JP2003208414A - Server with load distribution function and client - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent the inappropriate selection of load distribution destination caused by selecting a load distribution server on the basis of an overload condition of a proxy server which selects the load distribution destination, and a processing standby time, in a load distribution processing system. <P>SOLUTION: The server 2 holds the contents (concretely, execution date and time of processing, processing time of processing, data volume relating to processing, and a data transfer time in processing by every processing assigned to the server) in assigning the processing as a history 3 for every server 2 in the network including itself, the server 2 which can execute the processing in a short time is estimated in reference to the held history 3 when the processing request is given from a client 1 to select the server 2 as the load distribution server 2, and the execution of the processing is assigned to the selected server 2. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load distribution processing system in a network to which a plurality of servers are connected so that the processing in the server can be load-balanced to other servers in the network.

[0002]

2. Description of the Related Art In a network including a plurality of servers, when a processing request from a client to a server is distributed according to the load status of each server, conventionally,
The proxy server was supposed to select a server as the load distribution destination.

Specifically, in the conventional load balancing processing system, as shown in FIG. 2, the client A (1-
a), client B (1-b), client C (1
-C), server A (2-a), server B (2-b),
A proxy server 4 for monitoring the load distribution state of each server 2 is prepared between the server C (2-c), and when the proxy server 4 receives a processing request from the client 1 to the server 2, Select the server 2 as the load balancing destination,
The processing request was made to the selected server 2.

[0006] The proxy server 4 is used by each server 2
The CPU standby rate and the memory usage of the server are used to calculate the processing standby time, the load status of each server 2 is ascertained based on the calculated processing standby time, and the server 2 with the smaller load status is selected as the load distribution destination. It was like this.

[0005]

As described with reference to FIG. 2, in the conventional load balancing processing system, the proxy server 4 that monitors the load balancing state of each server 2 is
In general, when a processing request from the client 1 to the server 2 is received, the server 2 having a small load condition is selected and the processing request is sent to the selected server 2.

However, as shown in FIG. 3, when the number of clients 1 increases, processing requests from the clients 1 to the proxy servers 4 selected as load balancing destinations concentrate, so that the processing of the proxy server 4 is loaded. There was a problem that it became a bottleneck for the entire distributed processing system.

Further, as described with reference to FIG. 2, in the conventional load balancing processing system, the proxy server 4
However, when selecting the server 2 to be the load balancing destination, it is common to calculate the processing waiting time from the CPU usage rate and memory usage of each server 2 and perform the processing based on the calculated processing waiting time. It was target.

However, in actuality, when the proxy server 4 makes a processing request to the server 2 selected as the load distribution destination, the data required for executing the processing is also transferred. For this reason, when the amount of data required to execute the processing is large, there are some cases where the processing can be performed more efficiently by the processing system of its own than by performing the load distribution.

Therefore, an object of the present invention is to avoid the load balancing processing system by selecting the load balancing destination server based on the overload state of the proxy server that selects the load balancing destination or the waiting time of the processing. This is to prevent the selection of an appropriate load distribution destination.

[0010]

In order to achieve the above object, the present invention selects load balancing destinations from each server in order to prevent concentration of processing in a proxy server that selects load balancing destinations. Is trying to do.

The contents of each server outsourced to all servers in the network including itself (specifically, for each process outsourced to the server, the execution date and time of the process and the process). Processing time, the amount of data relating to the processing, and the data transfer time during the processing) are retained as a history, and when the processing is requested by the client connected to the network, the history information storage means Based on the stored history information, it estimates a server that will take a short time to execute the process, selects such a server as the load balancing server, and entrusts the selected server to execute the process. I am trying.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) FIG. 1 is a diagram showing a configuration example of a load distribution processing system according to the first embodiment.

As shown in FIG. 1, the load balancing processing system according to the first embodiment is a server A (2-a), a server B (2-b), and a server C (2) connected by a network.
-C), and the client 1 requesting processing to each server 2
And is configured. Although only one client 1 is shown in FIG. 1, it is general that two or more clients 1 actually exist.

Each server 2 has a history A (3-a) and a history B.
(3-b), the history C (3-c) is held, and the history contents when the processing was previously performed by the own server 2 and the processing history is previously entrusted to another server 2 in these history 3 The history contents at the time of doing are stored.

In the first embodiment, as shown in FIG. 1, a server (for example, server 2-a) that receives a processing request from the client 1 refers to its own history A (3-a), The best server to execute the process (for example, server C
(2-c)) is selected and the processing is outsourced to the server C (3-c).

The operation when the server A (2-a) receives a processing request from the client 1 will be described below.

The characteristics of each server 2 are as follows: Server A (2-a) ... Long processing waiting time, normal processing speed Server B (2-b) ... Short processing waiting time, Fast processing speed, Server A Data transfer time with (2-a) Ordinary server C (2-c) ... Features that the standby time for processing is short, processing speed is slow, and data transfer time with server A (2-a) is short. The time when the server 1 (2-a) receives the processing request from the client 1 is August 8
It is assumed that the day is 00:00:00.

4 to 6 show the contents of the history A (3-a) held by the server A (2-a). FIG. 4 shows the server A
(2-a) shows the history content of itself, FIG. 5 shows the history content of the server B (2-b), and FIG. 6 shows the history content of the server C (2-c).

As shown in FIG. 4, server A (2-a)
Is the history content of its own server A (2-a), time A ... the time taken to obtain the processing result, time B ... the time the processing was performed time C ... the waiting time of the processing Time D: Data transfer time of data to be transferred to the consignee Time E: Data transfer time of data transferred from the consignee to the consignor Data amount A ... Data amount of data required to execute the process Data amount B ... Process completion Holds the amount of data that will be generated later.

As shown in FIGS. 5 and 6, the server A (2-a) is a server B (2-b) and the server C is a server C (2-b).
As the history contents regarding (2-c), time A ... time was entrusted to another server to execute the process, and time taken until the processing result was received time B ... process was actually executed by the entrustee. Time Time C ... Waiting time for processing at the consignee Time D ... Data transfer time of data to be transferred to the consignee Time E ... Data transfer time of data transferred from the consignee to the consignee Source A ... Transfer to the consignee Data amount of data to be processed Data amount B ... Holds the data amount of data transferred from the entrustee to the entrustor.

Regarding the method of deriving the history contents, when the server A (2-a) receives a processing request from the client 1 and the server A (2-a) entrusts the processing to the server B (2-b). In the following, the history content held in the server A (2-a) will be described as an example with reference to FIG. 7.

As shown in FIG. 7, the server A (2-a)
First, when there is a processing request from client 1,
The time a1 at that time is held, and then the server B (2-
When the process is entrusted to b), the time a2 at that time is held, and the data amount of data necessary for executing the process, that is, the transfer data amount A1 to the server B (2-b) is held.

The server B (2-b) is connected to the server A (2-
When the process is entrusted from a), the time b1 at that time is held, and subsequently, when the process is executed, the time b2 at the start of the process and the time b3 at the end of the process are held. After finishing the entrusted processing, the server B (2-b) returns the processing result, that is, the data generated after the processing is completed, to the server A (2-a). At that time, While holding the time b4, the amount of data generated after the process is completed, that is, the server A (2-
The transfer data amount B1 to a) is held. Here, at the time of returning the processing result, the server B (2-
The time b1, the time b2, the time b3, the time b4, and the data amount B1 are also transmitted from b) to the server A (2-a).

The server A (2-a) is connected to the server B (2-
When the processing result is received from b), the time a3 at that time is held.

In this way, from the time of holding / acquiring, the server A (2-a) determines that time A = time a3-time a2 time B = time b3-time b2 time C = time b2-time b1 time D. = Time b1-Time a2 Time E = Time a3-Time b4 Data amount A = Data amount A1 Data amount B = Data amount B1 can be obtained and stored in the history content of the server B (2-b).

The server A (2-a) and the server B (2
It goes without saying that it is premised that the time is the same as that of -b). Further, the time a1 initially held by the server A (2-a) is the time b when the processing is executed by itself.
It is treated in the same manner as 1.

Now, in the first embodiment, the server A
(2-a), when receiving a processing request from the client 1, selects the server 2 out of all the servers 2 including itself that has a shorter time required to execute the processing requested by the client 1. It is assumed that such a server 2 is selected as the entrusted server 2 and the entrusted server 2 is entrusted with the processing.

This estimation is performed by calculating the value of the function f (X) described later based on the data amount of the processing request received from the client 1 and the contents of the history A (3-a).

That is, the server A (2-a) first
The coefficient used in the function f (X) is set. here,
The coefficient setting of the function f (X) will be described.

The server A (2-a) has a history A (3-a).
The coefficient ai, the coefficient bi, the coefficient ci, the coefficient di, and the coefficient ei are obtained based on the contents of. ai = time D / data amount A bi = time E / data amount B ci = time B / data amount A di = time C ei = data amount B / data amount A

Subsequently, the server A (2-a) determines that the history A
By referring to the date and time in (3-a), a record (No. in history A (3-a)) close to the date and time when the outsourcing server 2 is selected.
The data of unit) is regarded as more useful as a record for selecting an outsourcee, and each record in the history A (3-a) is weighted.

In the first embodiment, for example, importance 1 ... History importance within 1 hour 2 ... History importance within 1 day 3 ... History importance within 3 days 4 ... History exceeding 3 days 4 There are two importance levels, and the smaller the number, the higher the importance.

Then, the server A (2-a), as shown in FIG. 8, uses these importance levels and the coefficient ai, coefficient bi, coefficient ci, coefficient di, and coefficient ei previously obtained. The coefficient A, the coefficient B, the coefficient C, the coefficient D, and the coefficient E are obtained.

Then, the server A (2-a) uses the data amount of the data received from the client 1 for the processing as a variable X (kByte), and this variable X and the coefficient obtained as shown in FIG. A function f (X) shown below is calculated from A, coefficient B, coefficient C, coefficient D, and coefficient E.

F (X) = (A + C) × X + B × E × X + D The calculation for obtaining the f (X) value is performed for every record in the history record A (3-a).

Since the f (X) value serves as a measure of the time required to execute the process, the server A (2-a) entrusts the server 2 corresponding to the record with the small f (X) value. It is selected as the destination server 2.

FIG. 9 is a graph showing the relationship between the X value and the f (X) value of each server 2.

From this graph, the entrusted server 2 changes from server C (2-c) to server B (2-) according to the size of the amount of data required for processing received from the client 1.
It can be seen that the selection is made in the order of b)-> server A (2-a).

In the first embodiment, the function f
(X) and the weighting method are examples, and the user can freely set / change them without impairing the gist of the present invention.

(Second Embodiment)

In the first embodiment, when selecting the outsourcing server 2, weighting is applied to each record in the history 3 based on the date and time in the history 3, but not only the date and time. The weighting may be performed based on the data amount A in the history 3.

A case where this is done will be described below as a second embodiment.

In the second embodiment, the weighting based on the data amount A means that the record closer to the data amount of the data received from the client 1 for processing is considered to be more useful as the record selected by the outsourcing contractor. is doing.

In the second embodiment, for example, the history of importance 1 ... Within 1 hour, and the difference from the data amount A within 50 kByte history importance 2 ... History within 1 day and data amount A History importance within a difference of 50 kByte 2 ... History within 1 hour, and history difference within a data amount A of more than 50 kByte and within 300 kByte History importance 3 ... History within 3 days and data amount A Difference within 50 kBytes history history within 3 hours ... History with less than 1 hour and data amount A greater than 300 kByte within 1000 kByte history importance 4 ... History within 1 day and data amount A History importance level within a difference of more than 50 kBytes and within 300 kBytes 5 ... History exceeding 3 days, and history importance level within a difference of 50 kBytes within 5 kBytes 5 ... 1 hour or less History within 6 days, history history with a difference of more than 1000 kByte from data amount A ... History within 3 days, history history with a difference from data amount A of more than 50 kByte and less than 300 kByte 6 within 1 day History history importance level 7 that the difference from the data amount A is more than 300 kBytes and within 1000 kBytes ... History history that exceeds 3 days and the history importance level 7 that the difference from the data amount A is greater than 50 kBytes and within 300 kBytes 7 ... 3 History importance within 7 days and history difference with data amount A larger than 300 kBytes and within 1000 kByte 7 ... History history within 1 day and history importance with difference with data amount A greater than 1000 kByte 8 ... 3 days History and the history weight that the difference from the data amount A is more than 300 kByte and less than 1000 kByte. 8: History within 3 days, history importance with a difference of more than 1000 kByte from the data amount A 9: History importance of more than 3 days, and history with a difference of more than 1000 kByte from the data amount A Is set, and the smaller the importance number, the higher the importance.

The equations for obtaining the coefficient ai, the coefficient bi, the coefficient ci, the coefficient di, and the coefficient ei, and the coefficient A, the coefficient B,
The equations for obtaining the coefficient C, the coefficient D, and the coefficient E are the same as those in the first embodiment, but regarding W (xi), for example, FIG.
As shown in 0.

FIG. 11 shows the value of X and f (X) of each server 2.
It is a graph showing the relationship with a value.

From this graph, according to the size of the amount of data received from the client 1 for processing, the entrusted server 2 changes from server C (2-c) to server B (2-
It can be seen that the selection is made in the order of b)-> server A (2-a).

In the second embodiment, as in the first embodiment, the function f (X) and the weighting method are
This is merely an example, and the user can freely set / change it without impairing the gist of the present invention.

By the way, when the f (X) value is calculated in both the first embodiment and the second embodiment, a record having a small f (X) value corresponds to a plurality of servers 2. If it is estimated that there is no difference in the time required to execute the process regardless of which server 2 is selected as the entrusted server 2, the server 2 with a small history 3 (the number of records) is entrusted. It can be selected as the destination server 2. This makes it possible to enhance the history 3 held by the server 2 and improve the efficiency of the load balancing system.

In each of the first and second embodiments, the entrusted server 2 is selected based on the contents of the history 3, but all the servers including the self server 2 are selected. The load state of No. 2 may be collected, and the entrusted server 2 may be selected based on the collected load state and the history 3.

In each of the first and second embodiments, the entrusting source server 2 selects the entrusting destination server 2 and entrusts the selected entrusting destination server 2 with the processing, and then the entrusting destination server 2. The server 2 may estimate the time required for the processing and return the estimation result to the entrusting source server 2. By doing this, the consignment source server 2
Can re-election the entrusted server 2 depending on the returned estimation result.

Further, in both the first and second embodiments, for example, as shown in FIG.
When adding a new server D (2-d), the history A (3-a) of the existing server A (2-a) and the server B (2-
B) history B (3-b), server C (2-c) history C
The history content of the server D (2-d) is added to each of (3-c), and the history D of the server D (2-d) is added.
In (3-d), the history content of the server A (2-a), the history content of the server B (2-B), and the server C
The history content for (2-c) and the history content for server D (2-d) are retained.

Here, it is preferable that the history content of the server D (2-d) to be added to the history 3 of the existing server 2 be an average default value. Also,
History D (3-d) to be retained in the new server D (2-d)
May acquire the history 3 of the existing server 2 at the time of addition, but in reality it is difficult to make the history 3 exactly the same, so an average default value may be used.

Further, in any of the first and second embodiments, not only when the processing request is received from the client 1, but also when the processing is entrusted from another server 2, If the entrusted server 2 is selected by referring to its own history 3 and the most suitable entrusted server 2 is selected, the server 2
You may entrust to. For example, in FIG. 13, the server A (2-a) that has received the processing request from the client 1
When the server C (2-c) entrusts the processing to the server C (2-c), the server C (2-c) refers to its own history C (3-c),
When the server B (2-b) is selected as the entrustment destination server 2, an example in which the processing entrusted by the server A (2-a) is entrusted to the server B (2-b) is shown.

In this case, the number of times of entrustment is added to each entrustment, and the upper limit of the number of times of entrustment is set so that unnecessary processing is not performed. It is preferable. For example, in the example of FIG. 13, when the upper limit of the number of times of entrustment is set to 2, server A (2-a) → server C (2-c) → server B (2
-B) and the processing is outsourced, the server B (2-b)
In the above, since the number of times of entrustment is already 2, the server B (2-b) does not select the entrustee server 2,
Make sure to perform the entrusted processing.

Further, in both the first and second embodiments, the server 2 selects two or more servers 2 as the entrustment destination servers 2 and performs processing on these entrustment destination servers 2. You may divide and entrust. For example, in FIG. 14, the server A (2-a) that has received the processing request from the client 1 divides the processing 5-a entrusted by the client 1 into two, processing 5-b and processing 5-c, The server B (2-b) and the server C (2-c) are selected as the entrusted server 2, the process 5-b is entrusted to the server B (2-b), and the process C is performed to the server C (2-c). The example which entrusted -c is shown.

Further, depending on the processing capacity of the client 1, the client 1 may retain the history of each server 2. First, the client 1 itself refers to these histories and determines which server 2 should request processing. After the selection, the processing request may be issued to the selected server 2. For example, in FIG. 15, the client 1 has a history A (6-a) about the server A (2-a) and a server B.
History B (6-b) for (2-b), server C (2
An example is shown in which the server C (2-c) is selected as the processing request destination and a processing request is made by referring to the history C (6-c) for -c).

[0059]

As described above, according to the present invention, when load balancing is performed in an environment in which a plurality of servers exist in a network, processing concentration in a proxy server that determines a load balancing destination. In addition to solving the problem of processing degradation due to, there is a cost advantage due to not setting a proxy server.

Further, by considering not only the waiting time of the processing at the load distribution destination but also the data transfer rate, it is possible to select a more appropriate load distribution destination, and it is expected that the processing efficiency at the server will be improved. .

[Brief description of drawings]

FIG. 1 is a configuration diagram of a load distribution processing system according to a first embodiment.

FIG. 2 is a configuration diagram of a conventional load balancing processing system.

FIG. 3 is a configuration diagram of a conventional load balancing processing system.

FIG. 4 is an explanatory diagram showing an example of history contents regarding server A held by server A.

FIG. 5 is an explanatory diagram showing an example of history contents regarding server B held by server A.

FIG. 6 is an explanatory diagram showing an example of history contents regarding a server C held by the server A.

FIG. 7 is an explanatory diagram showing a method of deriving history contents.

FIG. 8 is an explanatory diagram showing an equation for obtaining a coefficient used when selecting a load distribution destination server.

FIG. 9 is an explanatory diagram showing a graph showing the relationship between the data amount of data required to execute the process and the time required for the process.

FIG. 10 is an explanatory diagram showing an equation for obtaining a coefficient used when selecting a load distribution destination server.

FIG. 11 is an explanatory diagram showing a graph showing the relationship between the amount of data required to execute the process and the time required for the process.

FIG. 12 is an explanatory diagram when a new server is installed.

FIG. 13 is an explanatory diagram of a case where subcontracting is performed.

FIG. 14 is an explanatory diagram of a case where divisional consignment of processing is performed.

FIG. 15 is an explanatory diagram of a case where a client retains a history.

[Explanation of symbols]

1: Client 2: Server 3: History 4: Proxy server 5: Data 6: History

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Takashi Kagei             216 Totsuka Town, Totsuka Ward, Yokohama City, Kanagawa Prefecture             Ceremony Company Hitachi Ltd. Communication Division (72) Inventor Haruhiko Araki             216 Totsuka Town, Totsuka Ward, Yokohama City, Kanagawa Prefecture             Ceremony Company Hitachi Ltd. Communication Division (72) Inventor Koji Suzuki             216 Totsuka Town, Totsuka Ward, Yokohama City, Kanagawa Prefecture             Ceremony Company Hitachi Ltd. Communication Division F term (reference) 5B045 BB28 BB42 GG02 GG05                 5B089 GA11 JA11 KA06 KC23 MA03                       MC03

Claims (5)

[Claims] 1. A server that is connected to a network together with a plurality of servers, and for each of all the servers including itself, the execution date and time of the processing, the processing time of the processing, A history information storage unit that holds at least the data amount relating to the processing and the data transfer time during the processing as history information, and the history information when the processing is requested by the client connected to the network. A load balancing destination server selecting means that estimates a server that will take a short time to execute the processing based on history information stored in the storage means and selects the server as a load balancing destination server; Load balancing, characterized in that the server selected by the selecting means has a consigning means for entrusting the execution of the processing requested by the client. Ability with the server. 2. The server with a load balancing function according to claim 1, wherein the load balancing destination server selecting means determines each history information according to the date and time in the history information held by the history information storing means. After selecting the load balancing destination server, or the amount of data in the history information held by the history information storage unit and the data of the data necessary for executing the processing requested by the client. A server with a load balancing function, characterized in that each history information is weighted according to its closeness to the amount and then a load balancing destination server is selected. 3. The server with a load balancing function according to claim 1, wherein the load balancing destination server selecting means loads a server with a small amount of history information when there are a plurality of load balancing destination server candidates. A server with a load balancing function that is selected as a distribution destination server. 4. The server with a load balancing function according to claim 1, 2, or 3, wherein the load balancing destination server selecting means is configured to execute the history information even when a process is entrusted by another server. Based on the history information stored in the storage means,
A server with a load balancing function, which estimates a server that will take a short time to execute the process and selects it as a load balancing destination server. 5. A client which is connected to a server with a load balancing function according to claim 1, 2, 3 or 4 on a network, and which is a process requested to the server for each of all the servers connected to itself. A history information storage unit that holds, as history information, at least the execution date and time of the process, the processing time of the process, the amount of data regarding the process, and the data transfer time at the time of each process, for each server. When a process that should be requested to be executed occurs, the server requesting the process is inferred based on the history information stored in the history information storage means, and a server that requires a short time to execute the process is estimated. And a process requesting unit that requests the server selected by the process requesting destination server selecting unit to execute a process. Client.