CN103178990A - A network equipment performance monitoring method and network management system - Google Patents

Abstract

The invention discloses a network device performance monitoring method and a network management system. The network device performance monitoring method includes: the network management system acquires key performance indicator (KPI) data of a network device periodically; the acquired KPI data are summarized according to set time windows to obtain KPI data summarizing values of the time windows; a linear regression algorithm model is utilized, KPI data of future N time windows including the current time window are predicted according to KPI data summarizing values of N time windows before the current window; whether the KPI data of the further N time windows exceed a set alarm threshold value is judged; and a alarm is given out during yes judgment. The network device performance monitoring method and the network management system can achieve early warning according to KPI historical data.

Description

A network equipment performance monitoring method and network management system

technical field

The invention relates to network management technology in the communication field, in particular to a network equipment performance monitoring method and a network management system.

Background technique

During the booming development of the telecom industry, the business support system has played an increasingly important role in customer service, service provisioning, service guarantee, billing and accounting, forecasting and planning, etc. With the needs of market development, new With the continuous introduction of business, the business support system is constantly surpassing itself with the growth of the company, and the basic node that supports the normal operation of the system is the mainframe, and the security problems of the mainframe are becoming increasingly serious.

The traditional host performance monitoring method uses the KPI (KPI is called KeyPerformance Indicator, which is the key performance indicator, such as CPU usage, memory usage, IO throughput, hard disk usage, database usage, etc.) to monitor the running of the host. ) to set the threshold value, when the system monitors that the KPI index exceeds the threshold value, it will issue an active warning. The advantage of this method is that it is simple and easy to implement.

From the perspective of long-term operation and maintenance practice, the existing technology has its obvious limitations, which are mainly reflected in the following aspects:

(1) The traditional monitoring method based on the threshold value does not take into account the changing trend of the system KPI index. During the operation of the system, if the KPI index changes suddenly but has not reached the set threshold value, no alarm will be generated. , and at this time, it actually needs to attract the attention of the system maintenance personnel, and it is necessary to take proactive intervention measures to prevent the system KPI indicators from further rising.

(2) The traditional threshold-based monitoring method is passive monitoring, and the alarm is issued after a situation occurs. At this time, the normal operation of the system may have been affected, and the purpose of active preventive monitoring cannot be achieved.

In short, the traditional host performance monitoring method adopts the monitoring method based on the threshold value. Once the system KPI index exceeds the threshold value, the system is often already in an overloaded state. An alarm at this time is difficult for customers to deal with, and the risk of system operation is relatively high.

Contents of the invention

Embodiments of the present invention provide a network device performance monitoring method and a network management system, which are used to implement early warning based on historical data of KPI performance indicators.

The network device performance monitoring method provided by the embodiment of the present invention includes:

The network management system periodically collects KPI performance index data of network equipment;

The network management system summarizes the collected KPI performance index data according to the set time window, and obtains the KPI performance index data summary value of each time window;

The network management system utilizes a linear regression algorithm model, and predicts the KPI performance index data of N time windows in the future including the current time window according to the summary value of KPI performance index data of N time windows before the current time window;

The network management system judges whether the KPI performance index data of the N time windows in the future exceeds a set alarm threshold, and sends out an alarm if it is judged to be yes.

The network device performance monitoring device provided by the embodiment of the present invention includes:

The collection module is used to periodically collect KPI performance index data of network equipment;

The summary module is used to summarize the collected KPI performance index data according to the set time window, and obtain the KPI performance index data summary value of each time window;

A prediction module, for utilizing the linear regression algorithm model, and predicting the KPI performance index data of N time windows in the future including the current time window according to the summary value of the KPI performance index data of N time windows before the current time window;

The alarm module is used to judge whether the KPI performance index data in the N future time windows exceeds the set alarm threshold, and issue an alarm when it is judged to be yes.

In the above-mentioned embodiment of the present invention, by summarizing the collected KPI performance index data according to the set time window, and using the linear regression algorithm model, according to the summary value of the KPI performance index data of N time windows before the current time window, Predict the KPI performance index data of N time windows in the future including the current time window, so as to use the historical KPI performance index data to predict the future KPI performance index data, and give an alarm according to the forecast situation, so as to issue an early warning before the actual problem occurs , in order to realize the pre-analysis processing of the system.

Description of drawings

FIG. 1 is a schematic diagram of a network device performance monitoring process provided by an embodiment of the present invention;

Fig. 2 and Fig. 3 are respectively the medium and long-term KPI performance index data prediction schematic diagrams in the embodiment of the present invention;

Fig. 4 and Fig. 5 are respectively the long-term KPI performance index data prediction schematic diagram in the embodiment of the present invention;

FIG. 6 is a schematic structural diagram of a network management system provided by an embodiment of the present invention.

Detailed ways

Aiming at the problems existing in the prior art, the embodiment of the present invention predicts the usage of system resources in a period of time in the future according to the monitored KPI performance index data, and performs active monitoring and alarm according to the predicted situation, so as to issue a warning before the actual problem occurs. Early warning, so as to realize the pre-analysis and processing of the system.

The KPI performance index data mentioned here may include one or any combination of various parameters that can characterize the performance of network equipment or business systems, such as CPU usage, memory usage, and network throughput.

Embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Referring to FIG. 1 , it is a schematic diagram of a network device performance monitoring process provided by an embodiment of the present invention. The process can be implemented by a network management system, and the process can include:

Step 101, periodically collect KPI performance index data of network equipment.

During specific implementation, various KPI performance index data can be collected from traditional monitoring systems.

Step 102, summarizing the collected KPI performance index data according to the set time windows to obtain the summary value of the KPI performance index data in each time window.

During specific implementation, the peak value of the KPI performance index data collected in a time window may be used as the summary value of the KPI performance index data in the time window. When there are multiple data statistics periods in a time window, average the KPI performance indicator data peak values collected in each data statistics period in a time window, and then average the KPI performance indicator data peak values in each data collection period The average value of is used as the summary value of KPI performance indicator data in this time window.

Step 103, using the linear regression algorithm model, and according to the summary value of the KPI performance index data of N time windows before the current time window, predicting the KPI performance index data of N future time windows including the current time window. The N time windows here refer to consecutive N time windows.

Step 104, outputting the predicted KPI performance index data of N time windows in the future.

Further, the above process also includes:

Step 105, judging whether the KPI performance index data of N time windows in the future meet the warning condition, and issuing a warning if it is judged to be yes. There is no strict timing requirement between this step and step 104 .

During specific implementation, if the KPI performance index data collected in step 101 includes multiple types, such as including CPU usage and memory usage, then in this step, the alarm conditions of the future N time windows need to be adjusted according to the CPU usage alarm condition. It is judged whether the CPU usage needs an alarm, and according to the alarm condition of the memory usage, it is judged whether the memory usage of N time windows in the future needs an alarm, and an alarm is issued according to the judgment result. Among them, according to the alarm conditions to determine whether to alarm, there are the following implementation methods:

Mode 1: If any data in the KPI performance index data of N time windows in the future exceeds the threshold of the KPI performance index (the threshold is a preset fixed value), an alarm is issued;

Method 2: The judgment method is the same as Method 1, but the threshold of the KPI performance index is based on the linear regression of the KPI performance index data of N time windows in the future and the summary value of the KPI performance index data of N time windows before the current time window The slope of the algorithm model is dynamically adjusted based on the preset threshold;

Mode 3: If there is data in the KPI performance index data of N time windows in the future that exceeds the threshold of the KPI performance index (the threshold can be preset or dynamically adjusted as in Mode 2), and the slope of the linear regression algorithm model satisfies If the condition is set, an alert is issued.

Method 4: Combine the above methods.

According to the length of the time window setting, the embodiment of the present invention can realize near-real-time pre-monitoring and alarming, mid- and long-term pre-monitoring and alarming, and long-term pre-monitoring and alarming.

The prediction period of the near-live pre-monitoring alarm is short, and the performance of the network equipment can be predicted in time to detect problems as early as possible. The length of the time window can be set between 10 and 20 minutes, such as 10 minutes, 15 minutes or 20 minutes, preferably 15 minutes, and the corresponding time window length can also be set according to different types of KPIs. The time window length of 10 to 20 minutes is an empirical value obtained from long-term operation and maintenance work, and the time window value that has the best effect on the prediction result obtained through repeated tests.

In the near-live pre-monitoring process, when counting the summary value of KPI performance index data in N time windows before the current time window, the peak value of KPI performance index data collected in each time window can be used as the KPI of the corresponding time window Performance metrics data summary value.

When predicting the KPI performance index data of N time windows in the future, according to the summary value of KPI performance index data of N time windows before the current time window, the serial number of the time window can be used as an independent variable, and the summary value of KPI performance index data can be used as a factor variables to obtain a linear regression equation; and then predict the KPI performance index data of N time windows in the future according to the equation.

When judging whether the KPI performance index data of N time windows in the future meets the alarm condition of the corresponding KPI performance index, the KPI performance index threshold value can be based on the KPI performance index data of N time windows in the future and the N time window before the current time window. The linear regression algorithm model slope of the KPI performance index data summary value of each time window is obtained by adjusting the preset KPI performance index threshold. The determination of KPI early warning usually needs to consider the threshold and slope. For example, the following alarm rules can be set:

(VALUE＞90)OR(VALUE＞80 AND SLOPE＞0.5)

This rule means: if the predicted value exceeds 90% (threshold 1), or if the predicted value exceeds 80% (threshold 2) and the trend slope is greater than 0.5, an alarm will be issued.

Furthermore, the values of these values can be continuously adjusted and optimized during use, mainly based on the following principles:

(1) If a monitoring alarm is generated in the original monitoring system, but no early warning is generated in this early warning system, it is necessary to analyze the KPI data before the alarm is generated, and appropriately reduce the absolute threshold or slope threshold.

(2) If a large number of early warnings are generated, and most of them are false alarms, the absolute threshold or the slope threshold needs to be increased.

(3) The ideal situation is that 80% of the monitoring alarms in the original monitoring system will have an early warning within 1-2 hours before they are generated, and the accuracy of the early warning should reach more than 70%.

The prediction cycle of medium and long-term pre-monitoring and alarm is moderate, which can predict the performance of network equipment in time without making predictions and alarms too frequently like near-live pre-monitoring. The length of the time window can be set to 1 day or several days, such as 1 day, 2 days or 5 days, preferably 1 day, and the corresponding time window length can also be set according to different types of KPIs. The 1-day time window length is an empirical value obtained from long-term operation and maintenance work, and the time window value that has the best effect on the prediction results obtained through repeated tests.

In the medium and long-term pre-monitoring process, when counting the summary value of KPI performance index data in N time windows before the current time window, the peak values of KPI performance index data collected in each data statistical period in a time window can be averaged , and then take the average value of the KPI performance index data peak values in each data collection period as the summary value of the KPI performance index data in this time window.

When predicting the KPI performance index data of N time windows in the future, in the N time windows before the current time window, the KPI performance index data of each time window can be subtracted from the KPI performance index of the previous time window Summarize the data to obtain an array containing N-1 incremental values; use the linear regression algorithm model, and according to the array, calculate the slope of the linear regression algorithm model of the N-1 incremental values (that is, the slope of the linear regression algorithm model in the array The element number is used as the dependent variable, and the corresponding element value is used as the independent variable to obtain ) in the linear regression equation; use the linear regression algorithm model, and calculate the future N time windows including the current time window according to the slope The incremental value of the KPI performance index data compared with the previous time window; according to the incremental value of the KPI performance index data of each time window in the next N time windows, and the KPI performance index data of the previous time window The predicted value is to obtain the predicted value of the KPI performance index data of each time window in the future N time windows; wherein, the predicted value of the KPI performance index data of the first time window in the future N time windows is the corresponding increment The sum of the value and the summary value of the KPI performance indicator data in the previous time window.

When judging whether the KPI performance indicator data in the next N time windows meets the alarm conditions of the corresponding KPI performance indicator, if the incremental slope is greater than 0 (indicating a steep upward trend), an alarm will be generated or the predicted data will be marked in a marked manner , hoping to get the attention of the network administrator.

Long-term pre-monitoring alarms have a long forecast period, and can predict the performance of network equipment for a long period of time in the future, so that corresponding processing strategies can be adopted according to the KPI change trend. The length of the time window can be set to one month or several months, preferably one month. Of course, the corresponding time window length can be set according to different types of KPIs. The 1-month time window length is an experience value obtained from long-term operation and maintenance work, and the time window value that has the best effect on the prediction results obtained through repeated tests.

In the long-term pre-monitoring process, when counting the summary value of KPI performance index data in N time windows before the current time window, the peak value of KPI performance index data collected in each data statistical cycle in a time window can be averaged, Then, the average value of the peak values of the KPI performance index data in each data collection cycle is used as the summary value of the KPI performance index data in the time window.

When predicting the KPI performance index data of N time windows in the future, according to the summary value of KPI performance index data of N time windows before the current time window, the serial number of the time window can be used as an independent variable, and the summary value of KPI performance index data can be used as a factor variables to obtain a linear regression equation; and then predict the KPI performance index data of N time windows in the future according to the equation.

For long-term pre-monitoring, there is no need to set up an alarm mechanism. Because long-term pre-monitoring is mainly used to check long-term performance trends, it can be used as a reference for capacity analysis. For example, if the CPU has been at a high level for a long time and has a slow upward trend, whether to consider hardware expansion.

The implementation process of the near-live and near-real-time pre-monitoring alarm, the mid- and long-term pre-monitoring alarm, and the long-term pre-monitoring alarm will be described below in conjunction with specific examples.

Example 1: Near real-time pre-monitoring alarm

Taking the monitoring of the CPU utilization rate of network equipment and the time window length as 15 minutes as an example, the near-live pre-monitoring and alarm process may include:

Data is collected from traditional monitoring systems every 5 to 15 minutes, such as every 5, 10, or 15 minutes. Each time window is 15 minutes, so that one or more CPU usages of a network device can be collected in each time window, and the maximum value of the CPU usage collected in each time window is taken as the time window Summary value of CPU usage. Then take 8 time windows forward from the current time window, that is, take the CPU usage rate of the previous 2 hours, and the summary value of the CPU usage rate of the 8 time windows can be shown in Table 1.

Table 1

time CPU usage (%) 20:00 73.2 20:15 72.1 20:30 74.3 20:45 74.6 21:00 75.1 21:15 75.8 21:30 78.3 21:45 77.2

The CPU usage summary value of these 8 time windows is used to predict the CPU usage forecast value of the next 8 time windows:

Process the CPU usage of these 8 time windows into the form of independent variable X and dependent variable Y, as shown in Table 2:

Table 2

x Y 1 73.2 2 72.1 3 74.3 4 74.6 5 75.1 6 75.8 7 78.3 8 77.2

The linear regression formula is y=a*x+b, and the values of a and b are calculated respectively according to Table 2 before forecasting. The embodiment of the present invention adopts the model of the LineRegression of Weka to calculate, can replace with two functions in Excel here:

a=SLOPE(dependent variable array, independent variable array)

b=INTERCEPT(dependent variable array, independent variable array)

According to the above linear regression algorithm model and the data in Table 2, b=71.65 and a=0.7619 were calculated respectively.

Then according to the formula y=a*x+b, the CPU utilization rate in the next 8 time windows can be predicted. Specifically, the CPU utilization of the ninth time window is:

Y ₉ ＝a*x+b＝0.7619*9+71.65＝78.50 (approximate value, keep 2 digits after the decimal point)

The CPU utilization of the 10th time window is:

Y ₁₀ ＝a*x+b＝0.7619*10+71.65＝78.27 (approximate value, keep 2 decimal places)

By analogy, the CPU utilization of the 9th to 18th time windows can be predicted, as shown in Table 3:

table 3

x Y 9 78.50 10 79.27

11 80.03 12 80.79 13 81.55 14 82.31 15 83.08 16 83.84 17 84.60 18 85.36

Determine whether the CPU usage threshold needs to be adjusted according to the preset CPU usage threshold adjustment policy. 3 prediction data to determine whether an alarm is required.

For example, the preset CPU usage threshold in this process is 80%. According to the above linear regression algorithm model y=0.7619*x+71.65, it can be seen that the slope is 0.7619. This value is relatively large, indicating that the CPU usage rate is on the rise. According to the forecast value given in Table 3, it can be seen that most of the forecast values have exceeded 80% and tend to 85%. In this case, you do not need to adjust the CPU usage threshold, because according to predictions, the CPU usage rises rapidly, and an alarm needs to be issued in time to remind the network administrator to handle it.

Example 2: Medium and long-term pre-monitoring and warning

Taking the monitoring of the CPU usage of network devices and the time window length as one day as an example, the long-term pre-monitoring and alarm process may include:

Collect the CPU usage rate from the traditional monitoring system according to the set data collection cycle, for example, collect once every 5 minutes, 10 minutes or 15 minutes, and count the maximum value of the CPU usage rate collected in this time period every hour, and set The peak value of the CPU usage counted every hour within a day is averaged to obtain the peak value of the CPU usage of the day.

Then take the peak average value of the CPU usage rate in 15 time windows forward from the current time window, that is, take the peak value value of the CPU usage rate in the previous 15 days. The peak value value of the CPU usage rate in the 15 time windows can be shown in Figure 2 Show. Wherein, the peak average value of the daily CPU usage from May 16 to May 30, 2009 is shown.

After that, subtract the CPU usage of the previous day from the CPU usage of the previous 15 days from the peak average value of the CPU usage to obtain an array containing 14 incremental values; then use the linear regression algorithm model, and According to the array, calculate the slope of the linear regression algorithm model of the 14 incremental values; then use the linear regression algorithm model, and calculate the increase of the CPU usage rate in the next 15 days compared with the previous day according to the slope. magnitude.

Afterwards, add the incremental value of the CPU usage rate for the next 15 days to the predicted value of the CPU usage rate of the previous day to obtain the predicted value of the CPU usage rate for this day. Wherein, the CPU usage forecast value of the first day in the next 15 days is the sum of the incremental value of the day and the peak average value of the CPU usage rate of the previous day.

According to the above prediction process, as shown in FIG. 2 , the predicted value of CPU usage from May 31 to June 14, 2009 is obtained.

Predicting the KPI performance index data through the slope of the slope can amplify the change trend of the KPI performance index data in order to attract the attention of the monitoring personnel.

As shown in FIG. 3 , the actual CPU usage from May 31 to June 14 is shown. Compared with the previous forecast, although the CPU usage did not exceed 75%, the CPU usage has increased significantly compared with the previous 15 days. In this case, although the threshold for monitoring alarms has not been reached, the CPU usage of this machine has exceeded The rapid increase requires monitoring personnel to focus on and investigate the cause of the rapid increase in CPU usage as soon as possible, so as to avoid affecting the operation of the system in the near future.

Example 3: Long-term pre-monitoring alarm

Taking the monitoring of the CPU utilization rate of network equipment and the time window length as one month as an example, the long-term pre-monitoring and alarming process may include:

Collect the CPU usage rate from the traditional monitoring system according to the set data collection cycle, for example, collect once every 5 minutes, 10 minutes or 15 minutes, and count the maximum value of the CPU usage rate collected in this time period every hour, and set Average the peak value of the CPU usage counted every hour within a day to obtain the peak average value of the CPU usage rate for that day; average the peak value value of the CPU usage rate every day within a month to obtain the CPU usage rate for a month peak mean.

Then take the peak average value of the CPU usage for three months forward from the current month, and the peak average value of the CPU usage for the three months can be shown in FIG. 4 . Wherein, the peak average value of the monthly CPU usage during May to July 2009 is shown.

Then, according to the peak average value of the monthly CPU usage from May to July 2009, use the linear regression algorithm to predict the CPU usage in the next three months.

It can be seen from Figure 4 that the CPU usage rate in the next three months will not change much compared with the previous months, and is basically stable. It can be considered that the server will meet the capacity requirements in the next three months.

Take the CPU usage rate of an application server in a company as an example, as shown in Figure 5:

This instance predicts the CPU usage in the next three months by calculating the average CPU usage in May, June, and July of 2009. It can be seen from Figure 5 that the CPU usage is showing a clear downward trend, and will drop below 50% in the next few months. It can be considered to re-allocate resources so that the machine can be used more reasonably and effectively.

Based on the same technical concept, an embodiment of the present invention also provides a performance monitoring device that can implement the above process.

Referring to FIG. 6 , it is a schematic structural diagram of a network management system provided by an embodiment of the present invention. This device can include:

A collection module 601, configured to periodically collect KPI performance index data of network equipment;

A summary module 602, configured to summarize the collected KPI performance index data according to the set time window, to obtain the summary value of the KPI performance index data of each time window;

The prediction module 603 is used to predict the KPI performance index data of N time windows in the future including the current time window according to the summary value of the KPI performance index data of N time windows before the current time window by using the linear regression algorithm model;

The alarm module 604 is configured to determine whether the KPI performance index data in the N future time windows exceeds the set alarm threshold, and issue an alarm if the determination is yes.

Further, an output module 605 may also be included, configured to output the predicted KPI performance index data of the N time windows in the future.

Specifically, the KPI performance index data summary value of each time window summarized by the summary module 602 is the peak value of the KPI performance index data collected in each time window, or the KPI collected in each collection period in each time window The average value of the peak value of the performance index data, the length of the collection period is shorter than the length of the time window.

Specifically, the prediction module 603 can use the linear regression algorithm model, and calculate the linear regression algorithm of the KPI performance index data summary value of the N time windows before the current time window according to the summary value of the KPI performance index data of the N time windows Model slope: using the linear regression algorithm model, and calculating the predicted values of the KPI performance index data of N time windows in the future including the current time window according to the slope. Its specific implementation can be described with reference to the aforementioned process.

Specifically, in the N time windows before the current time window, the prediction module 603 may subtract the KPI performance index data summary value of the previous time window from the KPI performance index data summary value of each time window to obtain the N -1 array of incremental values; using the linear regression algorithm model, and according to the array, calculate the slope of the linear regression algorithm model of the N-1 incremental values; using the linear regression algorithm model, and according to the slope respectively Calculate the incremental value of the KPI performance index data of the next N time windows including the current time window compared with the previous time window; according to the KPI performance index data of each time window in the future N time windows Incremental value, and the KPI performance index data prediction value of the previous time window, to obtain the KPI performance index data prediction value of each time window in the N time windows in the future; wherein, in the N time windows in the future The predicted value of the KPI performance indicator data in the first time window is the sum of the corresponding incremental value and the summary value of the KPI performance indicator data in the previous time window. Its specific implementation can be described with reference to the aforementioned process.

Specifically, the alarm module 604 is specifically configured to: if the KPI performance index data of the N time windows in the future exceeds the KPI performance index threshold, send an alarm; if the KPI performance index data of the N time windows before the current time window If the slope of the linear regression algorithm model of the data summary value is greater than the set threshold, an alarm is issued; if the slope of the linear regression algorithm model of the KPI performance index data summary value of the N time windows before the current time window is greater than the set threshold, and the If the KPI performance index data of the N future time windows exceeds the threshold of the KPI performance index, an alarm is issued.

Specifically, the threshold of the KPI performance index is based on the slope of the linear regression algorithm model of the KPI performance index data of the N time windows in the future and the summary value of the KPI performance index data of the N time windows before the current time window. It is adjusted based on the set KPI performance indicator threshold.

Specifically, the length of the time window is in minutes; or, the length of the time window is in days; or, the length of the time window is in months.

It can be seen from the above description that, compared with the prior art, the embodiments of the present invention have advantages in the following aspects:

(1) The embodiment of the present invention turns real-time monitoring into active early warning monitoring, so as to avoid the occurrence of existing technical problems, carry out in-depth mining and sorting of traditional monitoring data, and predict the future period by calculating the slope of historical data. For the use of time, active early warning is realized by configuring the alarm mechanism. According to the change trend of the system KPI indicators, when the system has a sudden change in performance, an early warning message will be issued in advance to inform the system maintenance personnel to take effective measures to prevent system performance failures.

(2) The embodiment of the present invention supports the configuration of the analysis time window, which defaults to the standard 15 minutes, and the 15-minute time window is the best time window configuration based on long-term operation and maintenance experience, and the interval of 15 minutes is predicted The effect is also the most ideal, and it is also possible to configure customized time windows according to different systems of different users, which is easy to change and has a wide promotion prospect.

(3) The embodiment of the present invention provides a near-real-time prediction process. According to the monitoring results of the actual application system, usually 60 to 120 minutes before the system KPI data exceeds the threshold, the pre-location of the fault can be realized and the management can be given in advance. Personnel send alarm information to prevent problems before they happen.

(4) The embodiment of the present invention provides a mid-to-long-term (15-day) forecasting process. The mid-to-long-term forecast can predict KPI trend data in the next 15 days or so, providing a basis for mid-to-long-term system monitoring.

(5) The embodiment of the present invention provides capacity prediction (6-12 months), according to the long-term (6-12 months) prediction, it can provide valuable basis for the host capacity planning of key business systems, and lay a good foundation for system expansion .

(6) The embodiment of the present invention realizes the configurable KPI of monitoring, hundreds of KPIs are available for configuration from CPU, memory to storage, and the network, and the scope of monitoring is relatively wide. Pre-monitoring of hundreds of KPIs such as memory, network, etc. At the same time, the monitoring KPI indicators can also be completely customized, providing a good human-computer interaction interface, and the configuration method is simple and easy, which is very convenient for system transplantation and promotion. A good foundation has been laid.

(7) The embodiment of the present invention is simple and convenient to deploy, occupies very low system resources, and has low cost, and is easy to be popularized and used in a large area.

(8) The rules of the early warning in the embodiment of the present invention can be defined, and can be defined according to the slope and the threshold of the predicted value. The two indicators of the KPI slope and the threshold can be customized according to the user's business characteristics, so as to fully meet the requirements of customization. Pre-failure positioning. When the operating resources of the system are relatively tight, and the KPI is above the operating threshold for a long time, but there is no significant change, the system can be considered to be running normally to a certain extent. By adjusting the threshold of the KPI performance index data used for alarms, the generation of alarms can be avoided. A large number of alarms reduce the burden on system maintenance personnel to deal with alarms.

Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be realized by hardware, or by software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present invention can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.), including several The instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute the methods described in various embodiments of the present invention.

Those skilled in the art can understand that the drawing is only a schematic diagram of a preferred embodiment, and the modules or processes in the drawing are not necessarily necessary for implementing the present invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description in the embodiment, or can be located in one or more devices different from the embodiment according to corresponding changes. The modules in the above embodiments can be combined into one module, and can also be further split into multiple sub-modules.

The above serial numbers of the present invention are only for description, and do not represent the advantages and disadvantages of the embodiments.

The above disclosures are only a few specific embodiments of the present invention, however, the present invention is not limited thereto, and any changes conceivable by those skilled in the art shall fall within the protection scope of the present invention.

Claims (16)

1. a performance of network equipments method for supervising, is characterized in that, comprising:

The KPI performance index data of network management system cycle collection network equipment;

Described network management system gathers the KPI performance index data that collect according to the time window of setting, the KPI performance index data that obtain each time window gather value;

Described network management system is utilized the linear regression algorithm model, and gathers value according to the KPI performance index data of N time window before the current time window, and prediction comprises the KPI performance index data of following N time window of current time window;

Described network management system judges whether the KPI performance index data of described following N time window surpass the setting alarm threshold, and sends alarm when being judged as YES.

2. the method for claim 1, is characterized in that, the KPI performance index data of described each time window gather value, is the KPI performance index data peaks that collects in each time window; Perhaps

The KPI performance index data of described each time window gather value, the mean value of the KPI performance index data peaks that collects for each collection period in each time window, and the length of described collection period is less than the length of described time window.

3. the method for claim 1, it is characterized in that, the described linear regression algorithm model that utilizes, and gather value according to the KPI performance index data of N time window before the current time window, prediction comprises the KPI performance index data of following N time window of current time window, comprising:

Utilize the linear regression algorithm model, and gather value according to the KPI performance index data of N time window before the current time window, the KPI performance index data that calculate this N time window gather the linear regression algorithm model slope of value;

Utilize described linear regression algorithm model, and calculate respectively the predicted value of the KPI performance index data of following N the time window that comprises the current time window according to this slope.

4. the method for claim 1, it is characterized in that, the described linear regression algorithm model that utilizes, and gather value according to the KPI performance index data of N time window before the current time window, prediction comprises the KPI performance index data of following N time window of current time window, comprising:

In N before a current time window time window, gather with the KPI performance index data of each time window the KPI performance index data that value deducts its previous time window and gather value, obtain including the array of N-1 increment size;

Utilize the linear regression algorithm model, and according to this array, calculate the linear regression algorithm model slope of this N-1 increment size;

Utilize described linear regression algorithm model, and calculate respectively according to this slope the increment size that the KPI performance index data of following N the time window that comprises the current time window are compared with previous time window;

Increment size according to the KPI performance index data of each time window in a described following N time window, and the KPI performance index data prediction value of previous time window, obtain the KPI performance index data prediction value of each time window in described following N time window; Wherein, the KPI performance index data prediction value of first time window in N time window in this future gathers the value sum for the KPI performance index data of corresponding increment size time window previous with it.

5. the method for claim 1, is characterized in that, when one of following situation, the KPI performance index data of the following N of a described network management system judgement time window surpass sets alarm threshold:

The KPI performance index data of described following N time window surpass this KPI Performance Counter Threshold;

The KPI performance index data of N time window before described current time window gather the linear regression algorithm model slope of value greater than setting threshold;

The linear regression algorithm model slope that the KPI performance index data of N time window before described current time window gather value is greater than setting threshold, and the KPI performance index data of described following N time window surpass this KPI Performance Counter Threshold.

6. method as claimed in claim 5, it is characterized in that, described KPI Performance Counter Threshold is the KPI performance index data according to a described following N time window, and the KPI performance index data of N time window before the current time window gather the linear regression algorithm model slope of value, and adjustment obtains on the KPI Performance Counter Threshold basis that sets in advance.

7. method as described in one of claim 1-6, is characterized in that, the length of described time window is take minute as unit; Perhaps

The length of described time window take in the sky as unit; Perhaps

The length of described time window is take the moon as unit.

8. method as described in one of claim 1-6, is characterized in that, described KPI performance index data comprise one of following or combination in any: CPU usage, memory usage and network throughput.

9. a network management system, is characterized in that, comprising:

Acquisition module is for the KPI performance index data of cycle collection network equipment;

Summarizing module, the KPI performance index data that are used for collecting gather according to the time window of setting, and the KPI performance index data that obtain each time window gather value;

Prediction module is used for utilizing the linear regression algorithm model, and gathers value according to the KPI performance index data of N time window before the current time window, and prediction comprises the KPI performance index data of following N time window of current time window;

Alarm module is used for judging whether the KPI performance index data of a pre-described following N time window surpass the setting alarm threshold, and sends alarm when being judged as YES.

10. network management system as claimed in claim 9, it is characterized in that, the KPI performance index data that described summarizing module gathers described each time window that obtains gather value, be the KPI performance index data peaks that collects in each time window, the mean value of the KPI performance index data peaks that perhaps collects for each collection period in each time window, the length of described collection period is less than the length of described time window.

11. network management system as claimed in claim 9, it is characterized in that, described prediction module specifically is used for, utilize the linear regression algorithm model, and gathering value according to the KPI performance index data of N time window before the current time window, the KPI performance index data that calculate this N time window gather the linear regression algorithm model slope of value; Utilize described linear regression algorithm model, and calculate respectively the predicted value of the KPI performance index data of following N the time window that comprises the current time window according to this slope.

12. network management system as claimed in claim 9, it is characterized in that, described prediction module specifically is used for, in N before a current time window time window, gather with the KPI performance index data of each time window the KPI performance index data that value deducts its previous time window and gather value, obtain including the array of N-1 increment size; Utilize the linear regression algorithm model, and according to this array, calculate the linear regression algorithm model slope of this N-1 increment size; Utilize described linear regression algorithm model, and calculate respectively according to this slope the increment size that the KPI performance index data of following N the time window that comprises the current time window are compared with previous time window; Increment size according to the KPI performance index data of each time window in a described following N time window, and the KPI performance index data prediction value of previous time window, obtain the KPI performance index data prediction value of each time window in described following N time window; Wherein, the KPI performance index data prediction value of first time window in N time window in this future gathers the value sum for the KPI performance index data of corresponding increment size time window previous with it.

13. network management system as claimed in claim 12 is characterized in that, described alarm module specifically is used for, if the KPI performance index data of described following N time window surpass this KPI Performance Counter Threshold, sends alarm; If the linear regression algorithm model slope that the KPI performance index data of N time window before described current time window gather value sends alarm greater than setting threshold; If the KPI performance index data of N time window before described current time window gather the linear regression algorithm model slope of value greater than setting threshold, and the KPI performance index data of described following N time window surpass this KPI Performance Counter Threshold, send alarm.

14. network management system as claimed in claim 13, it is characterized in that, described KPI Performance Counter Threshold is the KPI performance index data according to a described following N time window, and the KPI performance index data of N time window before the current time window gather the linear regression algorithm model slope of value, and adjustment obtains on the KPI Performance Counter Threshold basis that sets in advance.

15. network management system as described in one of claim 9-14 is characterized in that, the length of described time window is take minute as unit; Perhaps

The length of described time window take in the sky as unit; Perhaps

The length of described time window is take the moon as unit.

16. network management system as described in one of claim 9-14 is characterized in that, described KPI performance index data comprise one of following or combination in any: CPU usage, memory usage and network throughput.

Images