US 2003/0172093 A1Application
Server duplexing method and duplexed server system
Publication Date:2003-09-11
•4 Claims
•5 Drawing Sheets
Abstract
A method of duplicating servers and a duplicated server system that make seamless transition of service in case of server failures possible. The duplicated server system comprises first and second servers that are connected to a network and have the same network address, communication means for making high-speed communication between both servers possible, and switchover controller that designates the first server as the primary server put in operation for other computers and the second server as the secondary server in normal condition. When the first server receives a service request addressed to the system, it passes the service request to the application for processing. The first server sends the recovery data output by the application to the second server by means of the communication means. If the processing related to the service request involves an update of the data held on the server system, the second server executes the application using the recovery data sent from the first server to keep the data held on both servers identical. Service requests and the results of the processing are stored in the stacks in both servers in duplicate.
Metadata
Inventor
- Mitsugu Nagoya
Application Information
Application Number:US 10/312,499
Filing Date:2003-04-22
Priority Date:2003-04-22
Patent Drawings (5 sheets)
Description
Field of the Invention
[0001] The present invention relates to servers, and more particularly to the technology for enhancing the reliability of servers by duplicating them.
Background
[0002] Some of the servers that provide service to a large number of clients over a network are required to have a particularly high reliability for the nature of the services. A measure often employed to meet such a requirement is to duplicate the server by using another server (so called mirror server) which takes over the service task of the server in case of server failure. In such duplicated server systems, the data (herein referred to as application data) used in application processing on the two servers that make up a server system is always kept identical by connecting the two servers by means of high-speed communication means.
[0003] Even in such duplicated server systems, however, if there is a time difference in making the contents of the application data on both servers identical, the service being carried out must be started again from the beginning when the mirror server (secondary server) takes over the tasks in case of server failure while the server that normally provides service for other computers (referred to as the primary server) is actually performing service.
[0004] From this point of view, it is desired that the service being carried out at the occurrence of server failure is also continued by seamless transition such that the occurrence of failure is not perceived by the client who requested the service.
[0005] The object of the present invention is therefore to provide a method of duplicating servers that makes seamless transition of the service being carried out at the occurrence of server failure possible, and duplicated server system and database system mirror-backed-up according to the method.
Summary of the Invention
[0006] Claim 1 of the present invention for solving the above-described problem is a method of duplicating first and second servers of a server system consisting of the first and second servers which are connected to a network and have the same network address, the method being characterized by comprising the steps of: designating the first server as the primary server put in operation for other computers and the second server as the secondary server in normal condition; sending the recovery data output by the application as a result of the execution of the application in response to a service request to the second server when the processing related to the service request involves an update of the data held in the server system; and the second server executing the application using the recovery data sent from the first server, the application data held on the first and second servers thereby being always kept identical.
[0007] Claim 2 is characterized by further involving the steps of: storing the output data of the processing related to a service request into the storage area when the processing related to the service request does not involve an update of the data held in the server system; the first server sending the output data to the second server; and the second server storing the output data sent from the first server into the storage area of the second server for storing data log in order.
[0008] Claim 3 is a server system, comprising first and second servers which are connected to a network and have the same network address, communication means for making high-speed communication between the first server and the second server possible, and switchover control means that communicates with the first and second servers and designates the first server as the primary server put in operation for other computers and the second server as the secondary server in normal condition, and characterized by comprising means for sending the recovery data output by the application as a result of the execution of the application in response to a service request to the second server by means of the communication means when the processing related to the service request involves an update of the application data held in the server system, and means by which the second server executes the application using the recovery data sent from the first server, the application data held on the first and second servers thereby being always kept identical.
[0009] Claim 4 is characterized in that the first server and the second server have the first and second storage areas for storing data log in order therein, respectively, and further include means for storing the output data of the processing related to the network data into the first storage area when the processing related to the network data does not involve an update of the data held in the server system, means by which the first server sends the output data from the first server to the second server by means of the communication means, and means by which the second server stores the output data sent from the first server into the second storage area.
Brief Description of the Drawings
[0010] FIG. 1 is a simplified block diagram that shows the hardware architecture of the duplicated server system being an embodiment of the present invention.
[0011] FIG. 2 shows the data configuration in the HD system of FIG. 1 .
[0012] FIG. 3 illustrates the duplicate operation of the duplicated server system being an embodiment of the present invention. (A) shows the operation in case where processing involves an updating of data, and (B) shows the operation in the case of query-type processing.
[0013] FIG. 4 shows the manner in which service requests and the output data of the processing in response to the service requests are stored in the stacks 49 of the primary server 10 and the secondary server 20.
Description of Preferred Embodiments
[0014] FIG. 1 is a simplified block diagram that shows the hardware architecture of the duplicated server system being an embodiment of the present invention. In FIG. 1 , the duplicated server system 1 comprises two servers 10 and 20 that are connected to a network and have substantially the same hardware architecture and a switchover controller 30 that controls the switching of the operation mode of the servers 10 and 20.
[0015] Since the servers 10 and 20 have substantially the same configuration, both servers are hereinafter referred to like servers 10, 20 and only one server is described. The servers 10 and 20 each are computers suitable for a server. Each of the servers 10 and 20 has a processing circuit 11, 21 which includes a microprocessor and a memory not shown in the Figure, NIC (Network Interface Card) 12, 22 serving as the interface to the network, shared memory 13, 23 used for communication between the servers 10 and 20, HD (Hard Disk) system 14, 24, and communication interface 15, 25 for communication with the switchover controller 30.
[0016] The NICs 12 and 24 are assigned the same MAC (Media Access Control) address Amac so that the servers 10 and 20 appear as a single computer to the network. The shared memories 13 and 23 are dual-port RAMs (Random Access Memories) for high-speed communication between the servers 10 and 20, and have the same addresses assigned to form identical address spaces (referred to as shared memory spaces) on the individual servers. The shared memories 13 and 23 are configured so that, when one of the servers 10 and 20 writes data to its own shared memory space, the same data is written to both shared memories 13 and 23 substantially at the same time.
[0017] The HD systems 14 and 24 have the same configuration consisting of at least one hard disk drive. FIG. 2 shows an example of the contents that are stored in the HD systems 14 and 24. In each HD system, an OS (Operating System) 41 suitable for a server such as UNIX, application 43 used for specific service tasks and data 45, and the software 47 for the server duplication according to the present invention (referred to as the server duplication system herein) are stored as shown in FIG. 2 . In addition, a stack 49 in which every service request to the server system 1 and the output data in response to it are stored as data log is also stored in each HD systems 14, 24 for duplication of the HD systems as described later. The application 43 may be online banking service system and various data base management systems, for example.
[0018] As described above, both servers 10 and 20 have about the same configuration and are designed so as to be able to perform the equal functions. The mode of operation of the servers in which they are operating in service for other computers is referred to as the service mode, and that in which they are operating but not in service for other computers is referred to as the standby mode. The switching between the two modes is made by the switchover controller 30. The switchover controller 30 is also a PC (personal computer) provided with communication interfaces not shown in the figure and specifies the operation mode of each server by performing bidirectional communication with both servers 10 and 20.
[0019] Next, the operation of the duplicated server system 1 is described. When the servers 10 and 20 receive a service request from a remote client over the network, they normally perform the predetermined processing in response to the service request unless otherwise operated by the system administrator at need. For convenience' sake, let us suppose that the server 10 operates in the service mode and the server 20 in the standby mode when the whole system 1 is in normal condition. The server 1 is therefore referred to as the primary server and the server 2 as the secondary server.
[0020] When a service request to the duplicated server system 1 (with destination MAC address Amac) is sent over the network, the NICs 12 and 22 detect the service request, and normally the NIC 12 of the primary server 10 in the service mode actually receives the service request. The operation of the primary server 10 and the secondary server 20 is shown in FIG. 3 .
[0021] FIG. 3 (A) shows the operation when a service request for processing that updates the application data is given over the network 100. When receiving such a service request, the server in the service mode (primary server 10, for example) passes the request to the application 43 for processing (Step 102). The application 43 then performs the processing related to the service request, and outputs various types of data used in the processing as recovery data RD1. The recovery data RD1 is written to the shared memory 13 in order to pass it to the secondary server 20 (Step 104). At this time, the recovery data RD1 is also written to the shared memory 23 of the secondary server 20 (Step 106). The recovery data RD1 written in the shared memory 13 is then stored in the stack 49 in the HD system 14 (Step 108). Similarly, on the secondary server 20, the recovery data RD1 written in the shared memory 23 is stored in the stack 49 in the HD system 24 (Step 110). Further, the secondary server 20 always executes the application 43 using the recovery data RD1 stored in the stack 49 and thereby updates the application data 45 to keep it identical to the application data on the primary server 10.
[0022] FIG. 3 (B) shows the operation of the duplicated server system 1 when it receives a service request for processing that does not involve an update of the application data 45 from the network 100. In FIG. 3 , the service request received is passed to the application 43 (Step 102). The application 43 then performs the processing related to the service request in response and outputs output data OD1. The output data OD1 is sent back to the client over the network 100 (Step 202). The output data OD1 is also written to the shared memories 13 and 23 simultaneously (Step 204 and 206). The data OD1 written in the shared memory 13 and that in the shared memory 23 are mechanically stored in the stack 49 of the HD systems 14 and in the stack 49 of the HD systems 24, respectively (Step 208 and 210). By doing this, the stack 49 of the primary server 10 and that of the secondary server 20 always have the same log data stored as shown in FIG. 4 . Therefore, by always executing the application 43 on the secondary server 20 using the log data, the duplication of the application data 45 can be achieved.
[0023] Back in FIG. 3 (B), the secondary server 20 outputs the output data OD1 received by means of the shared memory 23 (Step 207). Although the application 43 on the secondary server 20 also outputs the output data because it is executed using the log data in the stack 49, this data is not used.
[0024] In both cases shown in FIGS. 3(A) and 3(B), the switchover controller 30 is monitoring the primary server 10 and the secondary server 20 with respect to the hardware. If a failure occurs on the server in the service mode (primary server 10 in this example), the switchover controller switches the primary server 10 to the standby mode and the secondary server 20 to the service mode.
[0025] Since a seamless duplication of the servers can be realized by the present invention as described above, the reliability of the servers can be enhanced.
[0026] Further, although two servers are used in the above embodiment, two or more servers can also be used as the secondary servers.
[0027] Furthermore, the stacks 49 may also be output to the internal memory of each server instead of being output to the HD system each and every time for high-speed processing.
[0028] Thus, by the present invention, a method of duplicating servers that makes possible seamless transition of the service being carried out at the occurrence of server failure, and duplicated server system and database system mirror-backed-up according to the method can be provided.
Claims
1. A method of duplicating first and second servers of a server system consisting of the first and second servers which are connected to a network and have the same network address, characterized by comprising the steps of:
designating said first server as the primary server put in operation for other computers and said second server as the secondary server in normal condition;
sending the recovery data output by the application as a result of the execution of the application in response to a service request to said second server when the processing related to the service request involves an update of the data held in said server system; and
said second server executing the application using said recovery data sent from said first server;
the application data held on said first and second servers thereby being always kept identical.
2. The method of duplicating servers of claim 1, characterized by further involving the steps of:
storing the output data of the processing related to a service request into said storage area when the processing related to the service request does not involve an update of the data held in the server system;
said first server sending said output data to said second server; and
said second server storing said output data sent from said first server into the storage area for storing data log in order of said second server.
3. A duplicated server system, comprising
first and second servers which are connected to a network and have the same network address,
communication means for making high-speed communication between said first server and said second server possible, and
switchover control means that communicates with said first and second servers and designates said first server as the primary server put in operation for other computers and said second server as the secondary server in normal condition,
and characterized by further comprising
means for sending the recovery data output by the application as a result of the execution of the application in response to a service request to said second server by means of said communication means when the processing related to the service request involves an update of the application data held in the server system, and
said second server having means for executing the application using said recovery data sent from said first server,
the application data held on said first and second servers thereby being always kept identical.
4. The duplicated server system of claim 3 wherein said first server and said second server have the first and second storage areas for storing data log in order therein, respectively, and which further includes
means for storing the output data of the processing related to the network data into said first storage area when the processing related to the network data does not involve an update of the data held in said server system,
means by which said first server sends said output data from said first server to said second server by means of said communication means, and
means by which said second server accumulates said output data sent from said first server into said second storage area.