CN117580089B - AMF overload detection and control implementation method - Google Patents

Abstract

The invention relates to a method for realizing AMF overload detection and control. The invention uses the busyness of the task and the utilization rate of the message node as the quantization parameters of the actual load of the AMF, calculates the load state of the AMF through the load monitoring and load decision process, and achieves the aim of OVERLOAD control by mapping the load state of the AMF to the OVERLOAD action and using the OVERLOAD START and the OVERLOAD STOP interface to adjust the service flow sent by the 5G-AN. When the AMF is changed from overload to normal state, the UAC of the 5G-AN is not stopped immediately, but is stopped when the actual load is far below the overload threshold, so that the ping-pong effect in overload control can be avoided.

Description

AMF overload detection and control implementation method

Technical Field

The invention belongs to the field of communication, and particularly relates to an AMF overload detection and control implementation method.

Background

An access and mobility management function (AMF) is a core network element of a 5G system (5 GC), which provides services such as identity authentication, service authorization, and mobility management for a UE to access a 5G network. The stable operation of AMF is a basic premise that 5GC provides service for UE. When the AMF is operating beyond its rated capacity, the AMF is considered to be in an overload state (also referred to as an overload state). The AMF overload may not only affect the user experience of the UE in the 5G network, but may even cause system disturbance of the entire 5G network in severe cases. AN interface for the AMF to regulate the traffic of the 5G-AN is provided in 3gpp ts38.413, but the basis for the AMF trigger regulating mechanism is not specified. In order to ensure that the AMF can stably run under rated capacity and provide good user experience for UE, and achieve the purpose of dynamically adjusting the service flow of 5G-AN by the AMF, the invention provides a method for detecting and controlling the AMF overload.

Disclosure of Invention

Aiming at the problems existing in the prior art, the invention aims to provide a technical scheme of an AMF overload detection and control implementation method.

The implementation method for AMF overload detection and control is characterized by comprising the following steps: the method comprises the steps of performing real-time monitoring by taking task busyness and message node utilization rate as quantization parameters of AMF actual load, combining preset rated capacity and OVERLOAD threshold, calculating the load state of AMF through load monitoring and load decision process, mapping the AMF load state onto OVERLOAD action, and adjusting the service flow of 5G-AN through AN OVERLOAD START and OVERLOAD STOP interface to realize OVERLOAD control;

when the AMF is converted from overload to normal state, the unified access control UAC of the 5G-AN is not stopped immediately, but is stopped when the actual load is far lower than the overload threshold;

the task in the task busyness is a thread in the AMF, and the task busyness is the CPU occupation condition of the thread in the AMF; the message nodes in the message node utilization rate are carriers for transmitting messages between different threads in the AMF, and the message node utilization rate is the use condition of the message nodes in the AMF.

The implementation method for AMF overload detection and control is characterized by comprising the following steps: the interactive flow of AMF dynamically adjusting the service flow of 5G-AN is as follows, when the service flow sent by 5G-AN increases suddenly, AMF overload operation is caused, AMF carries out overload detection and control, and then sends overload start containing overload action to 5G-AN; after receiving overload, the 5G-AN executes unified access control UAC according to overload action, thereby reducing the service flow sent to the AMF; when the service flow sent by the 5G-AN is obviously reduced, the AMF load state is recovered to be normal, the AMF sends overload stop to the 5G-AN after overload detection and control are carried out, and the 5G-AN stops executing unified access control UAC after receiving the overload stop.

The implementation method for AMF overload detection and control is characterized by comprising the following steps: the method comprises four parts, namely presetting rated capacity and a threshold value, load monitoring, load decision-making and overload control, wherein the execution process of AMF overload detection and control is that the three parts of load monitoring, load decision-making and overload control are repeatedly executed continuously in the AMF operation process except the preset rated capacity and threshold value, so that the AMF is ensured to be operated always under the rated capacity;

the preset rated capacity and the threshold value are related parameters read from local configuration or acquired from a network manager when AMF is started;

the load monitoring is carried out after the AMF is started, and is divided into monitoring of task busyness and monitoring of the utilization rate of message nodes;

the load decision is based on AMF actual load parameters, a load decision algorithm is used for obtaining the maximum load, and then the load state of the AMF is determined by combining with a preset overload threshold;

the overload control is based on the result of the load decision, and in combination with the overload threshold, a policy is made as to whether the AMF needs to ignore all the received traffic, and whether the AMF needs to interact with the 5G-AN, so as to adjust the traffic sent by the 5G-AN to the AMF.

The implementation method of AMF overload detection and control is characterized in that the related parameters, namely P1, P2, P3, P4 and P5, wherein P1 and P2 are 2 subdivision parameters of task busyness, and are respectively a CPU utilization rate threshold value and a CPU continuous occupation count maximum value; p1 is used for judging the excessive standard of CPU usage in the monitoring unit; p2 is used for indicating the continuous monitoring times of the CPU utilization rate; p3 is total number of message nodes, and is used for indicating maximum value of concurrency quantity for information interaction; p4 is an overload threshold and is used for judging the overload standard of the AMF; p5 is a serious overload threshold, and is a standard for judging serious overload of the AMF; p2 and P3 indicate rated capacities, and P1, P4, and P5 indicate thresholds.

The implementation method of AMF overload detection and control is characterized in that the parameters used in load monitoring comprise: c1 is CPU utilization, C2 is CPU continuous occupation count, and C3 is idle message node number; the AMF comprises a plurality of tasks for processing different service requests, wherein each task corresponds to one thread in the AMF, and a pair of C1 and C2 exists for each task; there is one and only one C3 in the AMF to calculate the message.

The implementation method of AMF overload detection and control is characterized in that the task busyness load monitoring is periodically executed by taking each second as a unit, and the implementation method comprises the following specific steps:

step 1-1: calculating the CPU utilization rate of the task, namely C1;

acquiring the total number of the CPU time slices monitored at this time and the number of the task thread CPU time slices from a system file, and then calculating C1, C1=the number of the task thread CPU time slices/the total number of the CPU time slices of each task;

step 2-1: calculating the CPU continuous occupation count of the task, namely C2;

updating the C2 value according to the C1; if the CPU utilization rate of the task is greater than or equal to the CPU utilization rate threshold value, the CPU continuous occupation count of the task is increased by 1, otherwise, the CPU continuous occupation count is set to be 0; that is, c2=c2+1 when c1+.p1, otherwise c2=0;

step 3-1: calculating task busyness

And updating the task busyness according to the C2 value, wherein the task busyness is the percentage of the CPU continuous occupation count of the task to the maximum value of the CPU continuous occupation count, namely, the task busyness=C2/P2.

The method for realizing the AMF overload detection and control is characterized in that a message node of the message node utilization rate is a carrier for transmitting messages among different threads in the AMF, is a limited resource in the AMF and is managed in a linked list mode; after the AMF is started, the number of nodes in the idle message node linked list is the preset total number of message nodes; when the sending task needs to transfer information, a message node is taken out of the linked list, and the process is called as distributing the message node; when the receiving task finishes the information processing, the message node is put back into a linked list, and the process is called releasing the message node; the monitoring of the usage rate of the message nodes is to update the number of idle message nodes when the message nodes are allocated and released; the message node usage is calculated as follows:

step 1-i: sending a task allocation message node or receiving a task release message node;

step 2-ii: calculating the total number of idle message nodes

C3=c3-1 when message nodes are allocated and c3=c3+1 when message nodes are released;

step 3-iii: calculating message node usage

And calculating the utilization rate of the message nodes according to the number of the idle message nodes, wherein the utilization rate of the message nodes is= (P3-C3)/P3.

The method for realizing AMF overload detection and control is characterized in that parameters used for load decision comprise C4, C5, C6 and C7 which correspond to a maximum load previous value, a maximum load current value, a load state previous value and a load state current value respectively; wherein, the value range of C4 and C5 is 0-100; the value range of C6 and C7 is 0-5,0 indicates normal load, 1 indicates overload level 1,2 indicates overload level 2,3 indicates overload level 3,4 indicates overload level 4,5 indicates serious overload;

the load decision algorithm is executed to obtain AMF load state, and the specific calculation steps are as follows:

step 1: calculating the maximum load previous value and the load state previous value, namely C4 and C6;

when the load decision algorithm is executed at the 1 st time, both the maximum load previous value and the load state previous value are initialized to 0, i.e., c4=0 and c6=0;

when the load decision algorithm is executed at the n+1th time, the maximum load previous value and the load state previous value are set as the nth execution result, namely c4=c5 and c6=c7, respectively;

step 2: calculating the current value of the maximum load, namely C5;

the maximum load current value is that the maximum value is selected from all the task busyness and the message node utilization, namely C5=max { the message node utilization, the busyness of the task 1, the busyness of the task 2, … … and the busyness of the task K };

step 3: calculating the current value of the load state, namely C7;

calculating a load state current value by combining the maximum load current value C5, the overload threshold value P4 and the serious overload threshold value P5;

step 3.1: if the maximum load current value is greater than or equal to the severe overload threshold, the load state current value is set to be severe overload, namely C7=5 if C5 is greater than or equal to P5; otherwise, executing the step 3.2;

step 3.2: if the maximum load current value is less than the overload threshold, the load state current value is set to load normal, i.e. c5< P4, c7=0; otherwise, executing the step 3.3;

step 3.3: the current value of the maximum load is between overload and serious overload, namely P4 is less than or equal to C5 is less than or equal to P5, and the current value C7 of the load state is determined as follows:

step 3.3.1: equally dividing the distance 4 between the overload threshold and the serious overload threshold and recording the equal distance as a grading interval, namely grading interval= (P5-P4)/4, and continuing to execute the step 3.3.2;

step 3.3.2: calculating C7 by combining C5, P4 and the grading interval;

step 3.3.2.1: if the current value of the maximum load is greater than or equal to the overload threshold and less than the sum of the overload threshold plus 1 time of the classification interval, the current value of the load state should be set to be the overload level 1, i.e. C5 satisfies p4.ltoreq.c5 < (p4+1×the classification interval), c7=1, otherwise step 3.3.2.2 is performed;

step 3.3.2.2: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 1 time of the grading interval and less than the sum of the overload threshold plus 2 times of the grading interval, the current value of the load state should be set to be the overload level 2, that is, C5 satisfies (p4+1×the grading interval) +.c5 < (p4+2×the grading interval), c7=2, otherwise step 3.3.2.3 is executed;

step 3.3.2.3: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 2 times the grading interval and less than the sum of the overload threshold plus 3 times the grading interval, the current value of the load state should be set to be the overload level 3, i.e. C5 satisfies (p4+2×grading interval) +.c5 < (p4+3×grading interval), c7=3, otherwise step 3.3.2.4 is performed;

step 3.3.2.4: the load state current value is set to overload level 4, i.e., c7=4.

The implementation method of AMF OVERLOAD detection and control is characterized in that the OVERLOAD control uses two interfaces, namely an OVERLOAD START and an OVERLOAD STOP, provided in 3GPP TS38.413, and is characterized in that when AMF is overloaded, a load state current value C7 is mapped to an OVERLOAD action parameter of the OVERLOAD START, and the load state current value C7 is divided into an OVERLOAD level 1, an OVERLOAD level 2, an OVERLOAD level 3, an OVERLOAD level 4 and serious OVERLOAD; overload level 1 corresponds to Reject RRC connection establishment for non-EMERGENCY MO DT, overload level 2 corresponds to Reject RRC connection establishments for Signalling, overload level 3 corresponds to Permit Emergency Sessions and mobile terminated service only, overload level 4 and severe overload both correspond to Permit High Priority Sessions and mobile terminated services only;

when the AMF is overloaded, the OVERLOAD control determines that the AMF sends AN over load START containing AN over load action to the 5G-AN, and instructs the 5G-AN to execute unified access control UAC according to the over load action requirement; when the AMF OVERLOAD state is greatly relieved, the OVERLOAD control decides that the AMF sends AN OVERLOAD STOP to the 5G-AN to instruct the 5G-AN to STOP the unified access control UAC.

The implementation method of AMF overload detection and control is characterized in that the implementation process of the overload control is as follows:

step 1-a: if the current value of the load state is a serious overload and the previous value of the load state is AN overload level 4 or a serious overload, namely C7=5 and 4.ltoreq.C6.ltoreq.5, ignoring all the received traffic but not interacting with the 5G-AN; otherwise, executing the step 2-a;

step 2-a: if the current value of the load state is severely overloaded and the previous value of the load state is less than the OVERLOAD level 4, namely C7=5 and 0 is less than or equal to C6<4, ignoring all the received traffic and sending AN over load START to the 5G-AN; otherwise, executing the step 3-a;

step 3-a: if the current value of the load state is OVERLOAD and the current value of the load state is greater than the previous value of the load state, namely C7>0 and C7> C6, sending AN OVERLOAD START to the 5G-AN; otherwise, executing the step 4-a;

step 4-a: if the current value of the load state is normal and the previous value of the load state is OVERLOAD and the current value of the maximum load is smaller than the difference value between the OVERLOAD threshold and 20, namely C7=0, C6 is larger than or equal to 1 and C5< (P4-20), sending AN OVERLOAD STOP to the 5G-AN; otherwise, executing the step 5-a;

step 5-a: no special treatment is done.

The invention uses the busyness of the task and the utilization rate of the message node as the quantization parameters of the actual load of the AMF, calculates the load state of the AMF through the load monitoring and load decision process, and achieves the aim of OVERLOAD control by mapping the load state of the AMF to the OVERLOAD action and using the OVERLOAD START and the OVERLOAD STOP interface to adjust the service flow sent by the 5G-AN. When the AMF is changed from overload to normal state, the UAC of the 5G-AN is not stopped immediately, but is stopped when the actual load is far below the overload threshold, so that the ping-pong effect in overload control can be avoided.

Drawings

FIG. 1 is AN interactive flow chart for AMF to dynamically adjust 5G-AN traffic;

FIG. 2 is a diagram illustrating the AMF overload detection and control;

FIG. 3 is a schematic diagram of load monitoring of task busyness;

FIG. 4 is a diagram illustrating dynamic changes in an idle message node list;

FIG. 5 is a schematic diagram of load monitoring of message node usage;

FIG. 6 is a diagram of the load decision algorithm performed 2 times before and after;

fig. 7 is a diagram showing an execution process of overload control.

Detailed Description

The invention is further described below with reference to the accompanying drawings:

the invention provides a realization method for AMF OVERLOAD detection and control, which takes the busyness of tasks and the utilization rate of message nodes as two quantization parameters of AMF load to monitor in real time, adopts a load decision algorithm to calculate the load state by combining with preset rated capacity and OVERLOAD threshold, and then dynamically adjusts the service flow of 5G-AN by using AN OVERLOAD START/STOP interface provided in 3GPP TS 38.413. The task in the task busyness is a thread in the AMF, and the task busyness is the CPU occupation condition of the thread in the AMF; the message nodes in the message node utilization rate are carriers for transmitting messages between different tasks (or threads) in the AMF, and the message node utilization rate is the use condition of the message nodes in the AMF.

The interactive flow of AMF dynamically adjusting 5G-AN service flow is shown in figure 1, when the service flow sent by 5G-AN increases suddenly, AMF overload operation is caused, AMF carries out overload detection and control, and then overload including overload action is sent to 5G-AN; and after receiving the overload start, the 5G-AN executes Unified Access Control (UAC) according to overload action, thereby reducing the traffic flow sent to the AMF. When the traffic flow sent by the 5G-AN is obviously reduced, the AMF load state is recovered to be normal, the AMF sends overload stop to the 5G-AN after overload detection and control are carried out, and the 5G-AN stops executing UAC after receiving the overload stop.

The invention discloses an AMF overload detection and control realization method which consists of four parts, namely preset rated capacity and threshold value, load monitoring, load decision and overload control. Fig. 2 depicts the execution process of the overload detection and control of the AMF, and the other three parts are repeatedly executed in the operation process of the AMF except for the preset rated capacity and the threshold value, so as to ensure that the AMF always operates under the rated capacity.

The different parts will be described one by one in the order of execution.

The first part of AMF overload detection and control is to preset rated capacity and threshold values. The preset rated capacity and threshold are parameters listed in table 1, i.e. initializing P1, P2, P3, P4 and P5, which are read from the local configuration or obtained from the network management at the start-up of the AMF. Wherein, P1 and P2 are 2 subdivision parameters of the busyness of the task, which are CPU utilization threshold and CPU continuous occupation count maximum respectively; p1 is used for judging the excessive standard of CPU use in the monitoring unit, and the value is 80; p2 is used for indicating the continuous monitoring times of the CPU utilization rate being too high, and the value is 300; p3 is the total number of message nodes, and is used for indicating the maximum value of the concurrence quantity for information interaction, and the value range is 1536-20480; p4 is an overload threshold, and is used for judging the overload standard of the AMF, and the value is 1-80; p5 is a serious overload threshold, and is used for judging the standard of serious overload of the AMF, and the value is 80; p2 and P3 indicate rated capacities, and P1, P4, and P5 indicate thresholds.

Table 1 rated capacity and threshold parameters

The second part of AMF overload detection and control is load monitoring, and the execution is started after AMF starting is completed. Load monitoring is divided into monitoring of task busyness and monitoring of message node usage, and table 2 lists parameters used in load monitoring. Wherein, C1 is the CPU utilization rate, and the value range is 0-100; c2 is CPU continuous occupied count, and the value range is 0-300; c3 is the number of idle message nodes, and the value range is 0-P3. The AMF comprises a plurality of tasks for processing different service requests, wherein each task corresponds to one thread in the AMF, and a pair of C1 and C2 exists for each task; there is one and only one C3 in the AMF to calculate message node usage.

Table 2 load monitoring parameters

In the present invention, task busyness load monitoring is performed periodically in units of seconds. As shown in fig. 3, the specific steps are described as follows:

step 1-1: calculating the CPU utilization rate of the task, namely C1;

acquiring the total number of the CPU time slices monitored at this time and the number of the task thread CPU time slices from a system file, and then calculating C1, C1=the number of the task thread CPU time slices/the total number of the CPU time slices of each task;

step 2-1: calculating the CPU continuous occupation count of the task, namely C2;

the C2 value is updated according to C1. If the CPU utilization rate of the task is greater than or equal to the CPU utilization rate threshold value, the CPU continuous occupation count of the task is increased by 1, otherwise, the CPU continuous occupation count is set to be 0; that is, c2=c2+1 when c1+.p1, otherwise c2=0;

step 3-1: calculating task busyness

And updating the task busyness according to the C2 value, wherein the task busyness is the percentage of the CPU continuous occupation count of the task to the maximum value of the CPU continuous occupation count, namely, the task busyness=C2/P2.

In the invention, the second part of load monitoring is the utilization rate of message nodes, wherein the message nodes are carriers for transmitting messages between different tasks (or threads) in the AMF, and are a limited resource in the AMF and are managed in a linked list mode. After the AMF is started, the number of nodes in the idle message node linked list is the preset total number of message nodes (i.e., P3), and fig. 4 describes the change process of the idle message node linked list. When the sending task needs to transfer information, a message node is taken out of the linked list, and the process is called as distributing the message node; when the receiving task completes the information processing, the message node should be put back into the linked list, and this process is called releasing the message node. The monitoring of the usage of the message nodes is to update the number of free message nodes when the message nodes are allocated and released. Fig. 5 depicts a message node usage calculation process, and the specific steps are as follows:

step 1-i: sending a task allocation message node or receiving a task release message node;

step 2-ii: calculating the total number of idle message nodes;

c3=c3-1 when message nodes are allocated and c3=c3+1 when message nodes are released;

step 3-iii: calculating the utilization rate of the message nodes;

and calculating the utilization rate of the message nodes according to the number of the idle message nodes, wherein the utilization rate of the message nodes is= (P3-C3)/P3.

And after the load monitoring is completed, obtaining an AMF actual load parameter, and then entering a third part of AMF overload detection and control. The load decision is based on AMF actual load parameters, a load decision algorithm is used for obtaining the maximum load, and then the load state of the AMF is determined by combining a preset overload threshold. Table 3 lists the parameters used in the load decisions: c4, C5, C6, C7 correspond to the maximum load previous value, the maximum load current value, the load state previous value, and the load state current value, respectively. Wherein, the value range of C4 and C5 is 0-100; the values of C6 and C7 are in the range of 0-5,0 indicates that the load is normal, 1 indicates overload level 1,2 indicates overload level 2,3 indicates overload level 3,4 indicates overload level 4, and 5 indicates severe overload.

TABLE 3 load decision parameters

The load decision algorithm is executed to obtain the AMF load state, and fig. 6 describes the execution process of the algorithm for 2 times, and the specific steps are as follows:

step 1: calculating the maximum load previous value and the load state previous value, namely C4 and C6;

when the load decision algorithm is executed at the 1 st time, both the maximum load previous value and the load state previous value are initialized to 0, i.e., c4=0 and c6=0;

when the load decision algorithm is executed at the n+1th time, the maximum load previous value and the load state previous value are set as the nth execution result, namely c4=c5 and c6=c7, respectively;

step 2: calculating the current value of the maximum load, namely C5;

the maximum load current value is that the maximum value is selected from all the task busyness and the message node utilization, namely C5=max { the message node utilization, the busyness of the task 1, the busyness of the task 2, … … and the busyness of the task K };

step 3: calculating the current value of the load state, namely C7;

calculating a load state current value by combining the maximum load current value (C5), the overload threshold value (P4) and the serious overload threshold value (P5);

step 3.1: if the maximum load current value is greater than or equal to the severe overload threshold, the load state current value is set to be severe overload, namely C7=5 if C5 is greater than or equal to P5; otherwise, executing the step 3.2;

step 3.2: if the maximum load current value is less than the overload threshold, the load state current value is set to load normal, i.e. c5< P4, c7=0; otherwise, executing the step 3.3;

step 3.3: the current value of the maximum load is between overload and serious overload (namely P4 is less than or equal to C5 is less than or equal to P5), and the current value C7 of the load state is determined as follows:

step 3.3.1: equally dividing the distance 4 between the overload threshold and the serious overload threshold and recording the equal distance as a grading interval, namely grading interval= (P5-P4)/4, and continuing to execute the step 3.3.2;

step 3.3.2: calculating C7 by combining C5, P4 and the grading interval;

step 3.3.2.1: if the current value of the maximum load is greater than or equal to the overload threshold and less than the sum of the overload threshold plus 1 time of the classification interval, the current value of the load state should be set to be the overload level 1, i.e. C5 satisfies p4.ltoreq.c5 < (p4+1×the classification interval), c7=1, otherwise step 3.3.2.2 is performed;

step 3.3.2.2: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 1 time of the grading interval and less than the sum of the overload threshold plus 2 times of the grading interval, the current value of the load state should be set to be the overload level 2, that is, C5 satisfies (p4+1×the grading interval) +.c5 < (p4+2×the grading interval), c7=2, otherwise step 3.3.2.3 is executed;

step 3.3.2.3: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 2 times the grading interval and less than the sum of the overload threshold plus 3 times the grading interval, the current value of the load state should be set to be the overload level 3, i.e. C5 satisfies (p4+2×grading interval) +.c5 < (p4+3×grading interval), c7=3, otherwise step 3.3.2.4 is performed;

step 3.3.2.4: the load state current value is set to overload level 4, i.e., c7=4.

After the load decision is finished, the fourth part of AMF overload detection and control, namely overload control, is to be performed. The OVERLOAD control of the present invention uses two interfaces, the OVERLOAD START and the OVERLOAD STOP, provided in 3gpp ts38.413, and it is critical that the current value of the load state, i.e. C7, is mapped to the OVERLOAD action parameter of the OVERLOAD START when the AMF is overloaded, and table 4 lists the mapping relationship of C7 to OVERLOAD action. Overload level 1 corresponds to Reject RRC connection establishment for non-overload MO DT, overload level 2 corresponds to Reject RRC connection establishments for Signalling, overload level 3 corresponds to Permit Emergency Sessions and mobile terminated service only, overload level 4 and severe overload all correspond to Permit High Priority Sessions and mobile terminated services only.

TABLE 4 mapping of load State Current values to overload actions

The overload control is based on the result of the load decision, and in combination with the overload threshold, makes a decision as to whether the AMF needs to ignore all traffic received, and whether the AMF needs to interact with the 5G-AN, and adjusts the traffic sent by the 5G-AN to the AMF.

When the AMF is overloaded, the OVERLOAD control determines that the AMF sends AN over load START containing AN over load action to the 5G-AN, and instructs the 5G-AN to execute UAC according to the over load action requirement; when the AMF OVERLOAD condition is greatly relieved, the OVERLOAD control decides that the AMF sends AN OVERLOAD STOP to the 5G-AN to instruct the 5G-AN to STOP the UAC. Fig. 7 depicts the implementation of overload control, and the specific steps are as follows:

step 1-a: if the current value of the load state is a serious overload and the previous value of the load state is AN overload level 4 or a serious overload, namely C7=5 and 4.ltoreq.C6.ltoreq.5, ignoring all the received traffic but not interacting with the 5G-AN; otherwise, executing the step 2-a;

step 2-a: if the current value of the load state is severely overloaded and the previous value of the load state is less than the OVERLOAD level 4, i.e. c7=5 and 0+.c6 <4, then ignoring all traffic received and sending AN OVERLOAD START to the 5G-AN (OVERLOAD action=c7 mapping value); otherwise, executing the step 3-a;

step 3-a: if the current load state value is OVERLOAD and the current load state value is greater than the previous load state value, i.e., C7>0 and C7> C6, then sending AN OVERLOAD START (OVERLOAD action=c7 mapping value) to the 5G-AN; otherwise, executing the step 4-a;

step 4-a: if the current value of the load state is normal and the previous value of the load state is OVERLOAD and the current value of the maximum load is smaller than the difference value between the OVERLOAD threshold and 20, namely C7=0, C6 is larger than or equal to 1 and C5< (P4-20), sending AN OVERLOAD STOP to the 5G-AN; otherwise, executing the step 5-a;

step 5-a: no special treatment is done.

Interpretation of the terms

5GC:5G Core, 5 th generation mobile communication technology Core network

5G-AN:5G Access Network,5G access network

AMF: access and Mobility Management Function Access and mobility management functionality

CPU: central Processing Unit, central processing unit

IPv4: internet Protocol version 4 Internet protocol fourth edition

NAS: non-Access Stratum

NF: network Function, network Function

NGAP: NG Application Protocol New generation application protocol

NRF: network Repository Function network storage function

UAC: unified Access Control unified access control

UE: user Equipment

Claims (8)

1. An implementation method for AMF overload detection and control is characterized in that: the method comprises the steps of performing real-time monitoring by taking task busyness and message node utilization rate as quantization parameters of AMF actual load, combining preset rated capacity and OVERLOAD threshold, calculating the load state of AMF through load monitoring and load decision process, mapping the AMF load state onto OVERLOAD action, and adjusting the service flow of 5G-AN through AN OVERLOAD START and OVERLOAD STOP interface to realize OVERLOAD control;

when the AMF is converted from overload to normal state, the unified access control UAC of the 5G-AN is not stopped immediately, but is stopped when the actual load is lower than the overload threshold;

the task in the task busyness is a thread in the AMF, and the task busyness is the CPU occupation condition of the thread in the AMF; the message nodes in the message node utilization rate are carriers for transmitting messages between different threads in the AMF, and the message node utilization rate is the use condition of the message nodes in the AMF;

the task busyness load monitoring is carried out periodically in units of every second, and comprises the following specific steps:

step 1-1: calculating the CPU utilization rate of the task, namely C1;

acquiring the total number of the CPU time slices monitored at this time and the number of the task thread CPU time slices from a system file, and then calculating C1, C1=the number of the task thread CPU time slices/the total number of the CPU time slices of each task;

step 2-1: calculating the CPU continuous occupation count of the task, namely C2;

updating the C2 value according to the C1; if the CPU utilization rate of the task is greater than or equal to the CPU utilization rate threshold value, the CPU continuous occupation count of the task is increased by 1, otherwise, the CPU continuous occupation count is set to be 0; that is, c2=c2+1 when c1+.p1, otherwise c2=0;

step 3-1: calculating task busyness

According to the C2 value, updating the task busyness, wherein the task busyness is the percentage of the maximum value of the CPU continuous occupation count and the CPU continuous occupation count of the task, namely, the task busyness=C2/P2;

the message node of the utilization rate of the message node is a carrier for transmitting messages among different threads in the AMF, is a limited resource in the AMF and is managed in a linked list mode; after the AMF is started, the number of nodes in the idle message node linked list is the preset total number of message nodes; when the sending task needs to transfer information, a message node is taken out of the linked list, and the process is called as distributing the message node; when the receiving task finishes the information processing, the message node is put back into a linked list, and the process is called releasing the message node; the monitoring of the usage rate of the message nodes is to update the number of idle message nodes when the message nodes are allocated and released; the message node usage is calculated as follows:

step 1-i: sending a task allocation message node or receiving a task release message node;

step 2-ii: calculating the total number of idle message nodes

C3=c3-1 when message nodes are allocated and c3=c3+1 when message nodes are released;

step 3-iii: calculating message node usage

And calculating the utilization rate of the message nodes according to the number of the idle message nodes, wherein the utilization rate of the message nodes is= (P3-C3)/P3.

2. The method for implementing AMF overload detection and control according to claim 1, wherein: the interactive flow of AMF dynamically adjusting the service flow of 5G-AN is as follows, when the service flow sent by 5G-AN increases suddenly, AMF overload operation is caused, AMF carries out overload detection and control, and then sends overload start containing overload action to 5G-AN; after receiving overload, the 5G-AN executes unified access control UAC according to overload action, thereby reducing the service flow sent to the AMF; when the service flow sent by the 5G-AN is obviously reduced, the AMF load state is recovered to be normal, the AMF sends overload stop to the 5G-AN after overload detection and control are carried out, and the 5G-AN stops executing unified access control UAC after receiving the overload stop.

3. The method for implementing AMF overload detection and control according to claim 1, wherein: the method comprises four parts, namely presetting rated capacity and a threshold value, load monitoring, load decision-making and overload control, wherein the execution process of AMF overload detection and control is that the three parts of load monitoring, load decision-making and overload control are repeatedly executed continuously in the AMF operation process except the preset rated capacity and threshold value, so that the AMF is ensured to be operated always under the rated capacity;

the preset rated capacity and the threshold value are related parameters read from local configuration or acquired from a network manager when AMF is started;

the load monitoring is carried out after the AMF is started, and is divided into monitoring of task busyness and monitoring of the utilization rate of message nodes;

the load decision is based on AMF actual load parameters, a load decision algorithm is used for obtaining the maximum load, and then the load state of the AMF is determined by combining with a preset overload threshold;

the overload control is based on the result of the load decision, and in combination with the overload threshold, a policy is made as to whether the AMF needs to ignore all the received traffic, and whether the AMF needs to interact with the 5G-AN, so as to adjust the traffic sent by the 5G-AN to the AMF.

4. The method for detecting and controlling AMF overload according to claim 3, wherein said related parameters are P1, P2, P3, P4 and P5, P1 and P2 are 2 sub-division parameters of task busyness, which are CPU utilization threshold and CPU continuous occupation count maximum respectively; p1 is used for judging the excessive standard of CPU usage in the monitoring unit; p2 is used for indicating the continuous monitoring times of the CPU utilization rate; p3 is total number of message nodes, and is used for indicating maximum value of concurrency quantity for information interaction; p4 is an overload threshold and is used for judging the overload standard of the AMF; p5 is a serious overload threshold, and is a standard for judging serious overload of the AMF; p2 and P3 indicate rated capacities, and P1, P4, and P5 indicate thresholds.

5. A method for implementing AMF overload detection and control according to claim 3, wherein said parameters used in load monitoring comprise: c1 is CPU utilization, C2 is CPU continuous occupation count, and C3 is idle message node number; the AMF comprises a plurality of tasks for processing different service requests, wherein each task corresponds to one thread in the AMF, and a pair of C1 and C2 exists for each task; there is one and only one C3 in the AMF to calculate the message.

6. The method for detecting and controlling AMF overload according to claim 3, wherein said parameters used for load decision include C4, C5, C6, C7, corresponding to maximum load previous value, maximum load current value, load state previous value, and load state current value, respectively; wherein, the value range of C4 and C5 is 0-100; the value range of C6 and C7 is 0-5,0 indicates normal load, 1 indicates overload level 1,2 indicates overload level 2,3 indicates overload level 3,4 indicates overload level 4,5 indicates serious overload;

the load decision algorithm is executed to obtain AMF load state, and the specific calculation steps are as follows:

step 1: calculating the maximum load previous value and the load state previous value, namely C4 and C6;

when the load decision algorithm is executed at the 1 st time, both the maximum load previous value and the load state previous value are initialized to 0, i.e., c4=0 and c6=0;

when the load decision algorithm is executed at the n+1th time, the maximum load previous value and the load state previous value are set as the nth execution result, namely c4=c5 and c6=c7, respectively;

step 2: calculating the current value of the maximum load, namely C5;

the maximum load current value is that the maximum value is selected from all the task busyness and the message node utilization, namely C5=max { the message node utilization, the busyness of the task 1, the busyness of the task 2, … … and the busyness of the task K };

step 3: calculating the current value of the load state, namely C7;

calculating a load state current value by combining the maximum load current value C5, the overload threshold value P4 and the serious overload threshold value P5;

step 3.1: if the maximum load current value is greater than or equal to the severe overload threshold, the load state current value is set to be severe overload, namely C7=5 if C5 is greater than or equal to P5; otherwise, executing the step 3.2;

step 3.2: if the maximum load current value is less than the overload threshold, the load state current value is set to load normal, i.e. c5< P4, c7=0; otherwise, executing the step 3.3;

step 3.3: the current value of the maximum load is between overload and serious overload, namely P4 is less than or equal to C5 is less than or equal to P5, and the current value C7 of the load state is determined as follows:

step 3.3.1: equally dividing the distance 4 between the overload threshold and the serious overload threshold and recording the equal distance as a grading interval, namely grading interval= (P5-P4)/4, and continuing to execute the step 3.3.2;

step 3.3.2: calculating C7 by combining C5, P4 and the grading interval;

step 3.3.2.1: if the current value of the maximum load is greater than or equal to the overload threshold and less than the sum of the overload threshold plus 1 time of the classification interval, the current value of the load state should be set to be the overload level 1, i.e. C5 satisfies p4.ltoreq.c5 < (p4+1×the classification interval), c7=1, otherwise step 3.3.2.2 is performed;

step 3.3.2.2: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 1 time of the grading interval and less than the sum of the overload threshold plus 2 times of the grading interval, the current value of the load state should be set to be the overload level 2, that is, C5 satisfies (p4+1×the grading interval) +.c5 < (p4+2×the grading interval), c7=2, otherwise step 3.3.2.3 is executed;

step 3.3.2.3: if the current value of the maximum load is greater than or equal to the sum of the overload threshold plus 2 times the grading interval and less than the sum of the overload threshold plus 3 times the grading interval, the current value of the load state should be set to be the overload level 3, i.e. C5 satisfies (p4+2×grading interval) +.c5 < (p4+3×grading interval), c7=3, otherwise step 3.3.2.4 is performed;

step 3.3.2.4: the load state current value is set to overload level 4, i.e., c7=4.

7. A method for implementing AMF OVERLOAD detection and control according to claim 3, wherein said OVERLOAD control uses two interfaces, namely OVERLOAD START and OVERLOAD STOP, provided in 3gpp ts38.413, and is characterized in that when AMF is overloaded, the load state current value C7 is mapped to an OVERLOAD action parameter of OVERLOAD START, and the load state current value C7 is divided into OVERLOAD level 1, OVERLOAD level 2, OVERLOAD level 3, OVERLOAD level 4, and serious OVERLOAD; overload level 1 corresponds to Reject RRC connection establishment for non-EMERGENCY MO DT, overload level 2 corresponds to Reject RRC connection establishments for Signalling, overload level 3 corresponds to Permit Emergency Sessions and mobile terminated service only, overload level 4 and severe overload both correspond to Permit High Priority Sessions and mobile terminated services only;

when the AMF is overloaded, the OVERLOAD control determines that the AMF sends AN over load START containing AN over load action to the 5G-AN, and instructs the 5G-AN to execute unified access control UAC according to the over load action requirement; when the AMF OVERLOAD condition is relieved, the OVERLOAD control decides that AMF sends AN OVERLOAD STOP to 5G-AN to instruct 5G-AN to STOP unified access control UAC.

8. The method for implementing AMF overload detection and control according to claim 7, wherein said overload control is performed as follows:

step 1-a: if the current value of the load state is a serious overload and the previous value of the load state is AN overload level 4 or a serious overload, namely C7=5 and 4.ltoreq.C6.ltoreq.5, ignoring all the received traffic but not interacting with the 5G-AN; otherwise, executing the step 2-a;

step 2-a: if the current value of the load state is severely overloaded and the previous value of the load state is less than the OVERLOAD level 4, namely C7=5 and 0 is less than or equal to C6<4, ignoring all the received traffic and sending AN over load START to the 5G-AN; otherwise, executing the step 3-a;

step 3-a: if the current value of the load state is OVERLOAD and the current value of the load state is greater than the previous value of the load state, namely C7>0 and C7> C6, sending AN OVERLOAD START to the 5G-AN; otherwise, executing the step 4-a;

step 4-a: if the current value of the load state is normal and the previous value of the load state is OVERLOAD and the current value of the maximum load is smaller than the difference value between the OVERLOAD threshold and 20, namely C7=0, C6 is larger than or equal to 1 and C5< (P4-20), sending AN OVERLOAD STOP to the 5G-AN; otherwise, executing the step 5-a;

step 5-a: no special treatment is done.