KR940001698B1 - Overload control method of electronic exchange - Google Patents

Abstract

The CPU occupation time of main processor is measured to detect an overload of a system and a service function of main processor is restricted according to the detected load condition. The method comprises the steps op: (a) counting a reference clock signal transmitted from an operating system to set a time unit by 60 seconds; (b) analyzing the load condition by receiving a CPU pause time data transmitted from an operating system and a time unit and repeating the step (a) when load condition is normal; and (c) analizing a current state and an one step ahead state of the load and processing the load according to a grade of the load condition.

Description

System overload control method using CPU share of electronic switch as state variable

1 is a block diagram of an all-electronic exchange processor to which the present invention is applied.

2 is a flowchart of a system overload control method according to the present invention.

3 is a flowchart of a CPU idle time detection control method according to the present invention.

4 is a flowchart of a method of changing an execution period according to the present invention.

* Explanation of symbols for main parts of the drawings

1: MPSH 2: OMP

3: MMP 4: NTP

5: INP 6: ASP

7: input / output device

The present invention relates to a system overload control method for controlling the load of the processor CPU (Central Processing Unit) in order to satisfy the continuous and stable service of the system among the total input load on the processor of the system.

In general, the overload is in the range of general overload, such as the time when the holiday discount rate starts, local overload where the call volume is concentrated in a specific area due to natural disasters, accidents, etc. It is classified as overload according to the cause and emergency overload, disaster overload, holiday overload, planned overload. The rapid increase in visibility due to the above factors causes service inability and service quality deterioration due to the lack of system available resources, which becomes incomplete, that is, invalid calls, which unnecessarily occupies system resources, further deteriorating the performance of the system and incomplete. Possitive feedback results in a more rapid increase in retries.

The function of monitoring the load by detecting the maximum value between the loads according to the degree of overload should be performed with the minimum load, and by the overload control method, the effect of overload should not spread to the whole system. It should be easy to determine and change overload thresholds and control variables to cope with system performance, installation location and environment.

Conventional overload control states the feedforward method and the length of the queue as the state variable with the input load limiting the input load by fixing the limit input load so that the system can maintain the processor occupancy at an appropriate level even in the case of overload. The feedback method was used as a variable, but since the number of input calls and the queue length were monitored, when the overload was determined, there was a problem that the performance of the system was reduced since the overload was already processed by the system.

The present invention devised to eliminate the above problems is to measure the CPU occupancy time of each MP (Main Processor) in the decentralized exchange system to control the service of the processor according to the load generation when a load above a certain threshold occurs Its purpose is to provide a system overload control method.

In order to achieve the above object, the present invention provides an NTP (Number) directly related to an operation and maintenance processor (Manipulation and Maintenance Processor) and MMP (Man Machine Processor), a load state and a call processing function in charge of maintenance and operation functions. In the system and load control method using the CPU occupancy of state of the electronic switchgear including a Translation Processor, an INP (Inter Network Processor), and an Access Switching Processor (ASP) as a state variable, a reference time from an OS (Operation System) A first step of calculating (40.960 microseconds) to calculate the unit time for calculating the CPU occupancy is 60 seconds, receiving a predetermined CPU idle time from the OS to analyze the load state by the CPU idle time and unit time If it is normal, the second step to continuously repeat the first step, and if the load is analyzed by the CPU idle time and unit time abnormality Analyzing current and previous measurements load state is performed by a third step of processing by dividing the rating.

Hereinafter, with reference to the accompanying drawings will be described an embodiment of the present invention;

1 is a configuration diagram of an electronic switch processor to which the present invention is applied, 1 is a Main Processor Supervision Handler (MPSH), 2 is an Operation and Maintenance Processor (OMP), 3 is a Machine Machine Processor (MMP), and 4 is a NTP (Number). Translation processor (5), 5 represents INP (Inter Network Processor), 6 represents ASP (Access Switching Processor), and 7 represents input / output device (CRT / TTY).

The all-electronic exchange processor to which the present invention is applied includes an OMP (2) which is in charge of maintenance functions as shown in FIG. 1, and an MMP (3), and an NTP (4) directly related to load conditions and call processing functions. , An INP 5, and an ASP 6, and the MPSH 1 software block of the present invention for overload detection and control is mounted in each of the processors to measure and control the load generated for each processor. In the event of an overload condition, the load is continuously monitored to take measures to reduce the load or output to the input / output device (7).

The overload detection target is all resources that can act as an overload factor, i.e., CPU occupancy, IPC occupancy, processor generation rate, and processor making up all the exchanges. It is to limit the functions performed in the Therefore, the control function periodically processes the overload generation information to determine what kind of overload condition has occurred.

2 is a flowchart of a system overload control method using the CPU occupancy as a state variable according to the present invention.

After receiving the base time of the OS primitive 40.960 microseconds and calculating the unit time for calculating the CPU occupancy is about 60 seconds (8), and receives 60 seconds of CPU downtime from the OS primitive loadbox (9) ). The load state is analyzed by the CPU idle time and the measured unit time of each processor (10) .If the load state is normal, the time is continuously calculated and the load is analyzed and monitored continuously. 12).

The load value is calculated by subtracting the CPU time per measurement time from the unit time by dividing the time value received from the C language of the OS with CHILL (CCITT High-Level Language) language, and dividing it by the measurement time as a percentage. The percentage load value is applied to the overload stage to suit the system characteristics to which it applies.

FIG. 3 is a flowchart of a method for monitoring and monitoring a CPU idle time according to the present invention and shows processes 10 and 12 of performing load analysis and overload control of FIG.

Receive the CPU idle time from the OS (13) and compare and analyze the load status (14) and search if the current load measurement status is different from the previous status more than three times (15), and if it is different, register the load status (16) After outputting the change details (17), control is performed according to the load state (18). If the current state and the previous state are the same, the load is continuously monitored (19).

4 is a flowchart of a method of changing an execution cycle according to the present invention, and shows an overload control process 18 of FIG.

In each MP, the load measurement is started to determine when a load is detected (20), and a class is determined (21), and when each class is determined, a corresponding action is taken for each class (22, 23, 24, 25). The levels of maintenance and operation are classified into Normal, Minor, Major, and Critical, and the actions according to the load rating provide continuous service in the normal state. (22) In the general state, limit the statistical functions not related to the test, periodic monitoring and testing by the operator's request (23), and in the case of the main state, the general function defined in maintenance, operation and call processing. Limit performance (24). In addition, actions according to the call processing function class handled in the major or emergency state have a great effect on the system, so they are defined and applied separately. In addition, the measured load check of the processor is performed every 60 seconds in the normal state (22), 50 seconds in the normal state (23), 40 seconds in the main state (24), 30 seconds in the emergency state Changes are registered in the timer (25) to be performed periodically (26) to check the system for instability early.

The present invention configured and operated as described above has the effect of providing a continuous and stable service by analyzing the load of the processor CPU to prevent a phenomenon that adversely affects the system performance and service, and to take step-by-step actions when an overload occurs.

Claims (7)

Operation and Maintenance Processor (2) and Man Machine Processor (3), responsible for maintenance and operation functions, and Number Translation Processor (NTP) directly related to load conditions and call processing functions. (4) An OS (Operation System) in a system overload control method in which the CPU occupancy rate of the electronic switchgear including the INP (Inter Network Processor) 5 and the Access Switching Processor (ASP) 6 is a state variable. Receiving a reference time (40.960 microseconds) from the first step of calculating a unit time for calculating CPU occupancy to 60 seconds; A second step of receiving a predetermined CPU idle time from the OS and analyzing a load state by the CPU idle time and a unit time and continuously repeating the first step if it is normal; And a third step of analyzing the load state by the CPU idle time and the unit time, and analyzing the current and previous load measurement state by dividing the class by grade in case of abnormality. System overload control method, characterized in that carried out by. The method of claim 1, wherein the third step comprises: returning to the step of calculating the unit time if the current load measurement state and the previous load measurement state are the same; Comparing the current load measurement state with the previous load measurement state and registering a load state when the number of times differs more than a predetermined number of times, and outputting a load change history; Determining a rating according to the load condition and performing a corresponding action; System and load control method, characterized in that carried out by. The method according to claim 1 or 2, wherein the CPU down time is 60 seconds. The method of claim 2, wherein the second step of registering the load state and requesting the load change history by comparing the current load measurement state with the previous load measurement state is different from the current and previous load measurement states continuously three or more times. System overload control method, characterized in that performed if. 3. The method of claim 2, wherein the third step of determining the rating according to the load condition and performing the corresponding action comprises: determining the rating according to the detected load condition; Changing the monitoring period according to the class; Registering the monitoring period with a timer; System overload control method, characterized in that carried out by. 6. The method according to claim 1, 2 or 5, wherein the class is classified into a normal state, a general state, a main state, and an emergency state according to the detected load state. 7. The system overload control method according to claim 6, wherein the monitoring period is 60 seconds in a normal state, 50 seconds in a normal state, 40 seconds in a main state, and 30 seconds in an emergency state.

Images