of nodes on the telemetry path and can implement telemetry in a larger range;

(2) After the current INT deployment, the large amount of INT header overhead will inevitably reduce the effective service load; however, the metadata stack in the SF-INT header of the present invention usually has only one metadata segment, which makes the INT header overhead drop sharply and greatly improves the effective utilization of bandwidth;

(3) The SF-INT of the present invention introduces many improvements, including using node registers for distributed storage to implement stateful INT, improving the computing power of NMS, and being fully compatible with the current INT working mode and achieving instant seamless switching;

(4) The existing INT lacks sufficient processing capability for packet loss. Consider the case where the telemetry path is R1->R2->R3->R4->R5->R6. When packet loss occurs at R3, the NMS under the existing INT cannot know that packet loss has occurred (because the service data packet carrying the network status information of R1 and R2 has been lost and cannot reach the NMS). However, under the SF-INT of the present invention, although packet loss occurs at R3, the historical status information is still stored by the previous node, so the NMS still has the opportunity to obtain the network status information of the forwarding nodes (R1 and R2) before the packet loss.

The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, but the stateful in-band network telemetry method and system of the present invention are not limited to the embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of the current process of in-band network telemetry (INT);

FIG2 is a schematic diagram of a stateful in-band network telemetry (SF-INT) process of the present invention;

FIG3 is a diagram showing the format of a valid header of a current INT;

FIG4 is a format diagram of an INT valid header of SF-INT of the present invention;

FIG5 is a flow chart of a stateful in-band network telemetry method of the present invention;

FIG6 is a diagram showing the basic process of telemetry processing of an INT forwarding node of the present invention;

FIG7 is a NMS processing flow chart of the present invention;

FIG8 is a diagram of the SF-INT process of six nodes of the present invention;

FIG9 is a schematic diagram of two example processes of SF-INT in the emergency mode of the present invention;

FIG. 10 is a module diagram of a stateful in-band network telemetry system of the present invention.

Detailed ways

The present invention is further described below by way of specific implementation methods. It should be noted that the specific embodiments described herein are only used to facilitate description and explanation of the specific implementation methods of the present invention and are not used to limit the present invention.

In order to make the purpose and technical solution of the present invention clearer, the present invention is further described below in conjunction with the accompanying drawings and examples. It should be understood that the examples described herein are only used to explain the present invention and are not used to limit the present invention.

First, some terms used in the present invention are listed and explained to facilitate understanding by those skilled in the art and to avoid ambiguity.

Metadata: The telemetry control module will issue telemetry instructions. For each telemetry instruction, the network node needs to provide its own network status information, which is called metadata.

Metadata segment: During a telemetry process, the telemetry control module will issue a set of telemetry instructions; and the network node provides different metadata for different instructions. The sum of all metadata is called the metadata segment.

The length of the metadata segment: the number of bits occupied by the metadata segment, in bytes. The present invention uses L0 to represent this.

Bare metadata segment: If the content of a metadata segment comes from the original network status information of a node, it is called a bare metadata segment.

XOR metadata segment: A new metadata segment generated by XOR calculation of multiple metadata segments is called an XOR metadata segment. This is also an innovative definition of SF-INT.

Metadata stack: The metadata stack contains zero or more metadata segments, which come from different network nodes. SF-INT stipulates that there can be at most one XOR metadata segment in a metadata stack (located at the bottom of the stack), and metadata segments other than the bottom are all bare metadata segments. In the current INT, the metadata stack is full of bare metadata segments.

INT metadata header: contains the telemetry instruction set, four fields customized by SF-INT, the number of nodes passed through, the length of the metadata segment, and other information.

INT valid header: consists of the INT instruction header and metadata stack.

The length of the INT effective header: the present invention uses L2 to represent it, and the unit is byte. Assuming that the number of metadata segments in the metadata stack is N, then L2=12+L0×N.

INT shim header: The INT shim header has different definitions in different scenarios. For example, the INT shim header in INT over GRE and INT over TCP/UDP scenarios has different definitions. The "length" field in the INT shim header is very important and is used to calculate the occupied length of the INT effective header.

INT header: Contains the INT mezzanine header, INT command header and metadata stack.

INT source node: refers to INT source node.

INT forwarding node: refers to INT transit node.

INT end node: refers to the INT sink node.

Network Monitoring System NMS: NMS receives INT reports sent by INT end nodes and parses and calculates INT metadata segments of each node on the telemetry path. These calculation results are stored and can be used for network measurement and network management.

Interested packets: The telemetry control module usually specifies which types of packets to sample to obtain network status information on their forwarding path. These packets that are required to collect information are called interested packets.

P4-INT: refers to the INT standard developed by the P4 working group. The latest one is the INT v2.1 standard (https://p4.org/specs/). The implementation of INT currently relies mainly on the programmable data plane, and the de facto standard of programmable language is P4. Therefore, the P4-INT standard is widely accepted by academia and industry and has become the current INT technical standard.

The ID of the network node: refers to the order in which the network node appears in the telemetry path. For simplicity, the present invention names the network node as "R+the ID of the network node". For example, the ID of the first network node in the telemetry path is 1, and the name is R1. R1 is also the INT source node, and the subsequent network forwarding nodes are successively recorded as R2, R3, ...

As shown in FIG1 , in the existing INT telemetry process, each forwarding node will generate INT metadata, so the length of the metadata stack increases with the number of forwarding times, and the length of the entire INT header will also become longer and longer. As shown in FIG2 , in the SF-INT telemetry process of the present invention, there is always only one metadata segment in the metadata stack of the network node (INT source node, INT forwarding node, and INT end node), so the INT header will not grow as the number of nodes passed by the data packet increases. The network information in the INT forwarding node register and the sent metadata are calculated and obtained through the SF-INT forwarding rules (Table 1), so it is also called stateful; in contrast, the current INT is considered to be stateless. Specifically, FIG2 shows the situation of SF-INT in standard mode; in emergency mode, the node will automatically switch to the current INT mode.

In order to realize the transmission of unit data segments, SF-INT defines new fields in the INT instruction header to realize stateful network telemetry. Specifically, as shown in Figures 3 and 4, the INT header reserved field (the Reserved field has 12 bits in total, as shown in Figure 3) in the current INT protocol is used to define the four new fields required by SF-INT (as shown in Figure 4). Since SF-INT does not change the meaning and position of other fields in the current INT instruction header, it is compatible with it.

The four new fields of SF-INT are:

(1) The first C field occupies one bit. If the value is 1, it indicates that it is a SF-INT header; otherwise, it is a current INT header.

(2) The second U field/urgent field occupies one bit. If the value is 1, it indicates that the SF-INT is in urgency mode; 0 indicates that the SF-INT is in standard mode.

(3) The third ST status field occupies 2 bits and can represent the following four INT states:

00: stands for Empty (abbreviated E), indicating that the metadata stack does not contain a metadata segment.

01: stands for High (abbreviated as H), which means high priority and also represents history information.

10: stands for Low (abbreviated L), which means low priority.

11: Represents XOR (abbreviated as X), which means exclusive OR.

The bottom metadata segment in the metadata stack is related to the status field and can be labeled with its abbreviation. For example, when the value of the status field is 11, the bottom metadata segment in the metadata stack is called the X metadata segment.

(4) The fourth "packetID" field occupies 8 bits and is used to indicate the packet ID. The packet ID generated by the INT source node ranges from 0 to 254 and can be used cyclically; 255 is a reserved number used to indicate the ID of the current packet. The purpose of establishing this field is to allow INT forwarding nodes and NMS to distinguish adjacent packets. There may be multiple metadata segments in a metadata stack. The packetID of the packet to which the bottom metadata segment in the metadata stack belongs is the value in the "packetID" field. The packet to which the non-bottom metadata segment in the metadata stack belongs is the current packet, where packetID = 255. If the NMS receives two INT reports, even if the packet ID in the INT command header in the report is 255, the NMS will assume that the two INT packet headers report telemetry information of two independent packets.

Obviously, the metadata segment in SF-INT needs to be associated with the INT forwarding node, INT status, and packetID. For ease of description, the present invention designs a notation for the SF-INT metadata segment, which is in the form of:

Status Symbols

Among them, M stands for Metadata, and the "status symbol" on the right represents the "status field" in the INT instruction header. Three states are considered here, namely X, H and L. Only the bottom metadata segment in the metadata stack needs to be marked with a status symbol, and the non-bottom metadata segment is irrelevant to the state and does not need to be identified. The packetID in the superscript of M represents the packetID of the data packet to which the metadata segment belongs; the IDs in the subscript represent the IDs of one to multiple nodes. IDs indicates that this metadata segment is the network status information collected after a certain data packet passes through which nodes. The standard writing regulations for IDs include:

1. If there is a comma, the number between the two commas represents an ID;

Second, each digit not between two commas represents a node ID;

3. Commas separate nodes.

For example, IDs="1,2" represents R1 and R2; IDs="12" also represents R1 and R2; IDs="12,12," represents R1, R2, and R12; IDs=",12,13," represents R12 and R13. Indicates that this metadata segment is the XOR result of the metadata segments collected when the data packet with packetID=1 passes through nodes R1 and R2. The 1 and 2 in the middle and lower part represent the node ID, that is, the metadata contains the network status information obtained when the data packet passes through nodes R1 and R2. Because two nodes are marked, it means that this metadata segment is an XOR metadata segment, which is formed by the XOR of the raw metadata segment of nodes R1 and the raw metadata segment of R2. If there is only one node ID, then this metadata segment is a raw metadata segment. The 1 in the middle and upper part represents the packetID, which is the ID value assigned by the INT source node.

As shown in FIG. 5 , in this embodiment, a stateful in-band network telemetry method includes: a telemetry control step S51 , a network node processing step S52 , and an NMS processing step S53 .

The telemetry control step S51 includes: based on the configuration data, the terminal APP sends a message to the INT source node; the message includes a data packet that needs to be telemetered, a telemetry instruction, a maximum number of telemetry nodes, and whether to enable SF-INT.

Specifically, the execution subject of the telemetry control step can be an independent program installed on the terminal, or it can be a northbound application on an SDN (Software Define Network) controller.

Through the telemetry control step, the INT source node is informed of which data packets to telemeter (i.e., which data packets are of interest); telemetry instructions are issued (which network status to collect), the maximum number of telemetry nodes, whether to start SF-INT, etc. The telemetry control step belongs to the control plane and is only executed when network telemetry is started, while the telemetry process is completely completed by the data plane.

The network node processing step S52 includes: an INT source node processing step S521, an INT forwarding node processing step S522 and an INT end node processing step S523.

The INT source node processing step S521 includes: the INT source node receives the data packet that needs to perform network telemetry, parses it and inserts the INT header, re-encapsulates it and sends it to the next node; the INT header includes an INT valid header, and the INT valid header includes an INT instruction header and an INT metadata stack.

Specifically, after receiving the data packet that needs to perform network telemetry, the INT source node inserts the INT header. According to the telemetry instruction issued, the INT source node sets the Instruction Bitmap field; calculates the required length of a metadata segment and sets the HopML field accordingly; sets the value of the length field in the INT mezzanine header to L0+12 bytes (equal to the length of the INT effective header), where 12 bytes are the fixed length of the INT instruction header. The INT source node also needs to set the SF-INT specific fields: set the C field to 1 to indicate to subsequent forwarding nodes and NMS that this is an SF-INT header; set the ST field to X to indicate to subsequent forwarding nodes that this INT header carries a metadata segment that needs to be XORed; insert a data packet sequence ID in the packeID field (the value range is 0 to 254, automatically incremented, and can be used cyclically); when encountering a network event, the U field is set to 1, otherwise it is set to 0.

At the same time, the INT source node inserts its own metadata segment, which is also the first metadata segment, into the metadata stack according to the requirements of the telemetry instruction, and updates the relevant fields in the INT header, including subtracting 1 from the value in the "RemainingHop Cnt" field. The metadata segment sent by the INT source node can refer to the metadata segment sent by R1 to R2 shown in Figure 8 in Example 1.

The INT forwarding node processes step S522, including:

S5221, when the U field is set to adopt the standard mode (the value in the U field is 0, and when no network event is encountered or notification of a network event is received, the forwarding node is in the standard mode of SF-INT), after the INT forwarding node receives the data packet from the previous node, it parses the valid INT header carried in the arriving data packet; compares the arriving INT valid header with the INT valid header in the register, and based on the SF-INT forwarding rule, obtains the INT valid header to be sent and the INT valid header to be stored in the register; the INT valid header to be sent is encapsulated and sent to the next node; the data packet to be stored is stored in the register;

After receiving the SF-INT header from the previous node (ie, the value in the C field is 1), the INT forwarding node processes the data packet according to the SF-INT forwarding rule.

As shown in Figure 6, the SF-INT forwarding node uses a register to store historical network status information, and its packet processing method involves the change of INT status. The node parses the information of the INT header carried in the received data packet, compares it with the information in the INT valid header in the register, and finally determines what kind of INT header should be carried in the sent data packet and the content of the INT valid header to be newly stored in the register according to the SF-INT forwarding rule table of the INT forwarding node (such as Table 1). The metadata segment stored in the local register is either an H metadata segment or an L metadata segment.

The symbol M(N) used in the rule table of Table 1, where N∈{Arrival (abbreviated as A), Storage (abbreviated as S)}, and the INT state set M∈{X,H,L,E}. Specifically, A represents the INT valid header carried in the arriving data packet; S represents the INT valid header stored in the register of each node. For example, X(A) represents that the "status" field in the arriving INT datagram header is X; in particular, E() represents that the INT valid header is not stored in the node. In addition, the C metadata segment (C stands for Current) is used to represent the current network status information of the node.

Taking forwarding rules 1, 5 and 6 as examples, how the rules in Table 1 are applied on the forwarding node is described.

Rule 1: In Rule 1, the data packet that arrives contains the X metadata segment; when it arrives, the node stores the H metadata segment. In this case, the node will send out the H metadata segment in storage, and the state will remain H; while the X metadata segment that arrives will be stored, and the state will be L, that is, the X metadata segment that arrives will be converted to the L metadata segment for storage. Here we can see that the H metadata segment has a higher priority than the X metadata segment. For example: Suppose on a forwarding node, the received data packet carries At this time, the register stores /> Then rule 1 applies, so the result of packet processing is: the metadata segment stored in the register is/> The data packet sent out carries the metadata segment/> (Reference Example 1)

Table 1 SF-INT forwarding rules of INT forwarding nodes

Table 1 SF-INT forwarding rules of INT forwarding nodes

*Indicates that E() is encountered on the INT forwarding node, which means that the path of the data packet may have changed.

Let's look at rule 5. In rule 5, H(A) arrives, which means that the arriving data packet carries H metadata segment; when it arrives, L metadata segment is stored in the node. The received metadata segment and the stored metadata segment belong to the same data packet, which means that the "XOR" process of the data packet has been blocked. At this time, the stored metadata segment will be sent out. For example: Assume that the metadata segment stored on the R3 node is The received data packet carries the metadata segment The received and stored metadata segments all belong to the packet with packetID 2, so rule 5 applies. After the packet is processed according to rule 5, the metadata segment stored in the register is/> The metadata segment carried in the data packet sent out is/> (Reference Example 1)

The difference between Rule 6 and Rule 5 is that Rule 6 applies to scenarios where the received metadata segment and the stored metadata segment do not belong to the same data packet. Example: Assume that the metadata segment is stored on the R3 node. The received data packet carries a metadata segment /> The metadata segment stored afterwards is/> The metadata segment carried in the data packet sent out is/>

Table 1 designs ten forwarding rules, covering all possible situations that the INT forwarding node may encounter. These rules are clear and unambiguous, making it easy to write relevant codes on the switch. The basic idea is that if a data packet fails to obtain a new network state at a certain node, the state is set to a low priority L, so that the NMS can obtain the metadata of the entire process of other data packets as soon as possible. Therefore, the priority order is: H is higher than X, and X is higher than L. The NMS needs the H metadata segment that arrives later to restore the raw metadata segment of each node, so the H metadata segment needs to pass first; L means that the data packet to which the metadata belongs fails to obtain the network state of the node at a certain node, so L must give way to H and X.

It should be noted that in rules 2 and 3, the node adds the network status information of its own metadata segment to the metadata stack, and at this time, the value in the "RemainingHop Cnt" field needs to be subtracted by 1. In other rules, only the historical network status information is transmitted, and the current network status information is not added, so the value in the "RemainingHop Cnt" field is not modified.

S5222: When the value in the U field is 1, or a network event is detected or a notification of a related network event is received, SF-INT switches to emergency mode and transmits enough network status information as quickly as possible. If there is an INT valid header in the register of the node that triggers the emergency mode, the INT valid header stored in the register is encapsulated into an INT report and sent to the NMS in an independent data packet (that is, only one independent "postcard" message is generated). After that, the node will no longer use the register to store historical network status information, but directly send the network status information, that is, directly add its own metadata segment to the metadata stack carried by the arriving data packet, and update the relevant fields in the INT header.

The INT end node processing step S523 includes: after receiving the data packet from the previous node, the INT end node parses the INT header in the data packet, directly adds its own metadata segment in the INT metadata stack, generates an INT report, and encapsulates the INT report in an independent message and sends it to the NMS.

Specifically, after receiving a data packet with a telemetry header, the INT end node separates the INT header from the data packet, then directly adds its own metadata segment to the metadata stack, and updates the relevant fields in the INT header (for example, subtract 1 from the value in the "RemainingHop Cnt" field), then generates an INT report (INT report), and encapsulates the INT report in a separate message and sends it to the NMS. The INT end node also updates the relevant fields of other headers in the data packet, and then sends the data packet without the INT header to the destination. In order to facilitate NMS processing, SF-INT requires the end node to add its own metadata segment to the metadata stack. The current INT allows the end node to separate its own metadata segment from the received metadata stack, but still in the same INT report.

The NMS processing step S53 includes: after receiving the INT report of SF-INT, the NMS combines the multiple INT reports with the same packetID received before and after to process and obtain the raw data segment of each node.

Specifically, when NMS receives the INT report of SF-INT, it needs to combine multiple INT reports with the same packetID received before and after to calculate in order to obtain the original metadata segment of each node. The processing flow on SF-INT is shown in Figure 7.

NMS must first confirm how many metadata segments are in the metadata stack reported by INT. The length of the INT instruction header is fixed at 12 bytes, and the number of metadata segments contained in a metadata stack is recorded as N2:

N2＝(L2-12)/L0

Next, the NMS needs to confirm how many nodes are included in the metadata stack. When the value in the "RemainingHopCnt" field is R bytes, it is calculated that the metadata stack contains the network status information of N1 nodes, that is, N1 = MAX-R, where MAX is the maximum number of nodes allowed by the telemetry system. In SF-INT, the forwarding node will only subtract 1 from the value in the "RemainingHop Cnt" field after adding its own network status information to the metadata stack.

Finally, NMS needs to calculate how many nodes' network status is contained in the bottom metadata segment in the SF-metadata stack:

N3＝N1-(N2-1)＝N1+1-N2

If N3=1, then the bottom metadata segment is not an XOR metadata segment. If N3>1, then the bottom metadata segment is an XOR metadata segment.

The "XOR restoration with metadata segments with the same packetID from other packets" in Figure 7 refers to the following method: After receiving the report, the NMS will analyze it according to the "status" field in the INT header. When the status bit is X, all metadata segments belong to the same packet. But when the status bit is H or L, the bottom metadata segment belongs to the packetID in the INT instruction header, and the other non-bottom metadata segments belong to the current packet (i.e. packetID = 255).

Assume that NMS receives a data packet containing XOR metadata segments of 4 nodes, and also receives another XOR metadata segment of the same data packet containing 3 nodes (distinguished by packetID), then NMS XORs these two XOR metadata segments again to obtain the metadata segment of the 4th node. Therefore, by comparing the previous and next INT reports, NMS can eventually gradually recover the raw metadata segments of each node on the path.

Example 1: Figure 8 shows a SF-INT telemetry process with 6 nodes. This is the process of SF-INT in standard mode, during which no network events are detected and no switch to emergency mode is made. R1 is the first node in the telemetry path and is therefore the INT source node. R2, R3, R4, and R5 are INT forwarding nodes; R6 is the INT terminal node. The leftmost number in the figure represents the sequence ID of the data packet. The INT source node inserts the ID in the SF-INT instruction header.

Take the data packet with sequence number 3 as an example. The packetID field of the SF-INT header carried by R1 is 3, which is consistent with the sequence number of the data packet; however, when the data packet leaves R2, the metadata segment carried by it is represented as This indicates that the value of the packetID field in the SF-INT header carried by the data packet is 2, which means that the information in the metadata segment carried by the data packet is the network status information when the data packet with sequence number 2 passes through the network.

Now let's look at the gray part related to the R3 forwarding node. Represents the metadata segment carried by the arriving data packet; above R3 is/> It represents the metadata segment in the INT valid header stored in the R3 node when the data packet arrives; therefore, R3 will process the telemetry header according to the forwarding rules in Table 1, and the node determines the applicable rule 2. Therefore, the metadata segment carried in the sent data packet is/> (indicated on the right side of R3), after telemetry processing, what is stored in the R3 register is/> (Indicates below the node).

Next, observe the INT end node R6. In the first row of the figure, R6 receives Finally, add its own metadata segment to the metadata stack/> The reason why packetID is 255 is that as long as it is not the bottom metadata segment, its packetID is considered to be 255, which means that the metadata segment belongs to the data packet carrying the INT header. Because the SF-INT header is in the X state, it means that the packetID in the SF-INT instruction header is the packetID assigned to this data packet by the INT source node, so in this scenario, NMS will also treat it as/> See as/> In the second row of the figure, R6 receives /> Will add its own metadata segment to the metadata stack/> Because the SF-INT header is in the H state, the NMS will attribute the bottom metadata segment and the non-bottom metadata segment to different data packets.

The NMS receives the metadata segment in multiple consecutive INT reports. Therefore, NMS can recover the telemetry information of the data packet with packetID 0 passing through R1, R2, R3, R4, R5 and R6. In fact, NMS receives /> It is not known which five nodes' metadata segments are XORed into the metadata segment at the bottom of the stack. It is only calculated by N3 that it contains the network status information of the five nodes.

NMS will receive multiple Because the packetID is 255, these metadata segments are considered to be metadata segments generated when different data packets pass through the node.

As shown in Figure 8, there is only one metadata segment in the metadata stack in the INT header on the link between network nodes, and the overhead of the INT header is fixed. For these 8 data packets, the NMS can obtain the full telemetry information of data packet No. 0, and partial telemetry information of data packet No. 4, as well as the telemetry information of other data packets at the end node.

A packet convergence period is defined as the number of packets that the NMS needs to go through to obtain the full telemetry information of a packet without packet loss or packet disorder. The packet convergence period in SF-INT is always ^2a , which is related to the number of nodes F on the telemetry path and satisfies 2a ^-1 ≤F- ^2≤2a , where a is a positive integer.

If there are 5 nodes on a telemetry path, then its period is 4;

If there are 6 nodes on a telemetry path, then its period is 8;

If there are 9 nodes on a telemetry path, then its period is 8.

Embodiment 2: An actual SF-INT process (emergency mode): A node detects a network abnormality event or receives a notification and enters an emergency state.

When the data packet with sequence ID 6 reaches R3, R3 detects a network event or receives a notification of a network event, and R3 switches to emergency mode. As shown in Figure 9, R3 will store the It is sent to NMS as an independent report, and the register is no longer used to store historical status messages before the emergency mode ends. Therefore, you can see that the bottom of R3, R4, and R5 in the figure are blank, without the mark of the metadata segment, which also means that there is no new storage. R3 and subsequent nodes will directly add the metadata segment representing their current network status to the metadata stack.

When the data packet with sequence ID 7 reaches R3, R3 detects a network event or receives a notification of a network event, and R3 switches to emergency mode. As shown in Figure 9, R3 will store the It is sent to NMS as an independent report, and no longer uses the register to store historical status messages before the emergency mode ends. R3 will directly add the metadata segment representing its current network status to the metadata stack.

It should be noted that no matter whether the forwarding node that triggers the emergency mode receives the metadata segment in the X state or the metadata segment in the H/L state, the NMS can accurately distinguish and recover the raw metadata segment obtained when the data packet passes through the network according to Figure 7.

This embodiment illustrates that SF-INT can switch to the current INT technology instantly and seamlessly in emergency mode. In addition, if the network event is detected at the location of the INT source node, the entire telemetry process will be consistent with the current INT working process. The so-called instant seamless switching means that any node can directly switch to emergency mode when it detects a network event or receives a notification of a network event, and does not need to wait until the next interesting data packet arrives at the INT source node to start switching.

It can be seen that the stateful SF-INT is compatible with the stateless current INT, and the current INT can be regarded as a special case of the stateful SF-INT.

Referring to FIG. 10 , according to another aspect of the present invention, a stateful in-band network telemetry system includes:

Telemetry control module 101, used to send configuration data to the INT source node; the configuration data includes the data packet to be telemetered, telemetry instructions, the maximum number of telemetry nodes and whether to enable SF-INT;

The telemetry data packet processing module 102 includes a source node processing submodule 1021, a forwarding node processing submodule 1022 and an end node processing submodule 1023;

The source node processing submodule 1021 is used to receive a data packet that needs to perform network telemetry, insert an INT header after parsing, and re-encapsulate and send it to the next node; the INT header includes an INT valid header, and the INT valid header includes an INT instruction header and an INT metadata stack; the INT instruction header includes a C field, a U field, an ST field and a packetID field; the C field is set to use SF-INT; the U field is set to emergency mode or standard mode; the ST field is set to X, which is used to indicate to subsequent INT forwarding nodes that the INT header carries a metadata segment that needs to be XORed; the packetID field is set to the data packet ID; the INT metadata stack is used to insert the metadata segment of the INT source node itself;

The forwarding node processing submodule 1022 is used for, when the U field is set to adopt the standard mode, after the INT forwarding node receives the data packet from the previous node, parsing the valid INT header carried in the arrived data packet; comparing the arrived INT valid header with the INT valid header in the register, and obtaining the INT valid header to be sent and the INT valid header to be stored in the register based on the SF-INT forwarding rule; inserting the INT valid header to be sent into the INT header, re-encapsulating and sending it to the next node; and also for switching to the current INT protocol instantly and seamlessly when the U field is set to adopt the emergency mode;

The end node processing submodule 1023 is used to parse the INT header in the data packet after receiving the data packet from the previous node, directly add its own metadata segment in the INT metadata stack, generate an INT report, and encapsulate the INT report in an independent message and send it to the NMS;

The NMS collection and calculation module 103 is used to process, after receiving the INT report of SF-INT, a plurality of INT reports with the same packetID received before and after, to obtain the raw metadata segment of each node.

The specific implementation of each module of a stateful in-band network telemetry system can be found in a stateful in-band network telemetry method, which will not be repeated here.

It should be understood that those skilled in the art can make improvements and changes based on the above description, and all these improvements and changes should fall within the scope of protection of the appended claims of the present invention.

Claims (7)

1. A stateful in-band network telemetry method, comprising: INT source node processing steps: the INT source node receives the data packet that needs to perform network telemetry, inserts the INT header after parsing, and sends it to the next node after repackaging; the INT header includes the INT valid header; the INT valid header includes the INT instruction header and the metadata stack; the INT instruction header includes the C field, the U field, the ST field and the packetID field; the C field is set to whether to use SF-INT; the U field is set to emergency mode or standard mode; the ST field is set to X, which is used to indicate to the subsequent INT forwarding node that the INT header carries the metadata segment that needs to be XORed; the packetID field is set to the data packet ID; the metadata stack is used to insert the metadata segment of the INT source node itself; INT forwarding node processing steps: when the U field is set to the standard mode, after the INT forwarding node receives the data packet from the previous node, it parses the valid INT header carried in the arrived data packet; compares the arrived INT valid header with the INT valid header in the register, and obtains the INT valid header to be sent and the INT valid header to be stored in the register based on the SF-INT forwarding rule; the INT valid header to be sent is encapsulated and sent to the next node; the data packet to be stored is stored in the register; INT end node processing steps: After receiving the data packet from the previous node, the INT end node parses the INT header in the data packet, directly adds its own metadata segment to the metadata stack, generates an INT report, and encapsulates the INT report in an independent message and sends it to the NMS; NMS processing steps: After receiving the INT report from SF-INT, NMS combines the multiple INT reports with the same packetID received before and after to obtain the raw metadata segment of each node; Specifically, the value of the ST field can also be set to H, L and E; H indicates high priority; L indicates low priority; E indicates that the metadata stack does not contain a metadata segment; In the INT forwarding node processing step, the arriving INT valid header is compared with the INT valid header in the register, and based on the SF-INT forwarding rule, the INT valid header to be sent and the INT valid header to be stored in the register are obtained. The specific rules include: If the ST field of the arriving INT valid header is X and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register, the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L; If the ST field of the INT valid header that arrives is X, and the ST field of the INT valid header in the register is L, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the INT valid header that arrives and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the INT valid header that arrives, and the corresponding ST field is set to H; If the ST field of the arriving INT valid header is X, and there is no INT valid header stored in the register, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the arriving INT valid header and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the corresponding ST field is set to H; If the ST field of the arriving INT valid header is H and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include the same data packet ID, the INT valid header to be sent is set to the INT valid header in the register; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L; If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include different packet IDs, the INT valid header to be sent is set to the arriving INT valid header; and the INT valid header in the register remains unchanged; If the ST field of the arriving INT valid header is H and there is no INT valid header stored in the register, the INT valid header to be sent is set to the arriving INT valid header and the register remains empty; If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is L, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is L and there is no INT valid header stored in the register, the INT valid header to be sent is set to the arriving INT valid header, and the register remains empty; The NMS processing steps specifically include: If the current data packet ID is equal to 255, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, discard the bottom metadata segment, read the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, then read the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment, sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, the processing also includes: reading non-bottom bare data segments, and arranging all bare data segments, and ending the processing of the current data packet ID. 2. The stateful in-band network telemetry method according to claim 1, wherein the INT forwarding node processing step further comprises: When the U field is set to emergency mode, or when the node detects a network event or receives a related network event notification, it directly switches to emergency mode to achieve instant seamless switching; it is determined whether there is an INT valid header stored in the register, and if so, the INT valid header stored in the register is encapsulated into an INT report and sent to the NMS in an independent data packet; thereafter, the node no longer uses the register to store historical network status information, but directly adds its own metadata segment to the metadata stack carried by the arriving data packet. 3. The stateful in-band network telemetry method according to claim 1, characterized in that before the INT source node processing step, it also includes a telemetry control step, which is as follows: Based on the configuration data, the terminal APP or the SDN northbound APP sends a message to the INT source node; the message includes the data packet that needs to be telemetered, the telemetry instruction, the maximum number of telemetry nodes, and whether to enable SF-INT. 4. The stateful in-band network telemetry method according to claim 1, characterized in that the method further comprises: When an INT source node, an INT forwarding node, or an INT end node inserts or XORs its own metadata segment in the metadata stack, the value in the "RemainingHop Cnt" field in the INT instruction header is reduced by 1. 5. The stateful in-band network telemetry method according to claim 1, characterized in that the method for obtaining the number of nodes contained in the bottom metadata segment of the metadata stack is as follows: N3＝N1-(N2-1)＝N1+1-N2 Among them, N1 indicates that the metadata stack contains the network status information of N1 nodes; N1=MAX-R, MAX is the maximum number of nodes allowed by telemetry, and R represents the value in the "RemainingHop Cnt" field of the INT instruction header; N2 represents the number of metadata segments contained in the metadata stack, N2=(L2-12)/L0, where L0 represents the length of the metadata segment, and L2 is the length of the INT effective header. 6. The stateful in-band network telemetry method according to claim 1, wherein the NMS processing step further comprises: If the U field is in emergency mode, relevant processing of the emergency message is performed. 7. A stateful in-band network telemetry system, comprising: A telemetry control module, used to send configuration data to the INT source node; the configuration data includes a data packet to be telemetered, a telemetry instruction, a maximum number of telemetry nodes, and whether to enable SF-INT; A telemetry data packet processing module, including a source node processing submodule, a forwarding node processing submodule and an end node processing submodule; The source node processing submodule is used to receive data packets that need to perform network telemetry, insert an INT header after parsing, and re-encapsulate and send it to the next node; the INT header includes an INT valid header, and the INT valid header includes an INT instruction header and a metadata stack; the INT instruction header includes a C field, a U field, an ST field and a packetID field; the C field is set to indicate whether SF-INT is adopted; the U field is set to emergency mode or standard mode; the ST field is set to X, which is used to indicate to subsequent INT forwarding nodes that the INT header carries a metadata segment that needs to be XORed; the packetID field is set to the data packet ID; the metadata stack is used to insert the metadata segment of the INT source node itself; The forwarding node processing submodule is used for, when the U field is set to adopt the standard mode, after the INT forwarding node receives the data packet from the previous node, parsing out the valid INT header carried in the arrived data packet; comparing the arrived INT valid header with the INT valid header in the register, and obtaining the INT valid header to be sent and the INT valid header to be stored in the register based on the SF-INT forwarding rule; encapsulating the INT valid header to be sent and sending it to the next node; and storing the data packet to be stored in the register; The end node processing submodule is used to parse the INT header in the data packet after receiving the data packet from the previous node, directly add its own metadata segment in the metadata stack, generate an INT report, and encapsulate the INT report in an independent message and send it to the NMS; The NMS collection and calculation module is used to process the INT report received from SF-INT in combination with multiple INT reports with the same packetID received before and after to obtain the raw metadata segment of each node; Specifically, the value of the ST field can also be set to H, L and E; H indicates high priority; L indicates low priority; E indicates that the metadata stack does not contain a metadata segment; In the forwarding node processing submodule, the arriving INT valid header is compared with the INT valid header in the register, and based on the SF-INT forwarding rule, the INT valid header to be sent and the INT valid header to be stored in the register are obtained. The specific rules include: If the ST field of the arriving INT valid header is X and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register, the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L; If the ST field of the INT valid header that arrives is X, and the ST field of the INT valid header in the register is L, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the INT valid header that arrives and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the INT valid header that arrives, and the corresponding ST field is set to H; If the ST field of the arriving INT valid header is X, and there is no INT valid header stored in the register, the metadata segment of the INT valid header to be sent is set to the data segment generated by XORing the metadata segment of the arriving INT valid header and the metadata segment of the current node, and the corresponding ST field is set to X; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the corresponding ST field is set to H; If the ST field of the arriving INT valid header is H and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include the same data packet ID, the INT valid header to be sent is set to the INT valid header in the register; the metadata segment of the INT valid header to be stored in the register is set to the metadata segment of the arriving INT valid header, and the ST field of the INT valid header to be stored in the register is set to L; If the ST field of the arriving INT valid header is H, the ST field of the INT valid header in the register is L, and the arriving INT valid header and the INT valid header in the register include different packet IDs, the INT valid header to be sent is set to the arriving INT valid header; and the INT valid header in the register remains unchanged; If the ST field of the arriving INT valid header is H and there is no INT valid header stored in the register, the INT valid header to be sent is set to the arriving INT valid header and the register remains empty; If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is H, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is L and the ST field of the INT valid header in the register is L, the INT valid header to be sent is set to the INT valid header in the register; the INT valid header to be stored in the register is set to the arriving INT valid header; If the ST field of the arriving INT valid header is L and there is no INT valid header stored in the register, the INT valid header to be sent is set to the arriving INT valid header, and the register remains empty; The NMS collection and calculation module is specifically used for: If the current data packet ID is equal to 255, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, discard the bottom metadata segment, read the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, then read the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID; otherwise, read the bottom bare metadata segment and the non-bottom bare metadata segment, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, determine whether the number of nodes contained in the bottom metadata segment of the metadata stack is greater than 1. If it is greater than 1, read the bottom XOR metadata segment, perform XOR restoration with the metadata segments with the same packetID from other data packets, and sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; otherwise, read the bottom bare metadata segment, sort out all the bare metadata segments with the same packetID, and end the processing of the current data packet ID; If the current data packet ID is not equal to 255, and the ST field in the data packet is not equal to X, the processing also includes: reading non-bottom bare data segments, and arranging all bare data segments, and ending the processing of the current data packet ID.