JP2010068367A - Device for monitoring network performance - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for monitoring network performance capable of predicting the future even when the value of a parameter high in correlation used for calculating a prediction value of prediction object data is not fixed at a prediction point, and further accurately predicting the variation of network performance. <P>SOLUTION: In this device for monitoring network performance, prediction object data of network performance and data high in correlation to the prediction object data are collected from respective apparatuses connected to a network; the network performance is calculated at the same time by using a plurality of prediction models based on the data high in correlation; and a prediction model highest in prediction accuracy is selected to predict the network performance. In the device, when the value of the data high in correlation is not fixed at the prediction point, the network performance is predicted by using the prediction value at the prediction point calculated by the prediction model based on stored data of the data high in correlation collected in the past. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a network performance monitoring apparatus that grasps the performance of a network system to which each device is connected, and more particularly to predicting fluctuations in network performance with higher accuracy.

  In a network system in which multiple servers, relay devices such as routers and switches (hereinafter referred to as network devices) and various terminals are connected via a network, the performance of the system is increased due to excessive requests to servers, attacks, and failures. May deteriorate. Therefore, conventionally, in order to operate such a network system ideally, a technique for measuring the performance of the network system and optimally controlling communication resources has been proposed.

  In addition, a network performance monitoring apparatus has also been proposed in which “(average) response time” of servers, network devices, and the like is used as an index indicating network performance, and future prediction of this response time is performed.

  As a prior art document related to such a network performance monitoring apparatus, there is Patent Document 1 below. This Patent Document 1 describes a technique related to a network performance monitoring apparatus that monitors network performance by dynamically selecting an optimal prediction model.

JP 2002-099448 A

  FIG. 6 is a block diagram showing an example of a conventional network performance monitoring apparatus, which includes a network performance monitoring apparatus 1, a network 2, and a network device 3.

  The network performance monitoring apparatus 1 is connected to a network device 3 via a network 2 that is a wired network such as Ethernet (registered trademark).

  The network performance monitoring apparatus 1 includes a communication unit 11 that performs data communication with a network device 3 via a network 2, an arithmetic control unit 12 that controls the operation of each unit, a storage unit 13, and the like.

  The arithmetic control unit 12 performs performance basic data (for example, a time required for a response to a processing request of the network device 3 (hereinafter referred to as a response time) as a basis of a performance evaluation value indicating the performance of the network at a predetermined time interval or a predetermined timing. Information collecting means 12A that collects and calculates the time), and the response time obtained by the information collecting means 12A and the date and time when the measurement was performed are associated and stored in the accumulated information database 13 of the storage unit 13. An information storage unit 12B, a future prediction unit 12C for predicting a future response time based on a plurality of response times stored and stored in the storage information database 13A and a plurality of prediction models stored in the prediction model database 13B, etc. Have.

  The storage unit 13 includes an accumulated information database 13A (hereinafter referred to as accumulated information DB) in which the response time calculated by the information collecting unit 12A and the date of measurement are stored in association with each other, linear approximation, logarithmic approximation, A prediction model database 13B (hereinafter referred to as a prediction model DB) in which a plurality of prediction models based on methods such as polynomial approximation, power approximation, and exponential approximation are stored.

  FIG. 7 is a flowchart for explaining the operation of the network performance monitoring apparatus. In step SP101, the information collection unit 12A of the calculation control unit 12 of the network performance monitoring apparatus 1 collects the response time of the network device 3.

  Specifically, data for measuring the response time (hereinafter referred to as measurement request data) is transmitted to the network device 3. For example, the information collecting unit 12A of the arithmetic control unit 12 of the network performance monitoring apparatus 1 uses a packet using ICMP (Internet Control Message Protocol), that is, a packet using a Ping (Packet Internet Groper) program (Type 8: Echo Request). Is transmitted to the network device 3 and the transmission time is stored in the storage unit 13.

  After receiving the measurement request data from the network performance monitoring device 1, the network device 3 transmits the measurement response data to the network performance monitoring device 1 in order to respond to the request. Specifically, after receiving an ICMP packet (Type 8: Echo Request) from the network performance monitoring apparatus 1, the ICMP packet (Type 0: Echo Reply) is transmitted to the network performance monitoring apparatus 1.

  When receiving the ICMP packet (Type 0: Echo Reply) from the network device 3, the information collecting unit 12A calculates the response time of the network device 3 based on the reception time and the transmission time.

  In step SP102, the information storage unit 12B of the network performance monitoring apparatus 1 stores and stores the response time by the information collection unit 12A in the storage information database 13 of the storage unit 13 in association with the date of measurement.

  In step SP103, the future prediction means 12C of the network performance monitoring apparatus uses the plurality of prediction models stored in the prediction model DB 13B based on the plurality of response times stored and stored in the storage information DB 13A. Predict time.

  For example, the future prediction unit 12C calculates an average response time for one month based on the response time stored in the accumulated information DB 13A, and calculates (predicts) an average response time for the next month. FIG. 8 is an explanatory diagram showing the past monthly average response time. FIG. 9 is an explanatory diagram showing the predicted response time by each prediction model, where (A) is a prediction result by a model using linear approximation, (B) is a prediction result by a model using logarithmic approximation, and (C) is a polynomial. The prediction result by the model using approximation, (D) is the prediction result by the model using power approximation, and (E) is the prediction result by the model using exponential approximation.

  The future prediction means 12C calculates and stores the average monthly response time for the past seven months (January: 14 milliseconds, February: 18 milliseconds, March: 24 milliseconds, etc.) as shown in FIG. 9 and predicts the response time of the next month (August) using each prediction model as shown in FIG. 9 (for example, (A) linear approximation: 34 milliseconds, (B) logarithmic approximation: 30. 6 milliseconds, (C) polynomial approximation: 31.7 milliseconds, (D) power approximation: 31.9 milliseconds, (E) exponential approximation: 36.8 milliseconds) are stored in the storage unit 13.

  In step SP104, the future prediction means 12C compares each prediction accuracy of the prediction result calculated using each model.

  Specifically, the future prediction unit 12C calculates an R-2 power value in each prediction model, and compares the prediction accuracy using the R-2 power value. This R-2 power value indicates that the closer the value is to 1, the higher the reliability.

  In step SP105, the future prediction unit 12C selects the prediction model with the highest prediction accuracy as a result of the comparison of the prediction accuracy, and thereafter sets and stores it as a prediction model for calculating the response time prediction data. . For example, in the example of FIG. 8, the prediction model using the power approximation with the highest R-2 power value is selected.

  In step SP106, the future prediction unit 12C calculates response time prediction data using the prediction model selected in step SP105, and stores it in the storage unit 13. It is also displayed on a display unit (not shown).

  Specifically, the future prediction unit 12C calculates a predicted value (31.9 milliseconds) of the average monthly response time using a prediction model that uses power approximation, and stores it in the storage unit 13.

  Thus, the conventional network performance monitoring apparatus can predict the performance of the network system.

  However, the conventional network performance monitoring device uses data that is a prediction target for grasping the network performance (hereinafter referred to as prediction target data (for example, response time)) and has a strong correlation with the prediction target data. If the value of traffic (such as traffic volume) is not fixed at the time of prediction, there is a problem that future prediction cannot be performed.

  10 and 11 are explanatory diagrams of problems of the conventional network performance monitoring apparatus. FIG. 10 illustrates a case where prediction can be performed without any problem, and FIG. 11 illustrates a case where there is a problem in data used for prediction and future prediction cannot be performed. Yes.

  In FIG. 10, the average monthly response time for the past seven months (January to July) is shown, and if the monthly degree (or date) increases (January <February <March ... <July). Since the average response time also increases, the monthly degree (or date) and the response time have a strong correlation. Therefore, the network performance monitoring apparatus 1 can perform future prediction using the prediction model based on the relationship between the monthly degree and the average response time without any problem.

  In FIG. 11, the average monthly response time and the monthly traffic volume (%) for the past seven months (January to July) are shown. The traffic volume (%) is the “packet volume” and “communication speed” output from the network device 3 via the interface (not shown) of the network device 3 by the network performance monitoring device 1 at a predetermined time interval or at a predetermined timing. Is obtained by dividing the amount of packets by the communication speed.

  In the case as shown in FIG. 11, since the correlation between the average response time and the monthly degree (or date) is weak, it is impossible to predict the future from the relationship between the monthly degree (or date) and the average response time. On the other hand, since the monthly average response time and the monthly traffic volume (%) have a strong correlation, the network performance monitoring apparatus 1 can predict the future from the relationship between the monthly traffic volume and the average response time.

  However, the traffic volume (%) data has a problem that the network performance monitoring device 1 cannot predict because the value of the traffic volume (%) is not fixed at the time of prediction as with the response time.

  The present invention solves the above-mentioned problems, and its purpose is to perform network performance even when the numerical value of data having a strong correlation with the prediction target data used in the prediction calculation is not fixed at the time of prediction. It is an object of the present invention to realize a network performance monitoring apparatus capable of predicting future fluctuations of the network and predicting fluctuations in network performance with higher accuracy.

In order to achieve such a problem, the invention according to claim 1 of the present invention is:
Collect data to be predicted for network performance from each device connected to the network and data that has a strong correlation with the data to be predicted, and simultaneously calculate using multiple prediction models based on the data with strong correlation. In a network performance monitoring apparatus that selects a high prediction model and predicts the performance of the network,
When the data with strong correlation is uncertain at the time of prediction, the performance of the network is predicted using the prediction value at the prediction time calculated by the prediction model based on the accumulated data of the data with strong correlation collected in the past. A network performance monitoring apparatus.

The invention according to claim 2 is the network performance monitoring apparatus according to claim 1,
A communication unit for performing data communication via the network;
A storage unit having a storage information database storing the prediction target data and the data having a strong correlation in association with each other and a prediction model database storing the plurality of prediction models;
Predicted data for evaluating the performance of the network via the network by controlling the communication unit and data highly correlated with the predicted target data are collected and stored in the storage unit, and the data having a strong correlation If the numerical value is uncertain at the prediction time, the performance of the network is calculated by using the numerical value at the prediction time calculated based on the accumulated data of the highly correlated data collected in the past and stored in the storage unit. A calculation control unit to predict;
It is characterized by providing.

The invention according to claim 3 is the network performance monitoring apparatus according to claim 2,
The arithmetic control unit is
Information collecting means for controlling the communication unit to collect the data to be predicted and the data having a strong correlation;
Information storage means for associating the prediction target data obtained by the information collection means with the data having a strong correlation and storing them in the stored information database;
Parameter future prediction means for calculating a predicted value at the prediction time of the data with strong correlation based on the accumulated data of the data with strong correlation stored in the storage unit;
Based on the prediction target data stored in the accumulated information database and the highly correlated data and the predicted value thereof, the prediction model stored in the prediction model database is simultaneously calculated using all the prediction models, and the prediction accuracy is the highest. Future prediction means for selecting a prediction model and predicting the performance of the network;
It is characterized by providing.

The invention according to claim 4 is the network performance monitoring apparatus according to claim 3,
The parameter future prediction means includes
Based on the accumulated data for the past several years at the same time as the prediction target time among the strongly correlated data stored in the storage unit, at least one of the plurality of prediction models or the prediction model having the highest prediction system Is used to calculate a predicted value of the data having a strong correlation at the prediction target time.

The invention according to claim 5 is the network performance monitoring apparatus according to claim 3 or 4, wherein
The information collecting means includes
The response time of the network device connected to the network is collected as the prediction target data, and the traffic amount obtained by dividing the packet amount output by the network device by the communication speed is collected as the highly correlated data. It is characterized by that.

  According to the network performance monitoring apparatus according to the present invention, by predicting the future based on data accumulated from the past having a strong correlation with the prediction target data, the numerical value of the data having a strong correlation at the time of prediction is not confirmed. Even if it exists, the fluctuation | variation of a network performance can be estimated and the fluctuation | variation of a network performance can be estimated more accurately.

  FIG. 1 is a block diagram showing an embodiment of a network performance monitoring apparatus according to the present invention. Parts common to those in FIG. The difference between FIG. 1 and FIG. 6 is that in FIG. 1, the network performance monitoring device 1 collects and calculates the traffic volume of the network device 3 and the data value necessary for calculating the predicted value of the prediction target data. For example, holding parameter future prediction means for prediction / calculation.

  In FIG. 1, the network performance monitoring device is composed of a network performance monitoring device 1, a network 2, a network device 3 that is a relay device such as a router or a switch, and the like.

  The network performance monitoring apparatus 1 is connected to a network device 3 via a network 2 that is a wired network such as Ethernet (registered trademark).

  The network performance monitoring device 1 operates as a communication unit 11 that performs data communication with the network device 3 via the network 2, an arithmetic control unit (for example, CPU) 12 that controls the operation of each unit, an OS, and the network performance monitoring device 1. For example, a storage unit 13 such as a RAM or a ROM storing a storage information database 13A, a prediction model database 13B, and the like.

  The arithmetic control unit 12 includes information collecting means 12A that collects and calculates basic performance data such as response time and traffic volume, which is a basis for evaluating network performance. The information collecting unit 12A calculates the response time of the network device 3 by transmitting data to measure the response time of the network device 3 at a predetermined time interval or at a predetermined timing, for example. Traffic information data including information on the output packet amount and the communication speed is collected to calculate the traffic amount (%).

  Further, the arithmetic control unit 12 associates the performance basic data (for example, response time) obtained by the information collecting unit 12A with the date and time when the measurement is performed, and stores and stores the information in the accumulated information database 13 of the storage unit 13. Based on a plurality of prediction models stored in the prediction model database 13B by selecting prediction target data serving as an index for performance evaluation from the storage means 12B and data stored and stored in the storage information database 13A. Future prediction means 12C for predicting future numerical values, and parameter future prediction means 12D for predicting / calculating data values having a strong correlation with the prediction target data necessary for the network performance monitoring apparatus 1 to calculate the prediction values of the prediction target data. Etc.

  That is, the information collecting unit 12A of the arithmetic control unit 12 collects the prediction target data and data having a strong correlation with the prediction target data.

  The storage unit 13 stores, for example, prediction target data such as response time collected and calculated by the information collecting unit 12A, measurement date and other performance basic data such as traffic volume and number of error packets, and the like. A storage information database (hereinafter referred to as storage information DB) 13A, and a prediction model database (a plurality of prediction models based on each method such as linear approximation, logarithmic approximation, polynomial approximation, power approximation, exponential approximation, etc.) (Hereinafter referred to as prediction model DB) 13B.

  FIG. 2 is a flowchart for explaining the operation of the network performance monitoring apparatus of FIG. In step SP111, the information collecting unit 12A of the calculation control unit 12 of the network performance monitoring apparatus 1 outputs the prediction target data and the response time of the network device 3 including the data strongly correlated with the prediction target data and the network device 3 for example. Collect basic performance data such as packet volume and communication speed data.

  Specifically, the network performance monitoring device 1 transmits data for measuring response time (hereinafter referred to as RT measurement request data) to the network device 3. For example, the information collecting unit 12A of the arithmetic control unit 12 of the network performance monitoring apparatus 1 transmits a packet using ICMP, that is, a packet using a Ping program (Type 8: Echo Request) to the network device 3 and its transmission time. Is stored in the storage unit 13.

  After receiving the RT measurement request data from the network performance monitoring device 1, the network device 3 transmits the RT measurement response data to the network performance monitoring device 1 in order to respond to the request. Specifically, after receiving an ICMP packet (Type 8: Echo Request) from the network performance monitoring apparatus 1, the ICMP packet (Type 0: Echo Reply) is transmitted to the network performance monitoring apparatus 1.

  When the information collection unit 12A receives RT measurement response data (for example, ICMP packet (Type 0: Echo Reply)) from the network device 3, the information collection unit 12A determines the response time of the network device 3 based on the reception time and the transmission time. calculate.

  Further, the network performance monitoring apparatus 1 transmits data for collecting and calculating the traffic amount (hereinafter referred to as TF amount calculation request data) to the network device 3. For example, the information collecting unit 12A of the arithmetic control unit 12 of the terminal device 1 transmits an SNMP request to the network device 3 using an SNMP (Simple Network Management Protocol) protocol.

  After receiving the TF amount calculation data from the network performance monitoring device 1, the network device 3 transmits TF amount calculation response data to the network performance monitoring device 1 in order to respond to the request. Specifically, after receiving an SNMP request from the network performance monitoring device 1, traffic information data including the packet amount output by each interface of the network device 3 and its communication speed is transmitted to the network performance monitoring device 1.

  When the information collection unit 12A receives the response data for TF amount calculation from the network device 3, the information collection unit 12A calculates the traffic amount (%) of the network device 3 by dividing the packet amount by the communication speed.

  In step SP112, the information accumulating unit 12B of the network performance monitoring apparatus 1 associates the date of measurement of the performance basic data including the prediction target data and the strongly correlated data by the information collecting unit 12A with the storage unit 13 and the like. Is stored and stored in the storage information DB 13A.

  Note that the information collection unit 12A and the information storage unit 12B of the network performance monitoring apparatus 1 repeatedly perform the processing of steps SP111 to SP112 at a predetermined time interval or at a predetermined timing. For example, the accumulated information DB 13A of the storage unit 13 stores accumulated information for the past several years (for example, 1 to 5 years).

  In step SP113, the parameter predicting unit 12D acquires storage information from the storage information DB 13A in order to perform future prediction of network performance.

  The future predicting unit 12C, when the accumulated information DB 13A is accessed from the parameter predicting unit 12D, is based on the response time and traffic amount stored in the accumulated information DB 13A, and the average response time for one month and the traffic amount for one month. And the average response time and traffic volume are associated with each other and stored in the accumulated information DB 13A.

  FIG. 3 is an explanatory diagram illustrating an example of accumulated information used when performing future prediction. The parameter prediction means 12D is, for example, monthly (January to August), traffic volume (January: 21%, February: 40, March: 3%, ..., August: 41%), monthly average response Accumulated information including time (January: 40 milliseconds, February: 63 milliseconds, March: 11 milliseconds,..., August: 31 milliseconds) is acquired.

  In step SP114, the parameter predicting unit 12D selects data having a strong correlation with the prediction target data (for example, average response time) based on the acquired accumulated information. Further, when the value of data having a high correlation is not fixed at the prediction time point, the parameter prediction unit 12D calculates the value at the prediction time point using the prediction model stored in the prediction model DB 13B of the storage unit 13. Calculate and predict.

  For example, in the case shown in FIG. 3, since the correlation between the average response time and the monthly level (or date) is weak, it is determined that the future prediction is impossible from the monthly level (or date), but the average monthly response time and the monthly traffic volume (% )), It is understood that future prediction is possible from the relationship between the monthly traffic volume and the average response time, and the monthly traffic volume is selected as data having a strong correlation with the average response time.

  The parameter predicting unit 12D determines the value at the prediction time using the prediction model stored in the prediction model DB 13B of the storage unit 13 because the value of the monthly traffic volume is not fixed at the prediction time (August). Predict and calculate.

  FIG. 4 is an explanatory diagram of future prediction of data (traffic volume) having a high correlation with the average response time at the prediction time (August). As shown in FIG. 4, for example, the network performance monitoring apparatus 1 has a traffic volume in August of the past five years stored in the accumulated information DB 13A (42% for 5 years ago: 38% for 4 years ago: 38% for 3 years ago: 45 years ago). %, 2 years ago: 41%, 1 year ago: 39%), a regression equation such as linear approximation, logarithmic approximation, polynomial approximation, power approximation, exponential approximation, or moving average stored in the prediction model DB 13B. A predicted value (41%) of traffic volume in August is calculated and predicted using a prediction model using analysis or the like.

  In other words, the parameter predicting unit 12D moves, for example, based on the accumulated data for the past five years at the same time as the prediction target time (August) among the highly correlated data (traffic volume) stored in the accumulated information DB 13A. Based on a prediction model using average analysis or the like, a predicted value of data having a strong correlation at the prediction target time (August) is calculated and predicted.

  The network performance monitoring apparatus 1 stores basic performance data including traffic volume, response time, and the like for the past several years (for example, five years) in the storage information DB 13A of the storage unit 13.

  Further, the parameter predicting means 12D calculates the predicted value of the data at the same time using all of a plurality of prediction models using regression equations based on the accumulated data of the past five years of the data having strong correlation, and the prediction accuracy is high. It is also possible to select the highest prediction model and calculate / predict the prediction value of data having a strong correlation at the prediction target time (August).

  In step SP115, the future prediction means 12C of the network performance monitoring device is stored in the prediction model DB 13B based on the calculated parameters and the prediction target data stored and stored in the storage information DB 13A (for example, based on the monthly average response time). The prediction value of the future prediction target data is predicted using all the prediction models.

  Specifically, the future prediction unit 12C uses the prediction model to calculate the average of the next month based on the predicted traffic volume (%) calculated by the parameter future prediction unit 12D and the average response time stored in the accumulated information DB 13A. Calculate (predict) the response time.

  Further, the future prediction means 12C predicts, for example, a prediction model using linear approximation, a prediction model using logarithmic approximation, a prediction model using polynomial approximation, a prediction model using power approximation, a prediction model using exponential approximation, and the like. Use the model.

  For example, the future prediction unit 12C predicts the response time of the next month (August) using each prediction model (for example, linear approximation: 64 milliseconds, logarithmic approximation: 60.6 milliseconds, polynomial approximation: 61.7 milliseconds). Second, power approximation: 62.1 milliseconds, exponential approximation: 66.8 milliseconds), and stored in the storage unit 13.

  In step SP116, the future prediction unit 12C compares the prediction accuracy of the prediction results calculated using all the models in the prediction model DB 13B.

  Specifically, the future prediction unit 12C calculates an R-2 power value in each prediction model, and compares the prediction accuracy using the R-2 power value. This R-2 power value indicates that the closer the value is to 1, the higher the reliability.

  In step SP117, the future prediction unit 12C selects the prediction model having the highest prediction accuracy as a result of comparing the prediction accuracy. The future prediction unit 12C then sets and stores it as a prediction model for calculating response time prediction data.

  The future prediction unit 12C selects, for example, a prediction model using “power approximation” that is determined to have the highest R−2 value.

  In step SP118, the future prediction unit 12C calculates a prediction value of the prediction target data (for example, average response time) using the prediction model selected in step SP105, and stores it in the storage unit 13.

  FIG. 5 is an explanatory diagram of the predicted response time according to the selected prediction model. The future prediction unit 12C calculates a predicted value (62.1 milliseconds) of the average monthly response time using, for example, a prediction model (5.1732 * X to 0.6691) using “power approximation” and stores it in the storage unit 13. .

  The arithmetic control unit 12 controls a display unit (not shown) to display various performance basic data such as monthly degree, traffic volume (%), monthly average response time, etc. stored in the accumulated information DB 13A in a table format. Prediction target data (for example, monthly average response time) and data highly correlated with the prediction target data (for example, traffic volume), prediction result by each prediction model, its R-square value, prediction model determined to be the highest and its average The response time may be displayed in a table format.

  As a result, the network performance monitoring apparatus according to the present invention predicts future data based on data accumulated from the past based on data that has a strong correlation with the prediction target data used in the prediction calculation. Even if the numerical value is not fixed, the fluctuation of the network performance can be predicted, and the fluctuation of the network performance can be predicted with higher accuracy.

  In the network performance monitoring apparatus of the above-described embodiment, the traffic amount is given as an example of the highly correlated data used for calculating the predicted value of the prediction target data. However, the present invention is not limited to this, and the error packet The number, the number of sessions, etc. may be handled as highly correlated data used for calculating the predicted value of the prediction target data. At this time, the information collecting unit 12A of the network performance monitoring apparatus 1 may collect traffic information data including the number of error packets and the number of sessions detected by the network device 3 in step SP111 in FIG.

  In the network performance monitoring apparatus of the above-described embodiment, it is described that the network performance monitoring apparatus 1 and the network device 3 are connected to the network 2. However, the present invention is not limited to this, and a plurality of network devices In addition, a plurality of various terminals such as computers may be connected.

  In the network performance monitoring apparatus of the above-described embodiment, it is described that the response time is measured by using an ICMP packet and the network performance is grasped. When the server is connected to the network 2, the time taken to acquire the resources of the website provided by the web server may be measured as the response time.

  In addition, the arithmetic control unit 12 of the network performance monitoring apparatus according to the present invention starts up an OS or the like stored in the storage unit 13 and reads and executes a program stored on the OS to thereby execute the network performance monitoring apparatus. 1 (for example, the arithmetic control unit 11 reads and executes the program to control the information collection unit 12A, the information storage unit 12B, the future prediction unit 12C, the parameter prediction unit 12D, etc.) and correlates with the prediction target data. When the data with strong data is uncertain at the time of prediction, the network performance is predicted using the numerical value at the time of prediction calculated by the prediction model based on the accumulated data of data that has been collected in the past. It may be what performs.

  At this time, the storage unit 44 expands the program or application executed by the arithmetic control unit 41 in the program storage area, and temporarily stores the input data or data such as a processing result generated when the program or application is executed in the work area. To remember.

  As described above, the network performance monitoring apparatus according to the present invention can predict fluctuations in network performance with higher accuracy by predicting the future based on accumulated data from the past with data that has a strong correlation with the prediction target data, It is possible to control communication resources optimally based on the prediction result and contribute to appropriate operation of the entire network system.

It is a block diagram which shows one Example of the network performance monitoring apparatus which concerns on this invention. It is a flowchart explaining operation | movement of a network performance monitoring apparatus. It is explanatory drawing which shows an example of the accumulation | storage information utilized when performing future prediction. It is explanatory drawing of the future prediction of the parameter (traffic amount) of a prediction time point (August degree). It is explanatory drawing of the prediction response time by the selected prediction model. It is a block diagram which shows an example of the conventional network performance monitoring apparatus. It is a flowchart explaining operation | movement of a network performance monitoring apparatus. It is explanatory drawing which shows the past monthly average response time. It is explanatory drawing which shows the prediction response time by each prediction model. It is explanatory drawing of the problem of the conventional network performance monitoring apparatus. It is explanatory drawing of the problem of the conventional network performance monitoring apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Network performance monitoring apparatus 11 Communication part 12 Operation control part 12A Information collection means 12B Information storage means 12C Future prediction means 12D Parameter future prediction means 13 Storage part 13A Accumulated information database 13B Prediction model database 2 Network 3 Network equipment

Claims (5)

Collect data to be predicted for network performance from each device connected to the network and data that has a strong correlation with the data to be predicted, and simultaneously calculate using multiple prediction models based on the data with strong correlation. In a network performance monitoring apparatus that selects a high prediction model and predicts the performance of the network,
When the data with strong correlation is uncertain at the time of prediction, the performance of the network is predicted using the prediction value at the prediction time calculated by the prediction model based on the accumulated data of the data with strong correlation collected in the past. A network performance monitoring apparatus characterized by: A communication unit for performing data communication via the network;
A storage unit having a storage information database storing the prediction target data and the data having a strong correlation in association with each other and a prediction model database storing the plurality of prediction models;
Predictive data for controlling the communication unit to evaluate the performance of the network via the network and data highly correlated with the predictive data are collected and stored in the storage unit, and the correlation is strong When the data is uncertain at the time of prediction, the network performance using the value at the time of prediction calculated based on the accumulated data of the data that has been collected in the past and stored in the storage unit An arithmetic control unit for predicting
The network performance monitoring apparatus according to claim 1, comprising: The arithmetic control unit is
Information collecting means for controlling the communication unit to collect the prediction target data and the data having a strong correlation;
Information storage means for associating the prediction target data obtained by the information collection means with the data having a strong correlation and storing them in the stored information database;
Parameter future prediction means for calculating a predicted value at the prediction time of the data with strong correlation based on the accumulated data of the data with strong correlation stored in the storage unit;
Based on the prediction target data stored in the accumulated information database and the highly correlated data and the predicted value thereof, the prediction model stored in the prediction model database is simultaneously calculated using all the prediction models, and the prediction accuracy is the highest. Future prediction means for selecting a prediction model and predicting the performance of the network;
The network performance monitoring apparatus according to claim 2, further comprising: The parameter future prediction means includes
Based on the accumulated data for the past several years at the same time as the prediction target time among the strongly correlated data stored in the storage unit, at least one of the plurality of prediction models or the prediction model having the highest prediction system The network performance monitoring apparatus according to claim 3, wherein a predicted value of the data having a strong correlation at the prediction target time is calculated by using. The information collecting means includes
The response time of the network device connected to the network is collected as the prediction target data, and the traffic amount obtained by dividing the packet amount output by the network device by the communication speed is collected as the highly correlated data. The network performance monitoring apparatus according to claim 3 or 4, characterized in that

Images