JP2001092793A - Method for estimating proper arrangement of distributed object, method and device for determining proper arranging position - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method capable of properly standardizing loads of plural servers and contributing to securing of target responsiveness of client application or target performance in complicated execution environment. SOLUTION: Execution frequency of plural applications to use distributed objects in specified monitoring period is detected, the execution frequency of each distributed object per specified time is calculated based on the detected execution frequency of each application, information regarding mutual call relation between the distributed objects and information of the distributed objects to be used by each application, performance when each application is executed is further estimated based on information regarding execution characteristics of each server, information regarding execution characteristics of each distributed object and information of the execution frequency of the distributed objects per specified time after the loads of each server are estimated and its estimation result is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system or a computer which distributes objects to a plurality of server computers, sequentially calls each distributed object from a plurality of application programs mounted on a client computer, and executes predetermined processing. The present invention relates to a method for estimating a proper arrangement condition of a distributed object and a method and an apparatus for determining an appropriate arrangement in a network system.

[0002]

2. Description of the Related Art Computer usage has shifted from batch processing and TSS processing by a host to an architecture centered on a network such as a client / server system due to the development of PCs and workstations. Particularly in recent years, system development based on object orientation has been increasing in order to enhance the reusability of programs. Objects provide a more specific approach to modeling the real world handled by programmers, as well as conceal information by permitting access to data in objects only through the interfaces in the objects, and It is intended to prevent dangerous operations such as rewriting data.

[0003] In the conventional object-oriented technology,
Although the calculation is performed by exchanging objects stored in one computer, the distributed object environment for performing the calculation by distributing the objects to a plurality of computers on a network has been developed. Is coming.

As such a distributed object environment, there is an OMG (Object Management Group) in the United States.
And the Java-RMI attached to the Java language, and DCOM proposed by Microsoft. Using a distributed object environment has many advantages over conventional simple client-server systems, but it still has many problems.

[0005] In a distributed object environment, objects related to the execution of one application are located on many computers. Also, a plurality of applications are executed in the same network environment. Further, the plurality of applications have a very complicated execution environment such as sharing the same distributed object.

In such an environment, it is difficult to avoid an overload on the server and maintain an environment in which the application can be executed with appropriate performance (performance such as a response time until an execution state is reached in response to a start instruction). Work. For this reason,
The system administrator attempts to improve the response of the system by relocating the distributed objects located on the server computer (hereinafter abbreviated as a server) with a high load to another server.

However, in such a manual relocation, how to perform the relocation is entirely left to a human, and when a distributed object having a high load is simply moved to another server, the relocated server is relocated. May be overloaded.

On the other hand, in a network system, attempts have been made to dynamically change the arrangement of servers. For example, in Japanese Patent Application Laid-Open No. H10-105500 (dynamic reconfiguration of a network server), when the load on a server increases, a part of a request from a client computer (hereinafter abbreviated as a client) to the server is redirected to another server. Thus, a method has been proposed in which another server responds to a client request. However, even in this method, even if the proxy server when the server is overloaded is determined in advance, if the number of servers increases, the response performance to the client after the redirect may not be guaranteed.

[0009]

In a complex computer network environment composed of a large number of distributed objects and a large number of servers, the response time to a request from a client is greatly affected by which server the distributed objects are located. .

When a large number of distributed objects are dynamically or statically relocated only by the load of each server during execution of the distributed system, the load state is simply transferred to a server different from the server which has become a problem due to the relocation. May be the only result. Therefore, there is a problem how to properly arrange a large number of distributed objects in the entire system.

When considering such an appropriate arrangement, especially in an intranet used by a recent company or the like, the number of servers is large, and the network is divided by a device such as a router in order to secure a response of the network. Consideration must be given to the increasing complexity of distributed object execution environments.

That is, in an intranet used in recent enterprises and the like, there is a restriction that a specific distributed object must be placed on a specific server divided by a router in order to secure system security. May be imposed.

For example, there is a restriction that a distributed object related to an application that handles confidential matters must be placed on a server belonging to a specific department that is allowed to handle the application. In addition, since a large-capacity database can be operated only on a specific server and cannot be moved to another machine, distributed objects that directly use the database are allocated to a specific server that operates the database. There is a restriction that it must be.

According to the present invention, in such a complicated execution environment, the load of a plurality of servers is appropriately leveled, and the appropriateness of a distributed object that can contribute to ensuring the responsiveness or the target performance of a target client application. A first object is to provide a method for estimating an arrangement condition and a method and an apparatus for determining an appropriate arrangement position.

Further, even when there are restrictions on the arrangement of distributed objects, a method for estimating proper arrangement conditions of distributed objects and determining an appropriate arrangement position can contribute to realizing proper arrangement of distributed objects so as to satisfy the arrangement restrictions. It is a second object to provide a method and apparatus.

[0016]

In order to achieve the first object, a proper arrangement estimating method according to the present invention comprises a plurality of applications mounted on a client computer for sequentially calling a distributed object and executing a predetermined process. Application execution monitoring means for detecting the number of executions of the program during a predetermined monitoring period, a relation description file in which a mutual call relation between distributed objects distributed in a plurality of server computers is registered, and a plurality of applications mounted on the client computer An application description file that registers which distributed object the program uses, a server description file that registers the execution performance of each server computer, and a distributed object between multiple server computers in the network. A network call time relational description file in which information on a required time for taking out is registered, and an object characteristic description file in which execution characteristics of each distributed object distributed in the plurality of server computers are registered, and detected by the application execution monitoring means. Based on the information on the number of executions of each application program and information on the mutual call relationship between distributed objects registered in the relationship description file and information on distributed objects used by each application program registered in the application description file, A first step of calculating the number of executions of each distributed object per predetermined time, information on execution characteristics of each server computer registered in the server description file, and the object characteristic description A second step of estimating the load on each server computer based on the information on the execution characteristics of each distributed object registered in the file and the information on the number of executions of the distributed object per predetermined time calculated in the first step. Based on the information on the load estimation result and the information on the required call time between the distributed objects registered in the network characteristic description file, the performance at the time of executing each application program is estimated, and the estimation result is used to determine the proper arrangement of the distributed object. And a third step of outputting as information for determination.

Further, the distributed object arrangement position determining method according to the present invention is an application execution method for detecting the number of executions in a predetermined monitoring period of a plurality of application programs mounted on a client computer for sequentially calling a distributed object and executing a predetermined process. A monitoring means, a relation description file in which mutual call relations between distributed objects distributed in a plurality of server computers are registered, and which distributed object is used by a plurality of application programs mounted on the client computer are registered. An application description file, a server description file in which the execution performance of each server computer is registered, and information on the time required for calling between distributed objects in a plurality of server computers in the network. A network call time relation description file, an object characteristic description file in which execution characteristics of each distributed object distributed in the plurality of server computers are registered, and an application target in which information on target performance for each application program is registered. A performance description file, information on the number of executions of each application program detected by the application execution monitoring means, information on mutual call relationships between distributed objects registered in the relationship description file, and information registered in the application description file. A first step of calculating the number of executions of each distributed object per predetermined time based on information on the distributed objects used by each application program; The information on the execution characteristics of each server computer, the information on the execution characteristics of each distributed object registered in the object characteristic description file, and the information on the number of executions of the distributed object per predetermined time calculated in the first step. A second method for estimating the load on each server computer based on the
And a third step of estimating the performance at the time of executing each application program based on the information on the load estimation result and the information on the required call time between the distributed objects registered in the network characteristics description file. A fourth step of determining a distributed object placement server that satisfies the target performance based on the information on the performance estimation result at the time of program execution and the information on the target performance for each application program registered in the application target performance description file. And characterized in that:

Further, in order to achieve the second object,
The proper placement estimation method and the placement position determination method of the present invention,
A distributed object placement constraint file in which a distributed object restricted to be placed on a specific server computer or a server group is registered, and a distributed object placement server that satisfies a target performance in the fourth step or the fourth processing means; When determining the distribution object, a distribution server of a distributed object that satisfies the arrangement restriction conditions registered in the distributed object arrangement restriction file is determined.

Further, the proper arrangement estimating apparatus of the present invention comprises an application execution monitoring means for detecting the number of executions of a plurality of application programs mounted on the client computer in a predetermined monitoring period, and a distributed execution distributed to a plurality of server computers. A relation description file in which a mutual call relationship between objects is registered, an application description file in which a plurality of application programs mounted on the client computer use which distributed object, and an execution performance of each server computer are registered. A server description file, and a network call time relationship description file in which information on a required call time between distributed objects in a plurality of server computers in the network is registered. An object property description file in which the execution properties of each distributed object distributed in the server computer are registered, information on the number of executions of each application program detected by the application execution monitoring means, and a distributed object registered in the relation description file. First processing means for calculating the number of executions of each distributed object per a predetermined time based on information on a mutual call relationship and information on distributed objects used by each application program registered in the application description file; Information on the execution characteristics of each server computer registered in the description file, information on the execution characteristics of each distributed object registered in the object characteristic description file, and the predetermined value calculated by the first processing means. Second processing means for estimating the load on each server computer based on information on the number of times of execution of the distributed object per interval; information on the load estimation result and call requirements between the distributed objects registered in the network characteristic description file; And a third processing unit for estimating the performance at the time of executing each application program based on the time information, and outputting the estimation result as information for determining an appropriate arrangement of the distributed objects.

Further, the arrangement position determining apparatus of the present invention is an application execution monitoring means for detecting the number of executions in a predetermined monitoring period of a plurality of application programs mounted on a client computer for sequentially calling a distributed object and executing a predetermined process. A relationship description file in which a mutual call relationship between distributed objects distributed in a plurality of server computers is registered, and an application description in which a plurality of application programs mounted on the client computer use which distributed object File, a server description file in which the execution performance of each server computer is registered, and a network in which information on the time required for calling between distributed objects in a plurality of server computers in the network is registered. A work call time relation description file, an object characteristic description file in which execution characteristics of each distributed object distributed in the plurality of server computers are registered, and an application target performance description file in which information on target performance of each application program is registered And information on the number of executions of each application program detected by the application execution monitoring means, information on the mutual call relationship between the distributed objects registered in the relation description file, and each application program registered in the application description file. First processing means for calculating the number of executions of each distributed object per predetermined time based on the information of the distributed objects to be executed, and each server computer registered in the server description file. Based on the information on the execution characteristics of the distributed object, the information on the execution characteristics of each distributed object registered in the object characteristic description file, and the information on the number of times of execution of the distributed object per predetermined time calculated by the first processing means. Second processing means for estimating the load on each server computer, and the performance at the time of executing each application program based on information on the result of the load estimation and information on the required call time between distributed objects registered in the network characteristic description file. And a third processing means for estimating the target performance, and information on a performance estimation result at the time of execution of each application program and information on a target performance for each application program registered in the application target performance description file. Distribution of distributed objects And a fourth processing means for determining a server.

[0021]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a system configuration diagram showing an embodiment of an information network system to which the present invention is applied. This system includes a plurality of network systems NW1 to NW3 each including a plurality of server computers (hereinafter abbreviated as servers) 101 and a client computer (hereinafter abbreviated as clients) 102.
Are connected by routers 103 and 104 to form an intranet environment.

The distributed objects are arranged on each of the servers 101 constituting the intranet environment, and execute a specific business process by making a call between the distributed objects.

FIG. 2 is a block diagram showing a functional configuration of a proper placement determining apparatus which is placed on one of the servers 101 and realizes proper placement of a distributed object. 202, characteristic description file 203, network description file 204, server description file 20
5. An arrangement constraint description file 206 and an application target description file 207 are provided. Also, the monitoring device 208, the definition analysis unit 209, the server load table 21
0, an object arrangement table 211, a response table 212, an optimization unit 213, a display unit 214, and a display device 215.

Distributed object relation description file 201
Stores information on a calling relationship between distributed objects. The application description file 202 stores information on distributed objects used in an application executed on each client. The characteristic description file 203 stores the execution characteristics of each distributed object. Here, the execution characteristic is information that affects execution performance, such as the occupation time of a server resource per unit time, as described later.

The network description file 204 stores the execution time required for calling between the distributed objects arranged on each server 101. The server description file 205 stores the processing capability of each server 101.

In the arrangement constraint description file 206, when distributing a distributed object to a specific server, constraint conditions relating to the arrangement are stored.

Application target description file 207
The target value of the response time for each application and priority information indicating which application target value should be prioritized when determining an appropriate arrangement with the target value as a target. Information on target performance is stored.

The monitoring device 208 monitors and records the number of executions of an application executed on the intranet system at a preset monitoring time within a set time width.

The definition analysis unit 209 is a distributed object called by each application based on a relational description of a call relation between distributed objects stored in the distributed object relational description file 201 and an application description stored in the application description file 202. And the number of times of execution of each distributed object in a unit time from the number of times of execution of the application obtained by the monitoring device 208.

The server load table 210 records the load value of each server 101. In the object arrangement table 211, a server name on which each distributed object is arranged is set. Response table 21
In 2, response information of each application calculated by the optimization unit 213 is recorded.

The optimization unit 213 includes information on the number of executions of each distributed object analyzed by the definition analysis unit 209, the server performance of the server description file 205, and the performance of calling the distributed object between servers stored in the network description file 204. Is used to determine the arrangement of distributed objects on the server that optimizes the response of each application.

The display unit 214 displays the arrangement of the distributed objects on the server determined by the optimization unit 213 on the display device 215.
Is displayed on the screen.

FIG. 3 shows an example of data representing a related description of a distributed object stored in the distributed object relation description file 201 of FIG. In this example, the number of distributed objects to be called from each distributed object is described. For example, the distributed object Ob001 (30
1) are Ob002 (302) and Ob003 (30)
3) and Ob005 (305) are called once each.

FIG. 4A is an example of data indicating the correspondence between an application and a distributed object stored in the application description file 202. This data defines the relationship of the distributed objects called by the application executed on the network of FIG. Here, for example, the application A001 (4
11) calls the distributed objects Ob002 (402), Ob003 (403), and Ob005 (405) once, and does not call other distributed objects.

FIG. 4B shows an example of data stored in the application target description file 207. The application target description file 207 includes a target response time 423 which is a target value of a response time for each application, and a priority indicating which application target value is to be prioritized when determining an appropriate arrangement with the target value as a target. The information of the degree 422 is stored by being divided by the application ID 421. Here, the higher the numerical value of the priority 422, the higher the priority. The unit of the numerical value of the target response time 423 is, for example, millisecond (ms). In addition, a unit related to a target performance can be used.

FIG. 5 shows an example of the data of the execution characteristic description file 203 representing the processing of the server necessary for one call of the distributed object. The distributed object ID 501 and its execution characteristic 502 are defined. It has become. For example, distributed object Ob001
This means that one execution of (501) requires 25 units of server resources per unit time. Here, one unit represents the occupation time of the server resource per unit time,
The larger the numerical value, the longer the occupation time of the server resource.

FIG. 6 shows a network description file 204.
3 shows an example of the data. The network description is data expressing communication performance between servers when using the network system in FIG. In the example of FIG. 6, the time required for communication between the distributed objects arranged in each server 101 is stored. For example, the communication station time between distributed objects arranged in the server S001 (601) is one unit time, and S001 (601) and S001 existing in the same network segment.
The time required for the communication station between 002 (602) and the server S001 (601) and the server S004 (604) arranged in separate segments spanning the router require 7 unit time. ing.

FIG. 7 shows an example of data in the server description file 205, in which a server ID 701 and a processing performance 702 are defined. The server description expresses the processing capability of each server 101.
In the example of FIG. 7, the processing capacity (performance) per unit time of each of the servers S001 to S006,. For example, the processing capability value (70) of the server S001 (703)
4) indicates 35 units. The larger the numerical value of the processing capacity value (704), the higher the processing capacity.

FIG. 8 shows an example of data of the placement constraint description file 206 of the distributed object, in which a distributed object ID 801 and a placement constraint 802 are defined. When a distributed object is arranged on a server, as described above, it cannot be arranged completely freely on a network. For example, a distributed object related to an application that is allowed to be used only in a specific department from the viewpoint of security can be placed only on a server belonging to the department. Further, a distributed object accessing a database (DB) is required to be arranged on a server where the DB exists. Also, distributed objects that are frequently called by each other need to be arranged on the same server or a server in the same segment. The distributed object arrangement constraint defines the arrangement condition of such a distributed object. In the example of FIG. 8, for the distributed object Ob001, the server S0
01, S002, and S003. Also, the distributed object Ob0
02 indicates that it must be arranged in the server S002. Also, the distributed object Ob
Reference numeral 004 indicates that the allocation to the same server as the distributed object Ob001 is required.

FIG. 9 shows the load on each server based on the application start frequency (or the number of times of execution) obtained by the monitoring device 208 at the time when the allocation of the distributed object to the server is determined by the optimization unit 213. 9 shows an example of data of a server load table 210 storing calculated values. For example, a numerical value indicating that the load of the server S001 is 0.75 is stored. The larger the value of this numerical value, the greater the load.

FIG. 10 shows the object arrangement table 21.
1 shows an example of data of each distributed object I
D1001 and server ID 1002 where these are arranged
Is defined. In this table 211, the server ID of the most appropriate arrangement destination calculated by the optimization unit 213 is stored. In the example of FIG. 10, the distributed object Ob001 is the server S001 and the object Ob002 is the server S
002 and Ob003 are stored in the server S007 to be optimally arranged.

FIG. 11 shows an example of data in the response table 212.
1 and a corresponding response 1102 are stored. Response 1102 is the optimization unit 2
A queue model is used based on the load value obtained by calculating the load on each server at the stage when the arrangement of the distributed objects is determined in the distributed object arrangement calculation process in 13 and the number of times of execution of the application obtained by the monitoring device 208. Is calculated. Here, the unit of the numerical value of the response 1102 is, for example, millisecond (ms),
The smaller the numerical value, the shorter the response time.

FIG. 12 is a flowchart showing an outline of the processing of the optimizing unit 213 in FIG.

The optimizing unit 213 first sets a response time target value and priority for each application, and registers it in the application target description file 207 of FIG. 4B (step 1201).
Here, the target response time and priority to be registered are inputted by an administrator of the system from an input device (not shown).

Next, an optimal arrangement process is performed (step 120).
2), determine an optimal server to place each distributed object. The outline of this processing is as follows. First,
Information on the execution characteristics of each server computer registered in the server description file 205, information on the execution characteristics of each distributed object registered in the characteristic description file 203, and execution of the distributed object per predetermined time calculated by the definition analysis unit 209 The load of each server computer is estimated based on the information on the number of times.

Next, the performance at the time of executing each application is estimated based on the information on the load estimation result and the information on the required call time between the distributed objects registered in the network description file 204. Next, based on the information on the performance estimation result at the time of execution of each application and the information on the target performance for each application registered in the application target description file 207,
Determine the allocation server of the distributed object that satisfies the target performance. Finally, the result of the optimal arrangement processing is displayed on the display device 21.
5 is output. The process of actually arranging the distributed objects or the process of changing the current arrangement based on the processing result is executed by waiting for an instruction from the system administrator, or is automatically executed without waiting for the instruction. Although there is a method, any may be adopted.

Hereinafter, the optimal arrangement processing (step 1202)
Will be described in detail. In the present embodiment, the optimal arrangement processing is realized using a genetic algorithm. This is because a genetic algorithm has advantages such as easy handling of a function that is more difficult than linearity and easy solution even if the purpose is ambiguous.

In the genetic algorithm, the arrangement of distributed objects on the server is represented by a one-dimensional array as shown in FIG. A one-dimensional array that provides such a correspondence between one distributed object and a server is called a gene. The example shown in FIG. 13 is the object arrangement table 2 in FIG.
This is a representation of the arrangement of the distributed objects shown in FIG. 11 on the server as a genetic expression.

Next, as shown in FIG. 14, an initial population consisting of a plurality of genes is prepared, and the plurality of genes are imitated by imitating the evolution of an organism, such as selection, crossover, mutation, and generation of a next-generation population. Make crossovers and mutations to improve gradually. In FIG. 14, each circle represents one gene.

As shown in FIG. 15, “crossing” means that when two genes 1501 and 1502 are multiplied,
03,1504.

Referring to FIG. 16, the optimal arrangement processing by the genetic algorithm will be described in detail. First, an initial gene group G1 is created (step 1601). The initial population G1 contains a predetermined number of genes. Here, the number of generations is indicated by a variable t, and the initial group is t1. FIG. 17 shows the details of the method for generating the genes included in the initial population G1.
Is shown in In the processing of FIG. 17, each distributed object is randomly arranged within a range that satisfies the conditions specified in the arrangement restriction description file 206 of the distributed object shown in FIG.

For example, the distributed object Ob001 of FIG.
Has a restriction of being placed in the servers S001, S002, and S003, and randomly selects a server to be assigned from the three. Ob002 is assigned to S002 because it is limited to server S002. Since Ob003 has no placement restriction, the server description file 205 in FIG.
One server is randomly selected from all the servers registered in. The distributed object Ob004 is Ob0
01 because it is a constraint to be placed on the same server as Ob00
Select the same server as 1. Genes belonging to the initial population G1 are randomly generated in this manner. In the processing of FIG. 17, along with the generation of such random genes, the generation of genes by assignment in consideration of the processing performance of the server is also performed. This is because, if allocation to a plurality of servers is possible in the allocation constraint description file 206, the allocation is desirably prioritized to a server having a high processing capacity.

In the arrangement restriction description file 206 of the distributed object, Ob001 indicates the servers S001, S002,
There are three options of S003. The processing capacities of these servers are "35", "64", and "5", respectively, as defined in the server description file 205 of FIG.
8 ”, the probability that the server S001 is selected is 3
5 / (35 + 64 + 58), the selection probability of S002 is 64
/ (35 + 64 + 58), the selection probability of S003 is 58 /
(35 + 64 + 58) as S001, S002, S0
Select one server from 03.

Using such a random operation, a plurality of genes are created to construct an initial population G1. By constructing the initial population G1 of genes using such a random operation, various genes can be generated, and the search range of the optimal solution by the genetic algorithm can be expanded. In addition, by adding a gene that gives a server with high processing capacity a probability priority, improvement in solution search efficiency can be expected.

Next, a termination condition is determined (step 1).
602). The termination condition is when the performance target of each application registered in advance has been achieved or the number of generations t has reached a predetermined value.

Next, a new gene is generated by taking out the gene from the group of genes with a predetermined probability and executing the intersection (step 1603). Next, the gene is taken out from the group of genes with a certain probability and a mutation process is performed (step 1604). This process is a process of changing a code at a random position of a gene randomly extracted with a certain probability.

Next, the genes belonging to the group obtained by adding the new genes generated by the crossover and mutation to the genes belonging to the original group are evaluated (step 1605). Gene evaluation refers to checking how much the response time of each gene satisfies the target response time including priority, and the lower the evaluation value, the less satisfied. I have.

Next, the gene is propagated in accordance with the evaluation value for each gene to generate a next-generation gene population (step 1606).

Next, it is determined whether the end condition is satisfied (step 1602). If not, the same processing is repeated until the end condition is satisfied.

FIG. 17 shows the initial population Gt of the genes of FIG.
It is a flowchart which shows the detail of step 1601 which produces (t = 1). In the process of creating the initial population G1 of the genes, a part of the gene population is randomly generated, and the remaining genes are initialized, and the number of genes randomly generated in the variable M1 and the number of remaining genes are previously stored in the variable M2. First, 1 is assigned to a variable N (step 1
701). Next, the value of the variable N is compared with M1 (step 1702). If N is equal to or less than M1, a random initial solution is generated (step 1703).

The generation of this random initial solution is performed in the following procedure. The distributed object on the gene shown in FIG.
If the placement constraint is registered in the distributed object placement constraint file 206 described in the above section, a server that matches the placement constraint is selected at random. For example, the distributed object Ob001 in FIG.
S002 and S003), one of the three servers is randomly selected as a placement server.

For a distributed object for which no placement constraint is registered, the server description file 20 shown in FIG.
5. One server is randomly selected from the servers registered in 5.

Next, "1" is added to the variable N (step 1704). Next, the process returns to step 1702. Step 1
If the value of the variable N has exceeded M1 in step 702, the process moves to step 1705. In step 1705, “1” is substituted for the variable N. Next, the value of the variable N is compared with the variable M2 (step 1706). If the variable N is larger than M2, the process ends. If not, the following processing is performed.

In step 1707, in step 1703, the distributed objects are randomly placed on the server. In this case, however, the server description file 20 shown in FIG.
5 is assigned with priority on the server with the higher processing performance of the server registered in 5.

If the distributed object of the gene shown in FIG. 13 is registered in the location constraint file 206 of the distributed object in FIG. 8, allocation is performed according to the processing performance ratio in the server group that matches the placement constraint. . For example, in the example of the distributed object O001 in FIG. 8, the placement constraint is (S001, S002, S003), and the processing capacity of each server is 35, 64, 5 as shown in FIG.
8, the probability that the server S001 is selected is 35 /
(35 + 64 + 58), the probability that S002 is selected is 6
4 / (35 + 64 + 58), the server is selected with the probability of S003 being selected as 58 / (35 + 64 + 58).

For a distributed object having no placement restriction, a server is selected from all the servers registered in the server description file 205 in FIG. 7 with the probability of the processing performance ratio.

Next, 1 is added to the variable N (step 170).
8). Next, the process returns to step 1706 again.

Next, the intersection execution processing (step 1603) in FIG. 16 will be described in detail with reference to FIG. In the crossover process, first, the number of genes included in the population is multiplied by a predetermined crossover probability to obtain the number of crossings, and the value is substituted for a variable N (step 1801). Next, “1” is substituted for the variable I (step 1802).

The values of variable I and variable N are compared (step 1803). If I is greater than N, the process ends.

If not, the following processing is performed.
First, two genes are selected from a group of genes by random numbers, and the genes are substituted into variables G1 and G2 (step 1804).

Next, two intersections are set for two genes as shown in FIG. 15, and codes at each position are exchanged. That is, two intersections are determined by random numbers and substituted into variables C1 and C2 (step 1805). Then, the intersection C for the two genes of the variables G1 and G2
Crossover is performed at 1 and C2 to generate two new genes (step 1806). Next, "1" is added to the variable I (step 1807), and the routine proceeds to the next intersection processing.

Next, the details of the mutation execution process (step 1604) in the process flow of FIG. 16 will be described with reference to FIG. First, the number of occurrences of mutation is calculated by multiplying a preset mutation rate by the individual logarithm of the gene population, and is substituted into a variable N (step 1901).

Next, "1" is substituted for the variable I (step 1902). Next, the values of the variables I and N are compared (step 1903). If the variable I is larger than N, the process ends. If the variable I is N or less, the following processing is performed.

First, one gene is selected from the gene group by random numbers and assigned to a variable G (step 190).
4). Next, a mutation point in the gene is determined by a random number and assigned to a variable M (step 1905). Next, the code at the M-th position of the gene of the variable G is changed at random. In this process, if the distributed object corresponding to the M-th position has a constraint registered in the distributed object placement constraint file 206, a server is randomly selected within a range that matches the constraint. If there is no placement constraint, a random selection is made from all servers (step 190).
6). Next, "1" is added to the variable I (step 190).
7) Move to the next mutation process.

Next, the selection process in FIG. 16 will be described with reference to FIG. First, the number of genes added in the crossover process and the mutation process is added to the number of genes in the original population and substituted into a variable N (step 2001). Next, "1" is substituted for the variable I (step 2002).

The values of the variables I and N are compared (step 20).
03). If I is large, the process is terminated. When I is N or less, the following processing is performed.

First, according to the allocation of the distributed object specified by the I-th gene in the group to the server, the number of executions of each application calculated by the definition revision unit 209 and the application description file 20 shown in FIG.
2 and the execution characteristic (502) defined in the execution characteristic description file 203 of the distributed object shown in FIG.
The amount of server resources used by executing the distributed object in each server is calculated, and the server load is calculated by comparing it with the server processing capacity registered in the server description file shown in FIG. 7, and this is shown in FIG. Server load table 2
10 (step 2004).

Next, based on the load of each server registered in the server load table 210 and the information on the required communication time between each server stored in the network description file 204 shown in FIG. Obtain the response time of the application (step 200
5).

Next, the response time of each application is calculated based on the target response time and priority of each application registered in the application target description file 207 in FIG. 4B. Evaluate the assignment to
The evaluation here is the sum of the product of the delay from the target response time and the priority for all applications (step 2006). Next, 1 is added to the variable I, and the process proceeds to the next gene (step 2007).

FIG. 21 shows the next-generation gene population G of FIG.
It is a flowchart which shows the detail of the preparation process of t (step 1606). First, the number of genes in the original population is added to the number of genes added in the crossover process and the mutation process to obtain a variable N
(Step 2101). Next, 1 is substituted for the variable I (step 2102).

Next, the genes are sorted in ascending order of the evaluation values obtained by the flow of FIG. 20 (step 2103). Next, the values of the variables I and N are compared (step 2104). if,
If the value of the variable I is large, the process ends. If I is less than or equal to N, the following processing is performed.

From the rank table, an example of which is shown in FIG. 22, the number of duplicate copies corresponding to each rank of the evaluation value is obtained (step 2105). The I-th gene is duplicated and added to the next generation population (step 2106).
Next, 1 is added to the variable I, and the process proceeds to the next-ranked gene (step 2107).

As described above, the present embodiment detects the number of executions of a plurality of applications mounted on a client that sequentially calls a distributed object and executes a predetermined process in a predetermined monitoring period, and executes the detected execution of each application. Based on the information on the number of times, the information on the mutual call relationship between the distributed objects registered in the distributed object relationship description file 201, and the information on the distributed objects used by each application registered in the application description file 202, each The number of executions of the distributed object is calculated, and the information on the execution characteristics of each server registered in the server description file 205 and the information on the execution characteristics of each distributed object registered in the object characteristic description file 203 are compared with the predetermined time. Based of the information of the execution times of distributed objects, after estimating the load of each server, the estimation result of the load information and the network properties description file 204
The performance at the time of executing each application is estimated based on the information on the required call time between the distributed objects registered in the application, and the estimation result is displayed on the display device 215. Further, the information on the estimation result and the application target performance description file 207 are displayed. And a server for distributing objects that satisfies the target performance is determined on the basis of the information on the target performance for each application registered in.

As a result, even in a system in which distributed objects are executed in a complex environment, estimated information or decision information for appropriately leveling the load on a plurality of servers is provided to an administrator or the like. It can be presented and contribute to ensuring targeted client application responsiveness or targeted performance.

Further, a distributed object placement constraint description file 206 in which distributed objects restricted to be placed on a specific server computer or server group are further provided. When determining a distributed object placement server satisfying the target performance, Since the allocation server of the distributed object that satisfies the allocation constraint registered in the distributed object allocation constraint description file 206 is determined, even if there is the allocation restriction of the distributed object, the appropriateness of the distributed object is determined so as to satisfy the allocation constraint. Can contribute to the realization of the arrangement.

The present invention is not limited to the above-described embodiment, and can be implemented with various details modified in accordance with the actual network environment.
Further, information such as execution performance and characteristics is not limited to the information used in the above-described embodiment, and it is also possible to add more detailed information as needed to estimate an appropriate arrangement.

Further, when the number of times of execution of each distributed object per predetermined time is obtained as an execution log in an existing system, this can be used.

[0088]

As described above, according to the present invention, even in a system in which distributed objects are executed in a complex environment, estimation information for appropriately leveling loads on a plurality of servers is provided. Alternatively, the decision information can be presented to an administrator or the like, thereby contributing to ensuring the responsiveness or the target performance of the target client application.

Also, by deciding a distributed object placement server that satisfies the placement constraint registered in the distributed object placement constraint description file, even if there is a distributed object placement constraint, the distributed server can satisfy the placement constraint. This can contribute to realizing proper arrangement of objects.

By rearranging the distributed objects or arranging them at appropriate positions in accordance with such estimation results or determination results, it becomes possible to promote the efficiency of the computer network system using the distributed objects.

[Brief description of the drawings]

FIG. 1 is a system configuration diagram showing an embodiment of a computer network system to which the present invention has been applied.

FIG. 2 is a block diagram showing a functional configuration of a distributed object proper arrangement determining apparatus arranged in any one of server computers.

FIG. 3 is a diagram illustrating an example of registration data of a distributed object relation description file.

FIG. 4 is a diagram showing an example of registration data of an application description file and an application target description file.

FIG. 5 is a diagram showing an example of registration data of a distributed object execution characteristic description file.

FIG. 6 is a diagram showing an example of registration data of a network description file.

FIG. 7 is a diagram illustrating an example of registration data of a server description file.

FIG. 8 is a diagram illustrating an example of registration data of a distributed object placement constraint file.

FIG. 9 is a diagram illustrating an example of registration data of a server load table.

FIG. 10 is a diagram showing an example of registration data of an object arrangement table.

FIG. 11 is a diagram illustrating an example of registration data of a response table.

FIG. 12 is a flowchart illustrating an outline of processing of a distributed object optimization unit.

FIG. 13 is a diagram showing an example of gene expression.

FIG. 14 is an explanatory diagram of a genetic algorithm.

FIG. 15 is a conceptual diagram of the intersection processing.

FIG. 16 is a schematic flowchart of the suiting process using a genetic algorithm.

FIG. 17 is a flowchart showing a procedure for creating an initial gene population.

FIG. 18 is a flowchart illustrating a procedure of an intersection process.

FIG. 19 is a flowchart illustrating a procedure of a mutation process.

FIG. 20 is a flowchart illustrating a procedure of a selection process.

FIG. 21 is a flowchart showing a process of creating a next-generation gene population.

FIG. 22 is a diagram illustrating an example of a rank table.

[Description of sign]

101 server computer, 102 client computer, 103 router, 201 distributed object relationship description file, 202 application description file, 203 characteristic description file, 204 network description file, 205 server description file, 20
6: placement constraint description file, 207: application target description file, 208: monitoring device, 209: definition analysis unit, 210: server load table, 211: object placement table, 212: response table, 21
3. Optimizer, 215 ... Display device.

Claims (9)

[Claims] An application execution monitoring means for detecting the number of executions in a predetermined monitoring period of a plurality of application programs mounted on a client computer for sequentially calling a distributed object and executing a predetermined process, and is distributed to a plurality of server computers. A relationship description file in which registered mutual call relationships between distributed objects are registered; an application description file in which a plurality of application programs mounted on the client computer use which distributed object; and an execution performance of each server computer And a network call time relationship description file in which information about the call time between distributed objects in a plurality of server computers in the network is registered. And an object characteristic description file in which execution characteristics of each distributed object distributed in the plurality of server computers are registered. Each distributed object disposed in the plurality of server computers in the network is mounted on the client computer. A method for estimating a proper arrangement of a distributed object in a computer system that sequentially calls a plurality of application programs to execute a predetermined process, comprising: information on the number of times of execution of each application program detected by the application execution monitoring unit; and the relationship description. The information of the mutual call relationship between the distributed objects registered in the file and the information of the distributed objects used by each application program registered in the application description file. A first step of calculating the number of executions of each distributed object per predetermined time; information on execution characteristics of each server computer registered in the server description file and each distributed object registered in the object characteristic description file A second step of estimating the load on each server computer based on the information on the execution characteristics of the server and the information on the number of times of execution of the distributed object per predetermined time calculated in the first step; Based on the information on the required call time between distributed objects registered in the network characteristics description file, the performance at the time of executing each application program is estimated, and the estimation result is output as information for determining an appropriate arrangement of the distributed objects. And the third step. A method for estimating the proper arrangement of distributed objects to be featured. 2. An application execution monitoring means for detecting the number of times of execution of a plurality of application programs mounted on a client computer for sequentially calling a distributed object and executing a predetermined process during a predetermined monitoring period, and is distributed to a plurality of server computers. A relationship description file in which registered mutual call relationships between distributed objects are registered; an application description file in which a plurality of application programs mounted on the client computer use which distributed object; and an execution performance of each server computer And a network call time relationship description file in which information about the call time between distributed objects in a plurality of server computers in the network is registered. An object characteristic description file that registers execution characteristics of each distributed object distributed in the plurality of server computers, and an application target performance description file that registers information on target performance of each of the application programs. A method of determining an appropriate arrangement position of a distributed object in a computer system that executes predetermined processing by sequentially calling each distributed object arranged on a plurality of server computers in a network from a plurality of application programs installed on a client computer. Information on the number of executions of each application program detected by the application execution monitoring means, and information on mutual call relationships between distributed objects registered in the relationship description file. A first step of calculating the number of executions of each distributed object per a predetermined time based on information on distributed objects used by each application program registered in the application description file; and Based on the information on the execution characteristics of each server computer, the information on the execution characteristics of each distributed object registered in the object characteristic description file, and the information on the number of times of execution of the distributed object per predetermined time calculated in the first step, A second step of estimating the load of each server computer, and, based on the information on the load estimation result and the information on the required call time between the distributed objects registered in the network characteristic description file, the performance at the time of executing each application program. A distributed object satisfying the target performance based on information on the performance estimation result at the time of execution of each application program and information on the target performance for each application program registered in the application target performance description file. And a fourth step of deciding a placement server of the distributed object. 3. A distributed object placement constraint file in which a distributed object restricted to be placed on a specific server computer or server group is further registered, and the distributed object placement server satisfying the target performance in the fourth step. 3. The method according to claim 2, further comprising: determining a distribution object placement server satisfying the placement constraint registered in the distributed object placement constraint file. 4. When deciding a placement server of a distributed object satisfying the target performance in the fourth step, the priority information of the target performance of each application program additionally registered in the application target performance description file is used. The method according to claim 2 or 3, wherein an appropriate arrangement position is determined by giving priority to a distributed object having a higher priority. 5. When deciding a placement server of a distributed object that satisfies the target performance in the fourth step, a placement position of each distributed object with a probability proportional to an execution performance of each server computer registered in the server description file. 4. The method according to claim 2, further comprising: determining a proper arrangement of the distributed objects. 6. The method according to claim 2, wherein a proper arrangement position is determined by using a genetic algorithm. 7. Appropriate arrangement of distributed objects in a computer system that executes predetermined processing by sequentially calling each distributed object arranged on a plurality of server computers in a network from a plurality of application programs installed on a client computer. An application execution monitoring means for detecting the number of executions of a plurality of application programs mounted in the client computer during a predetermined monitoring period, and registering a mutual call relationship between distributed objects distributed on a plurality of server computers. Relationship description file, and an application description file that registers which distributed object is used by the plurality of application programs mounted on the client computer A server description file in which the execution performance of each server computer is registered; a network call time relation description file in which information about a call required time between distributed objects in a plurality of server computers in the network is registered; An object property description file in which the execution properties of each distributed object are registered, information on the number of executions of each application program detected by the application execution monitoring means, and a mutual call relationship between the distributed objects registered in the relation description file The number of executions of each distributed object per predetermined time is calculated based on the information of the distributed object used by each application program registered in the application description file. First processing means for outputting, information on execution characteristics of each server computer registered in the server description file, information on execution characteristics of each distributed object registered in the object characteristic description file, and the first processing means Second processing means for estimating the load on each server computer based on the information on the number of times of execution of the distributed object per predetermined time calculated in the above step; information on the load estimation result and the distribution registered in the network characteristic description file Third processing means for estimating the performance at the time of execution of each application program based on information on the required call time between objects, and outputting the estimation result as information for determining an appropriate arrangement of the distributed objects. Apparatus for estimating proper arrangement of distributed objects. 8. An appropriate arrangement of distributed objects in a computer system that executes predetermined processing by sequentially calling each distributed object arranged on a plurality of server computers in a network from a plurality of application programs installed on a client computer. An application execution monitoring means for detecting the number of executions of a plurality of application programs mounted on a client computer which sequentially calls a distributed object and executing a predetermined process during a predetermined monitoring period, and distributed to a plurality of server computers. A relationship description file in which a mutual call relationship between the arranged distributed objects is registered, and a plurality of application programs mounted on the client computer determine which of the distributed objects An application description file that registers whether to use it, a server description file that registers the execution performance of each server computer, and a network call time relationship description that registers information about the call time between distributed objects on multiple server computers in the network A file, an object property description file in which execution characteristics of each distributed object distributed in the plurality of server computers are registered, an application target performance description file in which information on a target performance for each of the application programs is registered, and the application execution The information on the number of executions of each application program detected by the monitoring means, the information on the mutual call relationship between the distributed objects registered in the relationship description file, and the First processing means for calculating the number of executions of each distributed object per predetermined time based on information on distributed objects used by each application program registered in the application description file; On the basis of the information on the execution characteristics of the server computer, the information on the execution characteristics of each distributed object registered in the object characteristics description file, and the information on the number of times of execution of the distributed object per predetermined time calculated by the first processing means, Second processing means for estimating the load of each server computer; and performance at the time of executing each application program, based on information on the result of the load estimation and information on the required call time between distributed objects registered in the network characteristics description file. Third to estimate Processing means, based on the information on the performance estimation result at the time of execution of each application program and the information on the target performance for each application program registered in the application target performance description file, a distributed object arrangement server that satisfies the target performance. And a fourth processing means for determining the position of the distributed object. 9. A computer network system for executing predetermined processing by sequentially calling each of distributed objects arranged on a plurality of server computers in a network from a plurality of application programs mounted on a client computer, comprising: Application execution monitoring means for detecting the number of executions during a predetermined monitoring period of a plurality of application programs mounted on a client computer which sequentially executes and executes a predetermined process; and a mutual call relationship between distributed objects distributed and arranged on a plurality of server computers And a description file that registers which distributed objects are used by the plurality of application programs mounted on the client computer. A description file, a server description file in which the execution performance of each server computer is registered, a network call time relation description file in which information about a required call time between distributed objects in a plurality of server computers in the network is registered, and the plurality of servers. An object characteristic description file in which execution characteristics of each distributed object distributed in the computer are registered; an application target performance description file in which information on target performance for each of the application programs is registered; and each application detected by the application execution monitoring means Information on the number of executions of the program, information on the mutual call relationship between the distributed objects registered in the relationship description file, and information registered in the application description file. First processing means for calculating the number of executions of each distributed object per predetermined time based on information on distributed objects used by each application program, and information on execution characteristics of each server computer registered in the server description file And estimating a load on each server computer based on information on execution characteristics of each distributed object registered in the object characteristic description file and information on the number of times of execution of the distributed object per predetermined time calculated by the first processing means. A second processing means for performing, and a third processing for estimating a performance at the time of executing each application program based on information on a load estimation result and information on a required call time between distributed objects registered in the network characteristic description file. Means and each application pro A fourth process of determining a distributed object allocation server that satisfies the target performance based on the information on the performance estimation result at the time of executing the program and the information on the target performance for each application program registered in the application target performance description file. And a means for deciding the arrangement position of the distributed objects constituted by the means.