JP2002268922A - Performance monitoring device of www site - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a performance monitoring device of a WWW site to realize a preventative countermeasure for maintenance of performance. SOLUTION: The performance monitoring device 1 of the WWW site is constituted of an access fluctuation monitoring part 11 to fetch and store access quantity in the WWW site per unit time as time-series data of access fluctuation, an access fluctuation predicting part 12 to predict the access fluctuation in the future by statistical technique based on the stored pieces of time-series data, a site performance monitoring part 13 to acquire and store performance data by which use states of components in the WWW site from the components, a performance calculating part 14 to calculate performance values by every access based on a model in which the components of the site are abstracted from prediction data of the access fluctuation and performance monitoring data of the site, and a performance judging part 15 to judge whether or not prescribed performance is maintained based on the performance values of the calculated performance values by every access.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a WWW (World
The present invention relates to a performance monitoring device for a WWW site that predicts fluctuations in performance of a Wide Web site in advance and performs preventive measures based on the results, thereby maintaining a predetermined performance of the WWW site.

[0002]

2. Description of the Related Art At present, business via the WWW is about to rise at a rapid pace. In such WWW sites, it can be considered essential to provide high-performance and reliable services for securing and maintaining customers. Therefore, at present, the WWW
An operation management method called SLM (Service Level Management) that quantitatively sets the service quality of a site and adheres to the quality is attracting attention. In order to maintain such a predetermined quality of service, it is necessary to take preventive measures before a problem occurs, instead of taking an ex post measure for some kind of failure.

[0003]

As described above, performance maintenance is very important in a commercial WWW site. However, a recent WWW site is composed of various elements such as a network, a server, a database, and an application, and it is very difficult to maintain the performance perceived by the user as a whole. W
In order to maintain and manage the performance of the WW site, it is essential to measure the performance and take measures when the performance deteriorates. However, if the performance can be predicted further and preventive measures can be taken promptly, a certain level of performance quality will be achieved. Can be maintained. Conventional management products for such performance maintenance management have conventionally managed them by using the following methods (1) and (2). (1) Measures and displays the response time of access to a predetermined WWW site at regular intervals, and issues a warning when the response time exceeds a predetermined value. In addition, the measurement data is accumulated as time-series data, and future performance fluctuations are predicted by some statistical analysis method. (2) Network, server, application, etc., W
Network usage rate and CPU of WW site components
It monitors internal status values such as utilization and issues a warning if they exceed a predetermined threshold. Further, the measurement data is accumulated as time-series data, and future performance fluctuations are predicted by some statistical analysis method.

[0004] However, according to the method shown in the above (1), in order to measure the response time from the outside of the WWW site, if any performance degradation is observed, it is caused by any element constituting the WWW site. It is difficult to identify the problem. In addition, except for the case where a failure occurs in the system, generally, the performance quality can be considered to be caused by a change in user behavior.
The response time is determined as a result of processing by complex components in the WWW site, and it is difficult to assume a model for the time-series change of the response time itself. Therefore, it can be said that it is difficult to generalize a method for predicting a change in the response time value by a statistical method using only the response time value. On the other hand, the above (2)
According to the method described in (1), since each component is complicatedly related to the WWW site, it is difficult to grasp the performance quality of the entire WWW site only by monitoring the state of each component. . For the same reason, it can be said that it is difficult to predict the quality of the hull of the WWW site hull using only the internal state.

The present invention has been made in view of the above circumstances, and models a WWW site in a form capable of taking into account its constituent elements (resources). A means for calculating performance data is implemented, and W
WWW by predicting access fluctuation of WWW site
It is an object of the present invention to provide a WWW site performance monitoring device that predicts site performance fluctuations and implements preventive measures for maintaining performance. In addition, it is determined whether or not the performance determined from the above result can be maintained in the future, and when it is determined that the performance cannot be maintained, it is determined whether or not an element constituting the WWW site is affecting the performance deterioration. It is another object of the present invention to provide a WWW site performance monitoring device that makes a decision based on data and notifies an external system.

[0006]

In order to solve the above-mentioned problems, the present invention provides a method for calculating the above-mentioned WWW site from predicted data of access fluctuation per unit time and monitoring data comprising the use status of components at the WWW site. It is characterized by comprising a performance predicting means for determining whether or not predetermined performance can be maintained based on a model in which constituent elements are abstracted.

Further, in the performance monitoring apparatus for a WWW site of the present invention, when it is determined that the performance determined by the performance prediction means cannot be maintained, the components of the WWW site which may not be able to maintain the performance are determined. It is characterized by predicting.

In the performance monitoring apparatus for a WWW site according to the present invention, the performance prediction means captures and accumulates the access amount per unit time at the WWW site as time-series data of the access fluctuation, and accumulates the data. Access fluctuation prediction means for predicting a future access fluctuation by a statistical method based on the obtained time-series data, and acquiring and accumulating performance data for grasping the use state of the component on the WWW site from the component A site performance monitoring unit, a performance calculation unit that calculates a performance value for each access based on a model in which site components are abstracted from the access fluctuation prediction data and a site performance monitoring data, and the performance calculation unit Judge whether the performance determined based on the performance value for each access calculated in the above can be maintained, and And performance determining means for inferring the components of the WWW site causes can not be maintained serial performance, characterized by comprising a.

In the WWW site performance monitoring apparatus according to the present invention, the access fluctuation predicting means detects an access fluctuation after a specified time from a prediction model obtained by regression analysis of the time series data accumulated by the access fluctuation monitoring means. And calculating a plurality of regression models simultaneously during the regression analysis, and using a regression model having the highest correlation function with the access variation result as a prediction model.

In the performance monitoring apparatus for a WWW site according to the present invention, the model in which the components of the site are abstracted includes a site model in which the components of the site are modeled by a queue model, and a model for each access to the site. , An access model that estimates the load on the components of the site, and the performance calculation means calculates an access response time at a desired time from the access fluctuation prediction and the performance monitoring data of the site based on each of the models. It is characterized in that it is calculated.

In the WWW site performance monitoring apparatus of the present invention, the access response time is based on the access fluctuation prediction, wherein all accesses are created as an event calendar in which their occurrence times are described, and the access time is specified. It is assumed that the state is stationary at intervals, the interior of the system is arbitrarily divided, and events that occur assuming that there is no problem even if sequential processing is performed on events occurring inside the system in the specified time, at the specified time. Is calculated as a step and the calculation for the simulation is performed up to the desired time.

With the above configuration, a performance value for each access is calculated based on a model in which the components of the site are abstracted from the predicted data of the access fluctuation per unit time and the monitoring data including the usage status of the components on the WWW site. Then, it is determined whether or not the performance determined based on the calculated performance value for each access can be maintained, and the component of the WWW site that can not maintain the performance is estimated. As a result, it is possible to perform a preventive measure by predicting the performance degradation of the WWW site, and to perform a more accurate prediction by predicting from the access fluctuation,
Further, since the components of the WWW site are modeled and calculated from the model, it is possible to easily realize which part of the components is the bottleneck when some performance deterioration is expected. .

[0013]

Embodiments of the present invention will be described below with reference to the drawings. First, a configuration example of an entire system in which the performance monitoring device of the present embodiment is used will be described with reference to FIG. In FIG. 1, reference numeral 1 denotes a performance monitoring device according to the present embodiment. Reference numeral 2 denotes a performance monitoring target WWW site, which is an information providing system that discloses information to an accessing client. Reference numeral 3 denotes a client (a customer of the WWW site 2). Symbol 4 is W
This is the Internet (computer network) that connects the WW site 2 and the client 3. Reference numeral 5 is a coping system. The coping system is a system that can perform some operation dynamically to maintain the performance when the performance deteriorates.

The WWW site 2 has a WWW server 21 and
Content database 22 and network device 23
It is composed of The WWW server 21 receives an information disclosure request from the client 3 and provides the corresponding information to the client 3. The content database 22 is a database that records information contents provided by the WWW site 2 in advance. The network device 23 is an interface device for a computer network including a router that connects the Internet 4 and the WWW server 21.

In FIG. 1, when the client 3 accesses the WWW site 2 via the Internet 4 and requests disclosure of information, the WWW site 2 extracts the requested information from the content database 22 and sends it to the WWW server. 21 provided. The performance monitoring device 1 calculates the predicted data of the access fluctuation per unit time and the WWW server 21
From monitoring data consisting of component usage in
Based on a model in which the components of the WWW server 21 are abstracted, it is determined whether the specified performance can be maintained. In addition, when it is determined that the predetermined performance cannot be maintained, the WWW server 2 which is a factor that cannot maintain the performance.
One component is estimated and supplied to the coping system 5. The coping system 5 dynamically increases resources such as servers,
Alternatively, a mechanism capable of reducing or reducing access to the WWW site 2 is provided, but is not directly related to the gist of the present invention, and therefore need not be described here.

FIG. 2 is a diagram showing the internal configuration of the performance monitoring device 1 shown in FIG. Each block shown below is specifically configured by a peripheral LSI including a CPU and a memory, and the CPU realizes its function by reading out a program recorded in the memory and sequentially executing the program. Here, among the components of the performance monitoring device 1, only the performance prediction unit 10 is extracted and shown. The performance prediction unit 10 includes an access fluctuation monitoring unit 11, an access fluctuation prediction unit 12, a site performance monitoring unit 13, a performance calculation unit 14, a performance judgment unit 15, an SLA recording unit 16, a site model DB 17, and an access model DB 1.
8. Access variation time-series data file 19 (hereinafter referred to as
DB19), component usage status file 2
0 (hereinafter simply referred to as DB20).

The access fluctuation monitoring unit 11 has a function of taking in the amount of access per unit time in the WWW site 2 as time series data of the access fluctuation, and storing it in the DB 19. The access change prediction unit 12 has a function of predicting a future access change by a statistical method based on the time-series data accumulated in the DB 19. The site performance monitoring unit 13 has a function of acquiring performance data for grasping the use state of the component in the WWW site 2 from the component and accumulating it in the DB 20. In addition, the access fluctuation prediction unit 12
As described later, the access fluctuation after a specified time is obtained from the prediction model obtained by the regression analysis of the time-series data accumulated by the access fluctuation monitoring unit 11, and a plurality of regression models are simultaneously calculated during the regression analysis. The regression model with the highest correlation function with the result is used as the prediction model.

The performance calculation unit 14 has a function of calculating a performance value for each access from the predicted data of the access fluctuation and the performance monitoring data of the site based on a model in which the components of the site are abstracted. The model in which the site components are abstracted includes a site model (site model DB 17) in which the site components are modeled by a queue model,
For each access to the site, the access model (access model DB 18) is used to estimate the load on the components of the site. The performance calculation unit 14 uses the access fluctuation prediction and the performance monitoring data of the site to determine the access response at the desired time. Time is calculated based on each model. Furthermore,
The access response time is based on the access fluctuation prediction, creates all accesses as an event calendar in which the time of occurrence is described, arbitrarily divides the inside of the system assuming that the state is stationary at a specified time interval, An event that occurs assuming that there is no problem in performing sequential processing in any order with respect to an event that occurs therein is obtained by performing a calculation for a simulation up to a time at which a specified time is first determined as one step. .

The performance determining unit 15 determines whether or not the predetermined performance set in the SLA recording unit 16 can be maintained based on the performance value for each access calculated by the performance calculating unit 14. In addition, it has a function of estimating a component of the WWW site 2 that is a factor that cannot maintain the performance. Note that SL
The A recording unit 16 quantitatively determines target amounts such as system availability and performance.

FIG. 3 is a flowchart cited for explaining the basic operation of the performance monitoring apparatus of the present invention. Hereinafter, the basic operation of the performance monitoring device 1 of the present invention will be described with reference to the flowchart shown in FIG. First, the access change monitoring unit 11 acquires the number of accesses per unit time for each access type grouped or accessed from the WWW server 21 at regular time intervals (step S31). Then, the data is accumulated in the DB 19 as time-series data of the access fluctuation (step S32). Next, the access change prediction unit 12 predicts future access changes based on the access change data stored in the DB 19 by using a statistical method such as regression analysis described later (step S3).
3). On the other hand, the site performance monitoring unit 13 periodically obtains performance data from which the current usage status of resources (components) constituting the site can be grasped from those resources and accumulates them in the DB 20 (step S34).

The performance calculation unit 14 abstracts the resource configuration of the site from the access fluctuation prediction data obtained from the access fluctuation prediction unit 12 and the site performance monitoring data obtained from the site performance monitoring unit 13. A response time value for each access is calculated based on the model DB 17 and the access model DB 18 (step S3).
5). Note that a model in which a site is abstracted includes a site model in which components of the site are modeled by a queue model, such as a WWW server 21 and a content database 22.
An access model indicating how much access to the site places on each element of the model is stored in the site model DB 17 and the access model DB 18, respectively. These models are to be constructed in advance before operating the apparatus of the present invention from load tests in advance, basic performance data of individual resources, program logic, and the like. A simulation, a statistical method, an analytical method, or the like can be used for calculating the response time value.

Next, in the performance judgment section 15, the performance calculation section 1
SL based on the performance prediction result for each access calculated in 4
It is determined whether or not a predetermined performance quality recorded in the A recording unit 16 can be maintained (step S36). Here, if it is determined that it can be maintained, the processing after step S31 is repeated, and if it is determined that it cannot be maintained, it is notified to the administrator or the coping system 5 connected externally. At this time, it is shown which part of the site model element is deteriorating in performance (step S3).
7) By doing so, it is possible to indicate the root cause of performance deterioration.

FIGS. 4 to 14 are diagrams cited for explaining the operation of the embodiment of the present invention, and are diagrams conceptually showing a performance model (FIG. 4), and prediction of access fluctuation by regression analysis, respectively. (FIGS. 5, 6, and 7), a conceptual diagram (FIG. 8) cited to explain the definition of a queue, and a diagram cited to explain modeling of a queuing network. FIG. 9 (FIG. 9), an example of modeling of a WWW site and a diagram (FIG. 10) cited for explaining types of access available on the WWW site at that time, and an example of setting values used for modeling (FIG. 11) ), An example of an event calendar (FIG. 12), a flowchart showing the flow of a response time value calculation process by simulation using a time tracking method (FIG. 13), and separation of a bottleneck part by the performance judgment unit 15 will be described. Is a diagram which is referred to in order (Figure 14).

Hereinafter, referring to FIGS. 4 to 14, FIGS.
The operation of the embodiment of the present invention shown in FIG. 3 will be described in detail. The basic idea of the present invention is to perform, for example, access fluctuation prediction by regression analysis and construct a performance model capable of predicting a response time from an access fluctuation predicted value and an internal state value. The reason is that the bottleneck part can be grasped by modeling the resource. FIG. 4 schematically shows the performance model.

Next, the prediction of access fluctuations by regression analysis of the access fluctuation prediction unit 12 will be described with reference to FIGS. FIG. 5 is a flowchart for explaining the operation of the access change prediction unit 12, and FIG.
7 is a schematic diagram illustrating the derivation of the prediction model of FIG.
FIG. 4 is a schematic diagram illustrating an example of an information table for deriving and selecting a prediction model in the access change prediction unit 12. The access fluctuation monitoring unit 11 first obtains time-series data on the number of accesses per unit time from the access log of the WWW server 21 and stores the data in the DB 19. The access change prediction unit 12 extracts data in a certain history range of the time-series data from the DB 19 (step S331 in FIG. 5), and derives an optimal model function by applying the data to a plurality of model functions by regression analysis. For example,
The determination of this model is calculated by the following model determination method. (1) Within a certain range, the measurement time and the number of accesses have a linear relationship. In this model, the relationship is

(Equation 1) It is expressed in the form. (2) Within a certain range, the measurement time and the number of accesses have a quadratic curve relationship. In this model, the relationship is

(Equation 2) It is expressed in the form.

The coefficients of these models are derived by regression analysis from the given data, and the models are determined (step S333). As each model, not only a simple regression model as described above, but also a method using a neural network or the like in addition to a non-linear regression model represented by an exponential function, a logarithmic function or the like can be considered. Then, the prediction model theoretical value calculated by each model is compared with the actual access value, and the most suitable model is determined as the current prediction model (step S334). Here, as a method of determining the most suitable model, for example, there is a method of obtaining a correlation coefficient between a predicted model theoretical value and an actual access value, and employing a model having the largest correlation coefficient. In the example shown in FIG. 6, the quadratic curve model (square model) is specified as the most suitable model.

In the present embodiment, it is considered that functions can be represented by functions defined by three types of models, <linear model>, <square model>, and <root model> shown in FIG. Then, optimal coefficients a and b such that the combination of x and y best fit the model are obtained by, for example, a minimum approximation method. The correlation coefficient of each model is calculated, and here,
The root model with the highest value of the correlation coefficient is determined as the most suitable prediction model this time. Using the prediction model determined as described above, the access variation prediction unit 12 calculates a predicted value of the access variation in the set prediction range (Step S335).

Next, the definition of a queue used when the response time is calculated by the performance calculator 14 will be described. Here, the queue is described using terms such as access, service mechanism, and service window. To define a queuing model, it is necessary to consider the size of the population, the arrival of accesses, the distribution of service times, the maximum system capacity, the number of windows, and the criteria for queuing. Although there are many models in queuing theory, it is noted here that some terms not used in actual queuing theory are included for the sake of simplicity. That is, the queue defined in the queuing theory is not correctly defined.

The operation of an access arriving at the queue will be described below with reference to FIG. First, if there is an empty queue in the system, the access arrives at the queue (step S81). If the service windows are all closed, the accesses are arranged in the queue in the order of arrival (step S82). Then, when the service window becomes available, the access at the head of the queue enters the available service window (step S83). At the service window, the access receives a service amount according to the processing capability of the service window derived from the performance of the service mechanism and the number of simultaneous accesses (step S84). Then, when the access receives a load (service amount) determined for each access at the service window, the access leaves the service mechanism (step S85). The time required for steps S81 to S85 is referred to as the system time, the time required for steps S81 to S83 is referred to as a waiting time, and the time required for steps S83 to S85 is referred to as a service time.

Next, modeling of the above-mentioned queue network will be described. Here, classification is performed as shown in FIGS. 9A, 9B, and 9C according to the setting conditions of the queue. 9 (a), 9 (b) and 9 (c) all show the configuration of the queuing network, a graph with the processing capacity on the vertical axis and the number of simultaneous accesses on the horizontal axis. That is, (a) is a category having no queue and a certain waiting time, (b) is a category having a queue and has a certain waiting time, and (c) is a category having a certain waiting time.
There are queues and they are divided into categories with processing capacity that is inversely proportional to the number of simultaneous services because they use shared resources. Therefore, in (b), the processing capacity of the service window is equal to the performance, and in (c), the processing capacity of the service window is indicated by performance / number of simultaneous accesses.

FIG. 10 shows an example of modeling a WWW site. In the figure, (a) shows the system configuration of the WWW site 2, and (b) shows the type of access (ID and URL (Uniform Resource) available at the WWW site 2.
(C) is a queuing network model created corresponding to the system components of the WWW site 2. In the figure, blocks denoted by the same reference numerals as the blocks shown in FIG. 1 are the same as those shown in FIG.
Here, the WWW site 2 includes a content database (DB) 22, a WWW server 21, and an application (AP) server 100 as main elements.
Both have a LAN (Local A) with a capacity of 100 Mbps.
rea Network).

FIG. 11 shows each of the set values used for the modeling. In FIG. 11, (a)
This is a set value for each URL obtained in advance from the program logic and the load test of the WWW site 2, and is applied to the system shown in FIG. Here, the WWW server 21 /
It is shown separately in the AP server 100 and the content database 22. (B) shows W as a service mechanism.
This is a performance value of the WW21 / AP server 100 or the content database 22. (C) is a queue setting value, in which the maximum number of queues, the number of simultaneous processing services, the number of shared resources, and the waiting time are set for the WWW server 21 / AP server 100, the content database 22, and the client 3. Here, the “performance value” indicates the processing capacity of the server indicated by the queue, and the “set value” indicates the behavior of the access determined in the queuing network determined by the type of access. Refers to the load on the queuing network.

Next, a method of calculating the response time value by the performance calculation unit 14 will be described. Performance calculator 14
As described above, has a function of calculating a performance value for each access based on a model in which site components are abstracted from predicted data of access fluctuation and performance monitoring data of the site. Here, it is assumed that the response time is calculated by simulation using the time tracking method. All accesses are described as an event calendar whose time of occurrence is described. It is assumed that the event calendar is created from the access fluctuation value predicted by the access fluctuation prediction unit 12. FIG. 12 shows an example of the event calendar 120. Here, focusing on a certain short time Δt, assuming that the state is stationary during that time, and dividing the inside of the system into several parts and performing sequential processing on events occurring inside the system in any order Events that occur assuming that there is no problem in doing so are calculated. Under such conditions, the process of performing the calculation up to the time T _END where Δt is to be obtained as one step is described in the flowchart shown in FIG.

In the flowchart shown in FIG. 13, first, an access occurrence time is set in a variable t (step S351). Then, based on the queue, the service processing at the service window is performed, and the access existing in the queue is allocated to the service window (steps S353 and S353). Next, access from the event calendar 120 is obtained (step S354),
The access to the queue is again assigned to the service window (step S355). Then, the variable t is updated by Δt and the step proceeds (step S3).
56) The calculation is repeatedly performed until the time T _{END at} which the above processing is desired to be obtained (step S357).

Finally, an example of the bottleneck separation performed by the performance judgment unit 15 will be described. The performance determining unit 15 determines whether or not the specified performance set in the SLA recording unit 16 can be maintained based on the performance value for each access calculated by the performance calculating unit 14, and The components of the WWW site 2 that may not be able to maintain the above are estimated. The response time value at WWW site 2 is
It can be expressed as the sum of the in-system time in each queue in the WWW site 2. Therefore, it is possible to know which part of the WWW site 2 is taking a long time to process from the in-system time of each queue. FIG. 14 shows the in-system time in the queue for each ID and the response time value.
It can be understood that the processing time of B22 affects the response time.

The performance monitoring apparatus 1 shown in FIG. 2 records a program for realizing the function on a computer-readable recording medium, and causes the computer system to read the program recorded on the recording medium. ,
By executing the functions, the functions of the respective devices described above may be realized.

Here, the "computer system" includes an OS and hardware such as peripheral devices.
If a WW (World Wide Web) system is used, a homepage providing environment (or display environment) is also included. The “computer-readable recording medium” refers to a portable medium such as a floppy (registered trademark) disk, a magneto-optical disk, an OM, a CD-ROM, or a storage device such as a hard disk built in a computer system. . Further, a “computer-readable recording medium” refers to a medium that dynamically stores a program for a short time, such as transmitting a program via a computer network such as the Internet or a communication line such as a telephone line. (Transmission medium or transmission wave), in which case a program holding a program for a certain period of time, such as a volatile memory in a computer system serving as a server or a client, is also included.

The above-mentioned program may be for realizing a part of the above-mentioned functions, and may be for realizing the above-mentioned functions in combination with a program already stored in a computer system, that is, a so-called program. It may be a difference file (difference program).

[0039]

As described above, the present invention holds the internal configuration model of the WWW site, and predicts the performance of the site by predicting the fluctuation of the site access on a regular basis. When it is determined that the performance cannot be maintained, a bottleneck portion in the WWW site is notified to the coping system to promote the improvement. For this reason, prediction data of access fluctuation per unit time,
A performance value for each access is calculated based on a model in which the components of the site are abstracted from the monitoring data including the usage status of the components on the WWW site, and the performance determined based on the calculated performance value for each access is maintained. WW, which determines whether or not it can be performed and is a factor that cannot maintain its performance
The components of the W site were assumed.

As a result, it is possible to predict the performance degradation of the WWW site and take preventive measures. According to the present invention, it is possible to perform more accurate prediction by predicting from the access fluctuation, and since the components of the WWW site are modeled and calculated from the model,
When some performance deterioration is expected, it is possible to estimate which part of the component is the bottleneck. Therefore, by predicting fluctuations in WWW site performance in advance and taking preventive measures based on the results, WWW
The effect that the predetermined performance of the site can be maintained is obtained.

[Brief description of the drawings]

FIG. 1 is a diagram cited for describing a configuration example of an entire system in which a performance monitoring device according to an embodiment of the present invention is used.

FIG. 2 is a block diagram showing the internal configuration of the performance monitoring device shown in FIG.

FIG. 3 is a flowchart cited for explaining a basic operation of the embodiment of the present invention.

FIG. 4 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a diagram conceptually showing a performance model.

FIG. 5 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a diagram cited for explaining prediction of access fluctuation by regression analysis.

FIG. 6 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a diagram cited for explaining prediction of access fluctuation by regression analysis.

FIG. 7 is a diagram cited for explaining an operation of the embodiment of the present invention, and is a diagram cited for explaining prediction of access fluctuation by regression analysis.

FIG. 8 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a conceptual diagram cited for explaining the definition of a queue;

FIG. 9 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a diagram cited for explaining modeling of a queuing network.

FIG. 10 is a diagram cited for explaining the operation of the embodiment of the present invention, which is an example for modeling a WWW site, and a diagram cited for explaining types of access available on the WWW site at that time. is there.

FIG. 11 is a diagram cited for describing the operation of the embodiment of the present invention, and is a diagram illustrating an example of a set value used for modeling.

FIG. 12 is a diagram cited for describing an operation of the embodiment of the present invention, and is a diagram illustrating a configuration example of an event calendar.

FIG. 13 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a flowchart showing a flow of a process of calculating a response time value by simulation using the time tracking method.

FIG. 14 is a diagram cited for explaining the operation of the embodiment of the present invention, and is a diagram cited for explaining separation of a bottleneck part by a performance judgment unit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Performance monitoring apparatus 2 WWW site 3 Client 4 Internet 5 Response system 11 Access fluctuation monitoring unit 12 Access fluctuation prediction unit 13 Site performance monitoring unit 14 Performance calculation unit 15 Performance judgment unit 16 SLA recording unit 17 Site model DB 18 Access model DB 19 Access variation time-series data file 20 Component usage status file 21 WWW server 22 Content database 23 Network equipment

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B042 GA12 HH07 HH20 MA08 MA14 MB03 MC22 MC25 MC28 MC35 5B089 GA01 GB02 HA10 JB16 KA06 KA07 KA12 KB04 KC48 MC02

Claims (6)

[Claims] 1. Based on predicted data of access fluctuations per unit time and monitoring data composed of the usage status of components on a WWW site, a predetermined performance is determined based on a model that abstracts the components of the WWW site. A performance monitoring device for a WWW site, comprising: a performance predicting unit that determines whether or not it can be maintained. 2. The system according to claim 1, wherein when it is determined that the performance determined by the performance prediction means cannot be maintained, a component of the WWW site that is a factor that cannot maintain the performance is estimated. A WWW site performance monitoring device as described. 3. The performance prediction unit includes: an access variation monitoring unit that captures and accumulates an access amount per unit time in a WWW site as time series data of an access variation; and a statistical information based on the accumulated time series data. Access fluctuation prediction means for predicting future access fluctuations by a technique; site performance monitoring means for acquiring performance data for grasping the use state of the component in the WWW site from the component and accumulating the data; A performance calculating means for calculating a performance value for each access based on a model in which site components are abstracted from the prediction data and the performance monitoring data of the site; and a performance value for each access calculated by the performance calculating means. Judge whether the performance stipulated in can be maintained, and
3. The performance monitoring device for a WWW site according to claim 1, further comprising: a performance determination unit configured to estimate a component of the WWW site that is a factor that cannot maintain the performance. 4. The access fluctuation prediction means obtains an access fluctuation after a specified time from a prediction model obtained by regression analysis of the time-series data accumulated by the access fluctuation monitoring means, and executes a plurality of regression models during the regression analysis. 4. The WWW site performance monitoring apparatus according to claim 3, wherein a regression model calculated at the same time and having the highest correlation function with the access fluctuation result is used as a prediction model. 5. A model in which the site components are abstracted includes a site model in which the site components are modeled by a queue model, and a load on the site components is estimated each time the site is accessed. 4. An access model according to claim 3, wherein the performance calculation means calculates an access response time at a desired time point from the access fluctuation prediction and the performance monitoring data of the site based on the models. A WWW site performance monitoring device as described. 6. The access response time is based on the access fluctuation prediction, wherein all accesses are created as an event calendar in which their occurrence times are described, and it is assumed that a state is stationary at the specified time interval. The inside of the system can be arbitrarily divided, and events that occur assuming that there is no problem even if sequential processing is performed on events occurring in the system in any order can be performed by the specified time as one step up to the desired time. 6. The WWW site performance monitoring apparatus according to claim 5, wherein the performance monitoring apparatus obtains the performance by performing calculation.