JP2006139578A - Load distribution system, server system, and method therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a scale-out type server system having a load distribution function capable of instantaneously responding to increase and decrease of transactions and a money charge function thereof. <P>SOLUTION: Each server forming a server group 2 is maintained in a state for processing transactions, and a load distribution system 1 is connected with a front end of the server group 2. The load distribution system 1 has a function (server status management section 7) for receiving instructions whether or not each server is set to be a transmission destination of the transactions, and a function (load distribution section 6) for transmitting a transaction to the server which is a transmission destination of the transaction. The load distribution system 1 transmits the transaction transmitted from a client 3 to the server group 2 side. The load distribution system 1 also comprises a server usage time measuring section 8 and a server money charge section 9, and executes meter-rate based money charge processing by totalling the amount of time specified for a transmission destination of the transaction. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a load distribution apparatus, a server system, and a method for performing load distribution of a server system, and more particularly to a load distribution apparatus, a server system, and a method for performing scale out in order to improve the performance of the entire system. .

  Conventionally, as a measure to improve the transaction processing capacity in a server system consisting of multiple servers, a load balancer has been placed on the front end of the server system, and transactions are sent preferentially to servers under light load. However, a technique for performing processing is known.

  For example, in Japanese Patent Application Laid-Open No. 2003-296289, a means for measuring CPU usage time is provided in a server that accepts and processes a transaction, the load balancer obtains the measurement result via a network, and based on the measurement result. A method of determining a server with a light server load state and transmitting a transaction preferentially to the determined server is introduced.

  However, in the system configuration shown in the publication, it is difficult to improve the overall processing capacity of the system according to the amount of transactions, and when a transaction exceeding the total processing capacity of the prepared servers is transmitted. , There is a problem that can not be addressed.

Therefore, a server system equipped with a scale-out type load balancing function that can improve the overall system performance by distributing the load by increasing or decreasing the number of servers related to the processing of interest is also known. Yes. For example, Japanese Patent Application Laid-Open No. 2002-259354 introduces a method of increasing or decreasing the number of servers that distribute content according to the number of distribution requests to specific content.
JP 2003-296289 A JP 2002-259354 A

  However, in the server system equipped with the conventional scale-out type load balancing function described above, the amount of transactions to be processed increases rapidly, increasing the number of load distribution destination servers, and quickly improving the processing capacity of the entire system. There is a problem that the responsiveness in the case of wanting to be made is lacking and the processing capacity of the entire system cannot be instantly optimized. For example, in the case of an operation for increasing the number of servers, the time from when the server is turned on or when the service is changed until the server actually reaches a state where it can process a transaction, and the load balancer recognizes the change in the server state. The current situation is that it takes a long time. Similarly, even when the number of servers is reduced due to a decrease in the number of transactions, it takes a long time for the load balancer to recognize a change in the server status, and the processing capacity of the entire system is instantaneously determined. There was a problem that it could not be optimized.

  In addition, in a server system equipped with this type of scale-out load balancing function, the number of servers increases and decreases, and the performance is improved. Therefore, a reasonable billing method that matches the resources actually used is established. Not.

  The present invention has been made in view of the above-described circumstances, and its object is to provide a load with excellent responsiveness that can be suitably used for a server system having a scale-out type load distribution function. A distribution apparatus, a server system including the distribution apparatus, and a load distribution method are provided.

  It is another object of the present invention to provide a load balancer capable of performing a reasonable charge calculation corresponding to the resource used, a server system including the load balancer, and a billing method that can be implemented using the load balancer. .

  A load distribution apparatus according to a first aspect of the present invention is an apparatus that is located at the front end of a plurality of server groups and that distributes and transmits transactions from clients, and includes a server state management unit and a load distribution unit. I have. The server state management unit manages whether or not each server is a transaction transmission destination. The load distribution unit transmits a transaction to a server that is a transaction transmission destination.

  The load distribution apparatus according to the second aspect of the present invention includes a server usage time measurement unit and a server billing unit, and provides a billing processing function. The server usage time measurement unit measures the time set as a transaction transmission destination for each server, and the server billing unit performs billing processing based on server usage time data output from the server usage time measurement unit and charges Is output.

  According to the present invention, it is possible to instantaneously optimize the processing capability of the entire server system. The reason for this is that when implementing scale-out to increase or decrease the number of load distribution destination servers, the number of servers that are the transaction destinations is increased or decreased while keeping the servers in a transaction processable state, and transactions are transmitted based on this. The configuration is adopted.

  According to the present invention, it is possible to construct a rational billing system that is suitably applied to a server system that increases or decreases the number of load distribution destination servers. The reason for this is that paying attention to the time at which the transaction is sent and adopting a configuration that enables pay-per-use billing processing based on this time.

  Subsequently, a preferred embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention. Referring to FIG. 1, a load distribution apparatus 1 according to the present embodiment is connected to a server group 2 and a large number of clients 3 for performing scale-out type load distribution via networks 4 and 5. The load distribution apparatus 1 includes at least a load distribution unit 6, a server state management unit 7, a server usage time measurement unit 8, and a server billing unit 9. These are configured by the hardware resources of the computer constituting the load balancer 1 and a program that drives the hardware resources.

  The server state management unit 7 is connected to the load distribution unit 6 and the server usage time measurement unit 8. The system administrator uses the administrator terminal or the like to input data that gives an instruction for accepting / not accepting transactions for the servers included in the server group 2 to the server state management unit 7. Each server can be designated by a server ID described later. Then, data about whether to accept or reject transactions for the servers included in the server group 2 is transmitted to the load distribution unit 6 and the server usage time measurement unit 8.

  The load distribution unit 6 includes at least communication control units 10 and 11, transaction buffers 12 and 13, a transaction transmission destination address generation unit 14, and a server state management table 15. The load distribution unit 6 can be connected to an external network including the networks 4 and 5 via the communication control units 10 and 11.

  FIG. 2 is a diagram for explaining the configuration of the server state management table 15 stored in a storage device (not shown) of the load distribution apparatus 1. Referring to FIG. 2, the server state management table 15 includes entries including at least a server ID field 31, a transaction transmission target flag field 32, and a server private IP address field 33 (RFU is reserved for Future Use). Abbreviation). Each entry can be searched by a server ID given to specify a server of the server group 2.

  Referring to FIG. 1 again, the server usage time measurement unit 8 includes at least a timer 16, a server usage time management table update unit 17, and a server usage time management table 18. The server billing unit 9, the server status management unit 7 Connected with.

  FIG. 3 is a diagram for explaining the configuration of the server usage time management table 18 stored in a storage device (not shown) of the load distribution apparatus 1. Referring to FIG. 3, the server usage time management table 18 includes at least a server ID field 41, a transaction transmission target flag field 42, a transaction transmission destination state start time field 43, a transaction transmission destination state end time field 44, and an accumulated time field 45. Consists of containing entries. Each entry can be searched by server ID.

  Referring again to FIG. 1, the server billing unit 9 includes at least a server billing management table 19, a server billing management table update unit 20, a timer 21, and a billing information providing unit 22, and is connected to the server usage time measuring unit 8. Has been.

  FIG. 4 is a diagram for explaining the configuration of the server charging management table 19 stored in a storage device (not shown) of the load distribution apparatus 1. Referring to FIG. 4, the server billing management table 19 includes entries including at least a server ID field 51, a date field 52, an accumulated time field 53, and an accumulated fee field 54. Each entry can be searched by server ID.

  Next, operations of the load distribution unit 6, the server usage time measurement unit 8, and the server billing unit 9 of the load distribution apparatus 1 having the above-described configuration will be sequentially described with reference to the drawings.

  FIG. 5 is a flowchart for explaining the operation of the load distribution unit 6. In the following description, it is assumed that 0 or 1 is set in the transaction transmission target flag field 32 of the entry of the server state management table 15 by the input data from the server state management unit 7. If the transaction transmission target flag set by the input data is 0, it indicates that the transaction is not transmitted to the server. Conversely, if the transaction transmission target flag field 32 is 1, the transaction is not transmitted to the server. Indicates that it will be sent.

  Referring to FIG. 5, first, when a transaction is received from the transaction buffer 12 on the client 3 side, the transaction destination address generation unit 14 of the load distribution unit 6 masks a part of the destination IP address of the transaction (step 102).

  Subsequently, the transaction transmission destination address generation unit 14 of the load distribution unit 6 reads the entry from the server state management table 15 and checks the transaction transmission target flag field 32 (step 103). Here, when the transaction transmission target flag is 0 (Yes in step 104), the server of the entry is not a transaction transmission target, so the transaction transmission destination address generation unit 14 determines the next entry from the server state management table 15. Is read.

  On the other hand, if the transaction transmission target flag is 1 in step 104 (No in step 104), the transaction destination address generation unit 14 is masked in step 102 because the server of the entry is a transaction transmission target. For the destination IP address of the transaction, the value of the private IP address field 33 of the server of the entry is set (step 105).

  Finally, the transaction transmission destination address generation unit 14 transmits a transaction to the transaction buffer 13 together with the generated transmission destination IP address (step 106).

  FIG. 6 is a flowchart for explaining the operation of the server usage time measuring unit 8. As will be described in detail later, 0 or 1 is set in the transaction transmission target flag field 42 of the entry of the server usage time management table 18 according to the input data from the server state management unit 7. When the transaction transmission target flag becomes 0, the transaction transmission destination state end time field 44 and the accumulated time field 45 become valid (can be updated) in order to calculate the period of the transaction transmission target. On the other hand, when the transaction transmission target flag is set to 1 by the input data, the transaction transmission destination state start time field 43 becomes valid (can be updated) so as to record the start point of the period that was the transaction transmission target.

  Referring to FIG. 6, the server usage time management table update unit 17 of the server usage time measurement unit 8 monitors inputs from the normal server state management unit 7 or the timer 16 (steps 122, 123, and 124).

  In step 124, upon receiving a notification from the server state management unit 7 that the server is designated as the transaction transmission destination, the server usage time management table update unit 17 uses the server ID from the server usage time management table 18. The corresponding entry is read (step 125), 1 is set in the transaction transmission target flag field 42 (step 126), and the time data obtained from the timer 16 is set in the transaction transmission destination state start time field 43 (step 127). Next, the server usage time management table update unit 17 writes the updated entry back to the server usage time management table 18 (step 128).

  On the other hand, upon receiving a notification from the server status management unit 7 that the server is to be removed from the transaction transmission destination in step 123, the server usage time management table update unit 17 uses the server ID to generate a server usage time management table. 18 reads the corresponding entry (step 129), sets 0 in the transaction transmission target flag field 42 (step 130), and sets the time data obtained from the timer 16 in the transaction transmission destination state end time field 44 (step 131). . Next, the server usage time management table update unit 17 calculates the difference between the transaction transmission destination state end time and the value of the transaction transmission destination state start time field 43 to obtain the time that has been the current transaction transmission destination, and the accumulated time Add to the value of field 45 (step 132). Finally, the server usage time management table update unit 17 writes the updated entry back to the server usage time management table 18 (step 128).

  In step 122, upon receiving a “reference time arrival signal” indicating that the expiration date and time of the predetermined counting period has arrived from the timer 16, the server usage time management table updating unit 17 sends the server charging unit 9 It is notified that the reference time for accounting has been reached, and the data for all entries in the server usage time management table 18 is read (see step 133 and FIG. 7). Then, the server usage time management table update unit 17 sets a reference time as a new starting point in the “transaction transmission destination state start time field 43” in the entry of the server usage time management table 18 (step 134), and accumulates it. Zero is set in the time field 45 (step 135).

  FIG. 7 is a flowchart for explaining the operation of the server billing unit 9. Referring to FIG. 7, the server billing management table update unit 20 of the server billing unit 9 monitors the input from the server usage time management table update unit 17 (step 142).

  When the server usage time management table update unit 17 receives a notification that the reference time has been reached from the server usage time management table update unit 17, the server billing management table update unit 20 stores the server usage time management table 18 in the server usage time measurement unit 8. The entries are read one by one (step 143).

  The server billing management table updating unit 20 performs charge calculation by multiplying the value of the accumulated time field 45 by, for example, a predetermined unit price for each read entry (step 144).

  Next, the server billing management table updating unit 20 sets the values of the server ID field 41 and the accumulated time field 45 in the read entry in the server ID field 51 and the accumulated time field 53 of the entry (steps 145 and 146).

  Next, the server billing management table updating unit 20 sets the date data received from the timer 21 in the date field 52 of the entry (step 147).

  Further, the server billing management table updating unit 20 sets the result of the charge calculation in step 144 in the accumulated charge field 54 of the entry (step 148).

  The server billing management table update unit 20 writes the entry generated in each of the above steps into the server billing management table 19 (step 149).

  The server billing management table update unit 20 performs the above series of processing for all server IDs (steps 143 to 150).

  The data of the server billing management table 19 in the server billing unit 9 obtained in this way is specified by the billing information providing unit 22 at the last day (reference time) of the server and billing calculation period, and the external billing terminal It is possible to view and output from etc.

  As described above, the server system according to the present embodiment performs transaction load distribution by scale-out, but the number of load distribution destination servers can be instantaneously increased or decreased by notification from the server state management unit 7. It is configured as follows. Therefore, the processing capacity of the entire system can be flexibly and promptly adapted according to the amount of transactions and other load conditions. As a result, it is also possible to reduce the need for securing an extra margin and the like, and to optimize the operating cost of the server system.

  Further, in the server system of the present embodiment, the usage fee can be calculated by regarding the time that is the load distribution target of each server as the usage time of the server. The conventional billing based on the connection time to the server and the usage time of the CPU does not provide a rational billing system when performing transaction load balancing by scale-out, but in the present embodiment, the load balancing configuration described above is used. This makes it possible to calculate reasonable pay-per-use charges.

  Further, in the server system of the present embodiment, the time when each server is a transaction destination is measured and held for each server and used for the above billing process. This can be useful when reviewing plans or considering future capital investment plans.

  Although the embodiments of the present invention have been described above, the above embodiments are merely examples that are easy to explain and that are considered preferable in practicing the present invention. For example, in the above-described embodiment, two timers 16 and 21 are prepared, but it is also possible to configure them with one timer. Also, the entire configuration including the client 3, the load balancer 1, and the server group 2 is merely an example of a configuration that is easy to understand visually, and can be applied to various modified configurations.

It is a block diagram showing the whole structure of one embodiment of this invention. It is a figure for demonstrating the structure of the server state management table 15 of one embodiment of this invention. It is a figure for demonstrating the structure of the server use time management table 18 of one embodiment of this invention. It is a figure for demonstrating the structure of the server charge management table 19 of one embodiment of this invention. It is a flowchart for demonstrating operation | movement of the load distribution part 6 of one embodiment of this invention. It is a flowchart for demonstrating operation | movement of the server usage time measurement part 8 of one embodiment of this invention. It is a flowchart for demonstrating operation | movement of the server accounting part 9 of one embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Load distribution apparatus 2 Server group 3 Client 4, 5 Network 6 Load distribution part 7 Server state management part 8 Server use time measurement part 9 Server billing part 10, 11 Communication control part 12, 13 Transaction buffer 14 Transaction transmission destination address generation part DESCRIPTION OF SYMBOLS 15 Server state management table 16, 21 Timer 17 Server use time management table update part 18 Server use time management table 19 Server charge management table 20 Server charge management table update part 22 Charge information provision part 31, 41, 51 Server ID field 32, 42 Transaction transmission target flag field 33 Private IP address field of server 43 Transaction transmission destination state start time field 44 Transaction transmission destination state end time field 45, 5 3 Cumulative time field 52 Date field 54 Cumulative charge field

Claims (9)

A load balancer that is located at the front end of a plurality of servers in a state where each server can process transactions, and that distributes and transmits transactions from clients,
A server state management unit for managing whether to send a transaction to each server;
A load distribution unit that transmits a transaction to a server that is a transaction destination,
A load balancer. The load distribution apparatus according to claim 1, further comprising:
A server usage time measurement unit that measures the time that is a transaction transmission destination for each server;
A load balancer. The load distribution apparatus according to claim 2,
The server usage time measurement unit
A server usage time management table including at least a server ID, whether or not a transaction destination, a time that is a transaction destination, and a cumulative time that is a transaction destination,
For each server, measure the cumulative time that was the destination of the transaction in a given period,
A load balancer. In the load distribution apparatus according to claim 2 or 3,
A server billing unit that performs billing processing based on server usage time data output from the server usage time measurement unit and outputs a billed amount;
A load balancer. The load distribution apparatus according to claim 4, wherein
The server billing unit
A server billing information management table including at least a server ID, whether or not a transaction destination, a time that is a transaction destination, a cumulative time that is a transaction destination, and cumulative fee information,
For each of the servers, executing a billing process based on a time set as a transaction destination in a predetermined period;
A load balancer.   The load distribution device according to any one of claims 1 to 5 is arranged at a front end of a plurality of server groups in which each server can process a transaction, and load distribution is performed by increasing or decreasing the number of servers to which transactions are transmitted. Server system to perform. Leave each server in the server group in a state that can process transactions,
A load balancer located at the front end of the server group,
Receiving an instruction as to whether or not to be a transaction destination for each server;
And a step of transmitting a transaction to a server that is a destination of the transaction, and a processing method for optimizing the processing capacity by increasing or decreasing the number of servers that are the destination of the transaction. Leave each server in the server group in a state that can process transactions,
A load balancer located at the front end of the server group,
Receiving an instruction as to whether or not to be a transaction destination for each server;
Sending a transaction to a server that is the destination of the transaction;
Repeating the step of totaling the time designated as the transaction destination for each server, and outputting the time totaled for each server as the usage time of each server included in the load balancing server system,
Server usage time measurement method characterized by Leave each server in the server group in a state that can process transactions,
A load balancer located at the front end of the server group,
Receiving an instruction as to whether or not to be a transaction destination for each server;
Sending a transaction to a server that is the destination of the transaction;
Repeating the step of counting the time specified for the transaction destination for each server,
Performing pay-as-you-go billing processing based on the aggregated time every predetermined time;
A billing method characterized by the above.

Images