JP2010061622A - Transaction load distribution device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable load distribution, while preventing a speed of transaction processing from getting low, consistency of a transaction from being deprived, and reliability of the transaction from getting low, even in the stateful transaction. <P>SOLUTION: A transaction reception part 11 receives a message, when an address of the message transmitted from a client in each stage of the stateful transaction is an address of own transaction load distribution device. An ID extracting part 12 extracts a resource ID inserted into the message. Transaction distribution parts 12-15 distribute the message to anyone of a plurality of subordinate processors, based on the resource ID. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a transaction load distribution apparatus and program, and in particular, a transaction load distribution apparatus that can perform load distribution even for a transaction that requires a plurality of stages of processing or procedures and needs to consider the state of the previous stage in each stage. Regarding the program.

  Conventionally, when the number of transactions with a processing device, for example, a Web server is enormous, a transaction load distribution device is provided in front of the Web server, and a plurality of Web servers are arranged under the transaction load distribution method. The transaction load balancer virtually has an IP address that serves as a contact point for multiple Web servers, accepts transactions from clients, and distributes them to multiple Web servers under its control. This distribution is performed by, for example, a round robin method or a method of giving priority to a Web server with a fast response time. Such a transaction load distribution device is described in Patent Document 1, and according to this, it is possible to cope with an increase in the number of transactions from a client.

Non-Patent Document 1 describes a transaction that requires a multi-step process or procedure. Such transactions are flexible, linked to transactions, for example, when using two-phase commit to improve the atomicity and durability of transactions, changing the conditions of previously requested transactions later, etc. Necessary for processing.
US Pat. No. 6,128,279 Tetsuo Nakagawa "OSI and UNIX (registered trademark) Distributed Transaction Processing Technology" Software Research Center 1994

  In the case of a transaction in which the conventional transaction load balancer does not consider the relationship between transactions such as transaction state transitions, so multiple stages of processing or procedures are required, and the state of the previous stage must be considered at each stage It cannot be applied as it is.

  Database synchronization (database synchronization: mirroring) and database centralization (sharing) between multiple load-distributed Web servers may be possible, but database synchronization and database concentration are also possible. However, there is a problem that the speed of transaction processing is reduced and the meaning of load distribution is reduced. In addition, transaction consistency may be lost and reliability may be reduced.

  The object of the present invention is to solve the above-mentioned problems and require a multi-stage process or procedure. Even in a transaction where the state of the previous stage needs to be taken into consideration at each stage, the transaction processing speed is reduced and the transaction is not consistent. It is an object of the present invention to provide a transaction load distribution apparatus and program capable of distributing a load without causing a decrease in reliability and reliability.

  In order to solve the above-described problems, the present invention requires a multi-stage process or procedure in a transaction load distribution apparatus that distributes a load from a client to a plurality of processing apparatuses and distributes the load. A transaction receiving unit that receives the message when the destination address inserted in the header of the message transmitted from the client at each stage of the transaction that needs to be considered is the address of the own transaction load balancer; , An ID extraction unit that extracts a resource ID inserted in the header of the message received by the transaction reception unit, and a plurality of processes under the message based on the resource ID extracted by the ID extraction unit Any of the equipment It is characterized in that it comprises a transaction distributing unit for distributing to.

  Further, according to the present invention, the resource ID is blank in a message at the first stage of a transaction, and in a response sent from the processing apparatus to which the message at the first stage is distributed to a client to a message at the subsequent stage. The transaction distribution unit distributes the message at the first stage to one of a plurality of processing apparatuses by an arbitrary method, and inserts the message at the subsequent stage into the message. It is characterized by sorting according to the resource ID.

  Further, according to the present invention, the resource ID is blank in a message at the first stage of a transaction, and in a response sent from the processing apparatus to which the message at the first stage is distributed to a client to a message at the subsequent stage. The transaction distribution unit inserts the message at the first stage into one of a plurality of processing devices by an arbitrary method, and the message at the subsequent stage is assigned to the message. It is characterized by sorting according to the inserted resource ID.

  According to the present invention, the resource ID is inserted by a client from a message at the first stage of a transaction, and the transaction distribution unit includes a resource ID in which a message at each stage of the transaction is inserted into the message. It is characterized by sorting according to.

  In addition, the present invention is characterized in that the resource ID inserted in the message at each stage of the transaction has a value in a range given in advance to each processing device.

  Further, the present invention is characterized in that the resource ID inserted in the message at each stage of the transaction has a value that identifies each processing device by a remainder value obtained by dividing the resource ID by the number of processing devices.

  Furthermore, the present invention is characterized in that the resource ID inserted in the message at each stage of the transaction has a value obtained from an integrated value or a hash value of the transaction entry and ID information having a unique value in the network. It is said.

  The present invention can also be realized as a program that is mounted on a computer and functions as a transaction load distribution apparatus.

  According to the transaction load distribution apparatus and program of the present invention, it is necessary to perform a multi-stage process or procedure, and even in a transaction that needs to consider the state of the previous stage in each stage, Without sharing the database, it is possible to distribute the load while maintaining independence between the processing apparatuses. Therefore, even if a transaction requires multiple stages of processing or procedures, and each stage needs to consider the state of the previous stage, it does not cause a decrease in transaction processing speed, loss of transaction consistency or reliability, The load can be distributed.

  The present invention will be described below with reference to the drawings. Hereinafter, a case where the present invention is implemented as a transaction load distribution apparatus will be described. However, the present invention can also be implemented as a program that is mounted on a computer and functions as a transaction load distribution apparatus.

  FIG. 1 is a block diagram showing a basic configuration of a basic transaction processing system to which the present invention is applied.

  The transaction load balancer 10 is provided between the client 1 and a plurality of distributed servers (processing devices) 2-1, 2-2, 2-3,..., Accepts transactions from the client 1, and Distribute to multiple distributed servers 2-1, 2-2, 2-3, ....

  FIG. 2 is a block diagram showing an embodiment of the transaction load distribution apparatus according to the present invention.

  The transaction load balancer 10 includes a transaction accepting unit 11, an ID extracting unit 12, an ID processing unit 13, a destination editing unit 14, a transaction switching unit 15, a policy database (DB) 16 and a destination DB 17, and processes transactions from the client 1. Accept and distribute to multiple distributed servers 2-1, 2-2, 2-3,.

  The transaction load balancer 10 virtually has an IP address as a distribution server 2-1, 2-2, 2-3,..., And the transaction reception unit 11 inserts this IP address into the header. The received transaction from client 1 is accepted.

  The ID extraction unit 12 extracts a resource ID from the transaction accepted by the transaction acceptance unit 11. For example, in the case of a transaction transferred by SOAP (Simple Object Access Protocol), the XML parser unit as the ID extraction unit 12 analyzes the transaction and extracts a Resource ID. The policy DB 16 stores a policy necessary for analyzing a transaction header and extracting a Resource ID.

  The following ID processing unit 13, destination editing unit 14, and transaction switching unit 15 send messages based on the resource ID extracted by the ID extraction unit 12 to subordinate distributed servers 2-1, 2-2, 2-3, It functions as a transaction distribution unit that distributes to any of.

  First, the ID processing unit 13 determines the Resource ID extracted by the ID extracting unit 12 in order to determine the destination distributed servers 2-1, 2-2, 2-3,. Calculate. At this time, the defined processing method (calculation algorithm) refers to the pre-processing definition database in which the resource ID calculation processing is defined. This pre-processing definition database definition can be arbitrarily changed, and depends on the configuration of the distributed server.

  The destination editing unit 14 refers to the resource ID profile stored in the destination DB 17, and edits the destination address based on the Resource ID calculated by the ID processing unit 13. In the resource ID profile, the relationship between the Resource ID and the destination distributed servers 2-1, 2-2, 2-3,... Is defined.

  The transaction switching unit 15 distributes the transaction to the distributed servers 2-1, 2-2, 2-3,... According to the destination address edited by the destination editing unit 14.

  FIG. 3 is a flowchart illustrating an example of the load distribution operation of the transaction load distribution apparatus.

  In this example, a range of Resource IDs is given in advance to each of the distributed servers 2-1, 2-2, 2-3,. For example, a Resource ID of 0 to 10000 is given to the distributed server 2-1, a Resource ID of 10001 to 20000 is given to the distributed server 2-2, and a Resource ID of 20001 to 30000 is given to the distributed server 2-3. give.

  In the following, a transaction that requires a multi-step process or procedure and needs to consider the state of the previous step in each step is referred to as a stateful transaction (stateful transaction (stateful trnsaction). Is "ID1", the address of the transaction load balancer 10 is "ID10", and the addresses of the distributed servers 2-1, 2-2, 2-3, ... are "IDA", "IDB", It is assumed that “IDC”,.

  First, the client 1 transmits a message (1) of the first stage of a transaction having a state. This first message (1) includes a source address (SA), a destination address (DS), and a Resource ID (ID) in the header. In this case, SA = “ID1”, DS = “ID10”, ID = "Blank (empty)". This message (1) is, for example, an inquiry as to whether or not processing constituting a transaction can be realized in the transmission destination processing device (DS = “ID10”)

  The transaction load balancer 10 accepts the message (1) according to the transmission destination address (DS = “ID10”) of the message (1). Next, the Resource ID (ID) in the header of the message (1) is extracted to identify the destination distributed server. However, since ID = “blank (empty)” and the destination distributed server is not specified, the message (1 ) Is distributed to any of the subordinate distributed servers 2-1, 2-2, 2-3,... According to, for example, the round robin method. For example, the message (1) is distributed to the distributed server 2-1 by modifying the transmission destination address DS in the header to “IDA” which is the address of the distributed server 2-1.

  The distributed server 2-1 receives the message (1) from the transaction load balancer 10 and returns an approval response (1 ′) to the transaction load balancer 10 if the process constituting the transaction can be realized. The distributed server 2-1 sets the source address (SA) and destination address (DS) in the header of the response (1 ′) at this time to “IDA” and “ID1”, respectively, and Resource ID (ID ) Is a value within the range given to the self-distributed server, for example, “1000”.

  Upon receiving the response (1 ′) from the distribution server 2-1, the transaction load distribution device 10 corrects the source address (SA) in the header of the response (1 ′) to “ID10” and transmits it to the client 1. .

  Upon receiving the approval response, the client 1 transmits a message (2) at the next stage to the transaction load balancer 10. The SA and DS of the message (2) are the same as the message (1), but the Resource ID is the one included in the header of the response (1 ′) (ID = “1000”).

  The transaction load distribution apparatus 10 accepts the message (2) according to the transmission destination address (DS = “ID10”) of the message (2). Next, the Resource ID (ID) in the header of the message (2) is extracted to specify the destination distribution server. In message (2), ID = “1000” and the destination distribution server 2-1 is specified, so the message (2) is sent with the destination address DS in its header as the address of the distribution server 2-1. Modify to “IDA” and distribute to distributed server 2-1. As a result, the message (1) at the first stage and the message (2) at the next stage of the transaction having a state are distributed to the same distributed server 2-1.

  The distribution server 2-1 receives the message (2) from the transaction load distribution apparatus 10, performs processing according to the instruction in the message (2), and returns a response (2 ′) to the transaction load distribution apparatus 10. SA, DA, ID in the header of the response (2 ′) at this time is the same as that of the response (1 ′).

  Upon receiving the response (2 ′) from the distributed server 2-1, the transaction load balancer 10 modifies the source address (SA) in the header of the response (2 ′) to “ID10” and transmits it to the client 1. .

  Similarly, messages following the transaction having a state can be distributed to the same distributed server 2-1.

  As described above, in this example, the distributed server to which the message at the first stage of the transaction having the state is distributed inserts the Resource ID given in advance to the own device into the header and returns a response, and the client However, the resource ID included in the response is inserted into the message at the subsequent stage, and the transaction load balancer distributes the resource ID to the distributed server as a clue. It can be executed independently without any contradiction.

  FIG. 3 is a flowchart showing another example of the load balancing operation of the transaction load balancing apparatus. In the example of FIG. 2, the Resource ID is assigned by the distributed server, but in this example, the Resource ID is assigned by the transaction load balancer.

  The message (1) of the first stage of the transaction with state is sent from the client 1, and the transaction load balancer 10 sends the message (1) according to the destination address (DS = “ID10”) of the message (1). The process is the same as the example of FIG. 2 until the message (1) is distributed to the distributed servers 2-1 according to, for example, the round robin method.

  The distribution server 2-1 receives the message (1) from the transaction load distribution apparatus 10 and returns an approval response (1 ′) to the transaction load distribution apparatus 10. The source address (SA) and destination address (DS) in the header of the response (1 ′) at this time are set to “IDA” and “ID1”, respectively, as in the example of FIG. The Resource ID (ID) is blank (empty) as it is, that is, the Resource ID is not assigned by the distributed server, which is different from the example of FIG.

  Upon receiving the response (1 ′) from the distributed server 2-1, the transaction load balancer 10 corrects the source address (SA) in the header of the response (1 ′) to “ID10”, and further adds a Resource ID ( ID) is transmitted to the client 1 as a value within the range given to the distributed server 2-1, for example, “1000”. That is, the Resource ID is assigned by the transaction load balancing apparatus 10. The subsequent operations are the same as those in the example of FIG.

  Also in this example, the transaction load balancer inserts the Resource ID given in advance to the distributed server that distributed the message into the header and returns a response to the message at the first stage of the transaction with state. The client inserts the Resource ID included in the response into the message at the subsequent stage, and the transaction load balancer distributes the Resource ID to the distributed server as a clue. Distributed server can be run independently, consistently and consistently.

  FIG. 4 is a flowchart showing still another example of the load balancing operation of the transaction load balancing apparatus. In this example, the Resource ID given to the distributed server is known in advance on the client side. In this case, it is not necessary to assign the Resource ID by the distributed server or transaction load balancer, and the Resource ID that identifies the distributed server is included from the message (1) in the first stage of the transaction having the state from the client 1. Can be made.

  The transaction load balancer identifies the distributed server that distributes the message (1) from the Resource ID of the message (1), for example, the distributed server 2-1, and the subsequent messages (2),. It can be specified and distributed to the distributed server 2-1.

  Although the embodiment has been described above, the present invention is not limited to the above embodiment, and can be variously modified. For example, in the above embodiment, a range of Resource IDs is given in advance to each of the distributed servers 2-1, 2-2, 2-3,..., And the distributed servers 2-1, 2-2,. 2-3, ... are identified, but various resource IDs can be used for each distributed server 2-1, 2-2, 2-3, ... it can.

  For example, the distributed servers 2-1, 2-2, 2-3,... Can be identified from the remainder value of the Resource ID. For example, when a transaction is distributed to three distributed servers 2-1, 2-2, and 2-3, the distributed server 2-1 or 2-2 depends on whether the remainder value of Resource ID is 0, 1, or 2. This can be realized by assigning a Resource ID to each of the distributed servers 2-1, 2-2, and 2-3 and storing this rule in the destination DB 17 so that it can be identified.

  Also, a value obtained from the integrated value or hash value of the transaction entry (for example, serial number) and ID information such as an IP address or MAC address that has a unique value in the network is generated as a unique ID. You may make it give to a distributed server as Resource ID. In this case, by calculating backward from the Resource ID, it is possible to derive a distributed server as a transaction destination.

  The present invention can be applied to a load distribution system for Web transactions, a load distribution system for transactions using SOAP, XML, SIP (Session Initiation Protocol), Diameter protocol, COPS protocol, H.248 protocol, and the like.

It is a block diagram which shows the basic transaction processing structure to which this invention is applied. It is a block diagram which shows one Embodiment of the transaction load distribution apparatus which concerns on this invention. It is a flowchart which shows the example of the load distribution operation | movement of a transaction load distribution apparatus. It is a flowchart which shows the other example of the load distribution operation | movement of a transaction load distribution apparatus. It is a flowchart which shows the further another example of load distribution operation | movement of a transaction load distribution apparatus.

Explanation of symbols

1 ... Client, distributed server ... 2-1,2-2,2-3, 10 ... Transaction load balancing device, 11 ... Transaction accepting unit, 12 ... ID extracting unit, 13. ..ID processing unit, 14 destination editing unit, 15 transaction switching unit, 16 ... policy database (DB), 17 ... destination DB

Claims (8)

In a transaction load balancer that distributes transactions from clients to multiple processing devices and distributes the load,
For transactions that require multiple stages of processing or procedures, and each stage needs to consider the state of the previous stage, the destination address inserted in the header of the message sent from the client at each stage is the local transaction. A transaction accepting unit that accepts the message when the address is the address of the load balancer;
An ID extraction unit that extracts a resource ID inserted in a header of a message received by the transaction reception unit;
A transaction load distribution apparatus comprising: a transaction distribution unit that distributes the message to one of a plurality of processing devices under the resource ID extracted by the ID extraction unit. The resource ID is blank in the first stage message of the transaction, and in the subsequent stage message, the processing apparatus inserts it in the response sent from the processing apparatus to which the first stage message is distributed to the client. The transaction distribution unit distributes the message at the first stage to any one of the plurality of processing devices by an arbitrary method, and the message at the subsequent stage is assigned according to the resource ID inserted in the message. The transaction load distribution apparatus according to claim 1, wherein distribution is performed. The resource ID is blank in the first stage message of the transaction, and in the subsequent stage message, the transaction load balancer during the response sent from the processing apparatus to which the first stage message is distributed to the client. The transaction distribution unit distributes the message at the first stage to one of the plurality of processing devices by an arbitrary method, and the resource ID at which the message at the subsequent stage is inserted into the message. The transaction load distribution apparatus according to claim 1, wherein distribution is performed according to: The resource ID is inserted by a client from a message at the first stage of a transaction, and the transaction distribution unit distributes a message at each stage of the transaction according to a resource ID inserted in the message. The transaction load distribution apparatus according to claim 1. 5. The transaction load distribution apparatus according to claim 1, wherein the resource ID inserted in the message at each stage of the transaction has a value in a range given in advance to each processing apparatus. 5. The resource ID inserted in a message at each stage of a transaction has a value for identifying each processing device by a remainder value obtained by dividing the resource ID by the number of processing devices. The transaction load balancer described in 1. The resource ID inserted in the message at each stage of the transaction has a value obtained from an integrated value or a hash value of the transaction entry and ID information having a unique value in the network. 4. The transaction load distribution device according to any one of 4 above. Computer
For transactions that require multiple stages of processing or procedures, and each stage needs to consider the state of the previous stage, the destination address inserted in the header of the message sent from the client at each stage is the local transaction. A transaction accepting means for accepting the message when it is the address of the load balancer;
ID extraction means for extracting a resource ID inserted in a header of a message received by the transaction receiving unit;
Based on the resource ID extracted by the ID extraction unit, the message functions as a transaction distribution unit that distributes the message to any of a plurality of subordinate processing devices,
A program for distributing transactions from clients to multiple processing devices for load distribution.

Images