JP2001051963A - Parallel server system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the loss of a resource managing function caused by the down (ordinary stop or fault stop) of a server and to distribute the load of resource management to plural servers. SOLUTION: In this system, plural servers (11 and so on) share plural resources through a network. Each of plural resources is managed by one corresponding server of the servers. Each of the servers is provided with a down detecting part 1 for detecting the down of a specified server, a resource management succeeding part 4 for succeeding the management of a resource managed by the specified down server and a load distributing part 7 for distributing the load of the management of the resource.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a resource sharing type parallel server system in which a plurality of servers share at least one resource.

[0002]

2. Description of the Related Art FIG. 22 is a block diagram showing the configuration of a conventional disk distributed parallel video server system.
The parallel video server system shown in FIG. 22 includes servers 51 to 53, magnetic disk devices 61 to 63 locally arranged in each of the servers 51 to 53, and a network (for example, ,
Local area network) 70.

Each of the servers 51 to 53 executes a video distribution service in response to a request from a user. If the video data requested by the user is in the local magnetic disk device, the server reads the video data from the local magnetic disk device. If the video data requested by the user is in a remote magnetic disk device connected to another server, the server requests the other server to transfer data and reads the video data via the network 70.

In such a parallel video server system, each server manages a local magnetic disk device, and when there is a data transfer request from another server, the server that receives the request transfers the data to the band of the magnetic disk device. Manage resources and determine whether data transfer is possible.

However, in this system, when the video data requested by the server exists in the magnetic disk device of another server, the video data is transferred to the network 70 via the other server.
Therefore, the two servers are related to each other, and the resource consumption of the memory and the bus is doubled as compared with the case where the video data exists in the local magnetic disk device. Therefore, when such communication occurs between servers, the service delivery performance of the entire parallel video server system decreases in proportion to the number of communication.

Therefore, in order to suppress the number of communications via the network 70, it is necessary to optimally arrange the video data according to the access bias for each server.

In order to avoid this problem, a disk sharing type parallel video server system in which a plurality of servers and a magnetic disk device are directly connected via a network has been considered.

FIG. 23 is a block diagram of the parallel video server system. In this parallel video server system, each of the servers 81 to 83 transmits each of the magnetic disk devices 91 to
93 can be directly accessed, and data transfer via another server does not occur.

For this reason, the service delivery performance of the entire parallel video server system does not decrease even if the video data in any of the magnetic disk devices is accessed. However, since the servers 81 to 83 share the magnetic disk devices 91 to 93, resource management such as bandwidth control of the magnetic disk devices and bandwidth control of the network 100 is required between the servers.

Usually, a specific server (for example, server 8
1) has such a resource management function, and the other server executes access by obtaining access permission from the server 81 having the resource management function when accessing the disk.

[0011]

However, in a disk sharing type parallel video server system, if only a specific server manages resources, the specific server cannot be stopped, and There is a problem that the parallel video server system loses its resource management function when it stops due to a failure. Also,
If the resource management function is concentrated on a specific server and the resource management function and the video distribution service function are concurrently performed on the specific server, the load on the resource management affects the video distribution service of the server. There is also a problem that the video is interrupted.

The present invention has been made in consideration of the above problems, and prevents loss of a resource management function due to a server down (normal stop or failure stop) and distributes the resource management load to a plurality of servers. It is an object of the present invention to provide a parallel server system capable of performing the following.

[0013]

A parallel server system according to the present invention is a parallel server system in which a plurality of servers share a plurality of resources via a network, wherein each of the plurality of resources includes the plurality of servers. Among the plurality of servers, each of the plurality of servers is managed by a down detection unit that detects that a specific server is down, and the down server is managed by the down specific server. A resource management takeover unit that takes over the management of the resources, and a load distribution unit that distributes the management load of the plurality of resources, thereby achieving the above object.

The load distribution unit specifies a first server having a maximum resource management load and a second server having a minimum resource management load at a fixed cycle, and determines a resource management load by the first server and the second server. If the difference from the resource management load by the second server is larger than a predetermined value, the first server
The resource management takeover unit of the second server may be instructed to take over the resource management from the server.

[0015] When the down detection unit detects that the specific server has gone down, the load distribution unit identifies the second server with the minimum resource management load, and May instruct the resource management takeover unit of the second server to take over the resource management from the second server.

When a new server including the down detection unit, the resource management handover unit, and the load distribution unit is added to the parallel server system, the new server load distribution unit will The first server having the largest load may be specified, and the resource management takeover unit of the new server may be instructed to take over the resource management from the first server.

The load distribution unit controls the resource management takeover unit so as to take over the management of the resources managed by the downed specific server within a range in which the resource management load amount is smaller than a predetermined upper limit value. You may.

Another parallel server system according to the present invention is a parallel server system in which a plurality of servers share network resources, wherein the plurality of servers are connected to a plurality of storage devices via the network. The resources of the network are managed by a corresponding one of the plurality of servers, and resource management information for managing the resources of the network is stored in a corresponding one of the plurality of storage devices. Each of the plurality of servers stores a server down detection unit that detects that a specific server among the plurality of servers has gone down, and the downed particular server manages resources of the network. In this case, a resource management takeover unit that takes over the management of the resources of the network A storage device down detecting unit that detects that the storage device storing the resource management information has gone down and rewrites the resource management information to another storage device, thereby achieving the above object. You.

[0019]

Embodiments of the present invention will be described below with reference to the drawings.

(Embodiment 1) In Embodiment 1, the configuration and operation of the present invention will be described using a parallel video server system 10 as an example. However, the application of the present invention is not limited to the parallel video server system. The present invention can be applied to any resource sharing type parallel server system in which a plurality of servers share a plurality of resources via a network. Here, “resource” refers to a resource that can be accessed from a plurality of servers. The resources are managed by one corresponding server among the plurality of servers. The resource is, for example, a band of the magnetic disk device. These are the same in Embodiments 2 and 3 described later.

FIG. 1 is a block diagram showing the configuration of the parallel video server system 10. In the parallel video server system 10, a plurality of servers 11 to 13
The bandwidths of the plurality of magnetic disk devices 21 to 24 are shared as a plurality of resources via the “0”.

The servers 11 to 13 provide a video distribution service. The magnetic disk devices 21 to 24 store video data. The network 30 interconnects the servers 11 to 13 and the magnetic disk devices 21 to 24. The network 30 is, for example, an FC-AL (Fiber Ch).
Ann Arbitrated Loop).

In the example shown in FIG. 1, the servers 11 to 13
Are server 11, server 12, server 13, server 11
The order is cyclic. The magnetic disk device 21 stores a video title A, the magnetic disk device 22 stores a video title B, the magnetic disk device 23 stores a video title C, and the magnetic disk device 23 stores a video title C. 24, a video title D is stored. The playback rate of each of the video titles A to D is 30 Mbps.

Each of the servers 11 to 13 accesses the magnetic disk devices 21 to 24 via the network 30 in response to a request from the user, and
To read the video data stored in.

FIG. 2 is a functional block diagram showing the configuration of the server 11. The configuration of the servers 12 and 13
The configuration is the same as

The server 11 includes a down detection unit 1, a band reservation application unit 2, a band management unit 3, and a band management takeover unit 4.
, A user request processing unit 5, a video distribution unit 6, and a load distribution unit 7.

The band reservation application unit 2 is a magnetic disk drive 2
When accessing each of 1 to 24, a bandwidth reservation message is sent to the server that manages the bandwidth of the magnetic disk device. The bandwidth reservation application unit 2 has a bandwidth management server table. The bandwidth management server table defines the correspondence between the magnetic disk devices and the servers that manage the bandwidth of the magnetic disk devices. The bandwidth management server table in the bandwidth reservation application unit 2 is updated in response to the update instruction.

FIG. 3 shows an example of the bandwidth management server table.
A server that manages the band is uniquely determined for each magnetic disk device. In the example shown in FIG. 3, the bandwidth of the magnetic disk device 21 is managed by the server 11, the bandwidth of the magnetic disk device 22 is managed by the server 12,
The band of the magnetic disk device 23 is managed by the server 13, and the band of the magnetic disk device 24 is managed by the server 11.

Referring again to FIG. 2, the band management unit 3 discretely manages the band of the magnetic disk device in the form of a ticket, receives a band reservation message from the server, and returns the ticket if any. . One ticket is 30Mbps
And a plurality of disks are prepared according to the band of the magnetic disk device. The usage status of the ticket and the distribution status to the server are managed by the bandwidth management unit 3 as ticket management information. Each time the ticket management information is updated, it is written to the corresponding magnetic disk device.

The user request processing unit 5 receives a video distribution request from the user and instructs the bandwidth reservation application unit 2 to apply for a bandwidth reservation, and instructs the video distribution unit 6 to distribute a video if a ticket is obtained.

The video distribution unit 6 distributes a video in response to an instruction from the user request processing unit 5.

The down detecting unit 1 periodically checks the operation status of the server by periodically transmitting / receiving a message to / from the server which is ordered immediately before, and determines whether the server is down (normal stop or failure stop). Upon detection, it instructs the band management takeover unit 4 to take over the band management.

The band management handover unit 4 includes the down detection unit 1
In response to the bandwidth management takeover instruction from the server, the server that went down takes over the bandwidth management of the magnetic disk device, and sends an instruction to update the bandwidth management server table to the bandwidth reservation application unit 2 of the other server.

The load distribution unit 7 checks the amount of bandwidth management load (the number of disks in charge) of each server at regular intervals, and determines which server has the largest bandwidth management load and which server has the least bandwidth management load. To identify. If the difference between the load on the server with the highest bandwidth management load and the load on the server with the lowest bandwidth management load is greater than a predetermined value, the server takes over the bandwidth management from the server with the highest bandwidth management load. It instructs the bandwidth management takeover unit 4 of the server with the smallest bandwidth management load.

Hereinafter, an operation example of the parallel video server system 10 will be described.

When the user request processing unit 5 of the server 12 receives the distribution request of the video title A from the user, it instructs the band reservation application unit 2 of the server 12 to apply for the band reservation. The bandwidth reservation application unit 2 of the server 12
Sends a bandwidth reservation message to the bandwidth management unit 3 of the server 11 that manages the bandwidth of the magnetic disk device 21 in which is stored. The bandwidth management unit 3 of the server 11 checks the ticket of the magnetic disk device 21, and if there is a ticket, returns the ticket to the bandwidth reservation application unit 2 of the server 12, updates the ticket management information, and writes the updated information to the magnetic disk device 21. When the ticket is obtained, the user request processing unit 5 of the server 12 sends a video distribution start instruction to the video distribution unit 6 of the server 12. If the ticket cannot be obtained, the user request processing unit 5 of the server 12 notifies the user that the service cannot be performed. Upon receiving the video distribution start instruction, the video distribution unit 6 of the server 12 periodically distributes the video data. As a result, the video is reproduced.

Further, the down detection unit 1 of each server exchanges messages with the down detection unit 1 of the server which is ordered immediately before at predetermined time intervals (for example, 10 seconds), and exchanges messages. Detect server down due to failure. Server 1 down detection unit 1
3 is detected, the down detecting unit 1 of the server 11
Instructs the bandwidth management takeover unit 4 of the server 11 to take over the bandwidth management so as to take over the bandwidth management of the down server 13.

The band management takeover unit 4 of the server 11 receives the instruction to take over the band management, reads out the ticket management information from the magnetic disk device 23, and takes over the band management of the server 13. Then, the server 13
except for.

Also, the server 13 is notified to the bandwidth reservation application section 2 of the server 12 that the server 13 has gone down and the server 11 has taken over the bandwidth management of the magnetic disk device 23. Upon receiving the notification, the bandwidth reservation application unit 2 of the server 12 sets the bandwidth management server of the magnetic disk device 23 in the bandwidth management server table to the server 1
1 and the server 13
except for. As a result, the server 11 is in charge of the band management of the magnetic disk device 23 in addition to the magnetic disk devices 21 and 24 (FIGS. 4 and 5).

In this state, the operation of the load distribution unit 7 will be described.

Server 11 (first server by ordering)
The load distribution unit 7 checks the bandwidth management server table at predetermined time intervals (for example, three minutes), and determines which server (server 11) has the largest bandwidth management load and which server has the least bandwidth management load (server 12). ) And the amount of bandwidth management load (ie, the difference in the number of magnetic disk devices responsible for bandwidth management).
Check if is greater than or equal to 2. When the difference is 2 or more, the bandwidth management takeover unit 4 of the server 12 is instructed to take over the bandwidth management from the server 11 to the server 12. The band management takeover unit 4 of the server 12 that has received the band management takeover instruction reads out the ticket management information from the magnetic disk device 24 (the highest numbered magnetic disk device in which the server 11 is in charge of the band management) and executes the band management server table. Is updated, and the band management is taken over, and the band reservation application unit 2 of the server 11 is notified that the server 12 has taken over the band management of the magnetic disk device 24. Upon receiving the notification, the bandwidth reservation application unit 2 of the server 11 changes the bandwidth management server of the magnetic disk device 24 in the bandwidth management server table to the server 12 (FIG. 6).

FIG. 7 shows an operation flow of the load distribution unit 7 and the bandwidth reservation application unit 2 at this time. FIG. 8 shows an operation flow of the band management takeover unit 4.

When the down detecting section 1 of the server 11 detects that the server 13 has gone down, the server 11 does not always need to take over the band management of the magnetic disk device 23 which was in charge of the server 13. For example, server 1
If the load amount of one bandwidth management is equal to or more than the predetermined upper limit, the server 11 may request the server 12 to be ordered next to take over the bandwidth management from the server 13 that has gone down.

Hereinafter, an operation example of the parallel video server system 10 will be described.

The down detection unit 1 of each server exchanges messages with the down detection unit 1 of the server that is ordered immediately before at predetermined time intervals (for example, 10 seconds), and the failure of the message exchange fails. To detect server down. When the down detection unit 1 of the server 11 detects that the server 13 is down, the down detection unit 1 of the server 11 instructs the band management takeover unit 4 of the server 11 to take over the band management so as to take over the band management of the down server 13. I do.

The bandwidth management takeover unit 4 of the server 11 notifies the load distribution unit 7 of the server 11 that the instruction has been received.

The load distribution unit 7 of the server 11
It is determined whether or not the load amount of the bandwidth management (that is, the number of magnetic disk devices for which the server 11 is in charge of the bandwidth management) is equal to or larger than a predetermined upper limit (for example, 2). The predetermined upper limit is calculated by, for example, (the number of magnetic disk devices / the minimum integer equal to or more than the number of servers). When the load amount of the bandwidth management of the server 11 is equal to or greater than the predetermined upper limit, the load distribution unit 7 of the server 11 sends a non-permission message for disallowing the handover of the bandwidth management to the bandwidth management takeover unit 4 of the server 11. To return. If the bandwidth management load of the server 11 is smaller than the predetermined upper limit, the load distribution unit 7 of the server 11
Returns a permission message for permitting the handover of the band management to the band management handover unit 4 of the server 11. For example, the predetermined upper limit is 2, and as shown in FIG.
If the number of magnetic disk devices 1 is in charge of bandwidth management is 2, the load distribution unit 7 of the server 11 returns a non-permission message to the bandwidth management takeover unit 4 of the server 11.

In response to the rejection message, the bandwidth management takeover unit 4 of the server 11 instructs the bandwidth management takeover unit 4 of the server 12, which is ordered next to the server 11, to take over the bandwidth management of the downed server 13. .

The bandwidth management takeover unit 4 of the server 12 notifies the load distribution unit 7 of the server 12 that the instruction has been received. If the bandwidth management load of the server 12 is equal to or greater than the predetermined upper limit, the load distribution unit 7 of the server 12
Returns a non-permission message to the band management handover unit 4 of the server 12 to disallow the handover of the band management. If the load amount of the bandwidth management of the server 12 is smaller than the predetermined upper limit, the load distribution unit 7 of the server 12 returns a permission message permitting the handover of the bandwidth management to the bandwidth management handover unit 4 of the server 12. For example, the predetermined upper limit is 2,
As shown in FIG. 3, when the number of magnetic disk devices for which the server 12 is in charge of bandwidth management is one, the server 12
The load distribution unit 7 returns a permission message to the bandwidth management takeover unit 4 of the server 12.

Upon receiving the permission message, the band management takeover unit 4 of the server 12 reads out the ticket management information from the magnetic disk device 23, updates the band management server table, and takes over the band management of the server 13. Thereafter, the server 13 is removed from the server ordering.

Further, the server 13 is notified to the band reservation application section 2 of the server 11 that the server 13 has gone down and the server 12 has taken over the band management of the magnetic disk device 23. Upon receiving the notification, the bandwidth reservation application unit 2 of the server 11 sets the bandwidth management server of the magnetic disk device 23 in the bandwidth management server table to the server 1.
2 and the server 13
except for. As a result, the server 12 is in charge of managing the bandwidth of the magnetic disk device 23 in addition to the magnetic disk device 22 (FIG. 9).

FIG. 10 shows an operation flow of the band management handover unit 4 at this time. FIG. 11 shows an operation flow of the load distribution unit 7.

When the down detecting unit 1 of the server 11 detects that the server 13 is down, the down detecting unit 1 of the server 11 notifies the server 11 that the server 13 has gone down without passing through the band management takeover unit 4. May be notified. In this case, the server 11
The load distribution unit 7 checks the bandwidth management load of each server, and issues a bandwidth management takeover instruction to each of the plurality of servers so that the bandwidth management load of each server becomes smaller than a predetermined upper limit. Good.

According to the first embodiment, the down detecting unit 1 of the server 11 detects the down of the server 13, and the band management takeover unit 4 of the server 11 takes over the band management of the magnetic disk device 23 from the server 13. By giving the instruction, the loss of the band management function of the magnetic disk device 23 due to the server 13 going down can be prevented. In this case, by controlling the server 13 to take over the bandwidth management of the server 13 within a range in which the bandwidth management load of the server 11 is smaller than the predetermined upper limit value, the bandwidth management load can be distributed among the servers.

Further, the load distribution unit 7 periodically compares the load of the bandwidth management of the disk in each server, and the server 12 having the minimum bandwidth management load is shifted from the server 11 having the maximum bandwidth management load to the magnetic disk drive. By giving an instruction to take over the bandwidth management of 24, the load of the bandwidth management can be distributed among the servers.

(Embodiment 2) FIG. 12 is a functional block diagram showing a configuration of a server 11 according to Embodiment 2 of the present invention. The configuration of the server 11 shown in FIG. 12 is the same as the configuration of the server 11 shown in FIG. 2 except for the load distribution unit 7a.

When the servers 11 and 12 are connected to the network 30 (for example, in the state shown in FIGS. 4 and 5), the server 14 is newly connected to the network 30.
Is assumed to be added. Here, it is assumed that the servers 12 and 14 have the same configuration as the server 11.

In this case, the load distribution unit 7a of the server 14
Notifies the server 11 and the server 12 that the server 14 has been added. Server 14 is ordered next to server 12 (before server 11) (FIG. 13). The load distribution unit 7a of the server 14 acquires the bandwidth management server table from the immediately preceding server 12, and takes over the bandwidth management from the server 11 having the largest bandwidth management load (the number of magnetic disk devices in charge of bandwidth management). As described above, the bandwidth management takeover instruction is instructed to the bandwidth management takeover unit 4 of the server 14.

The band management takeover unit 4 of the server 14 receives the band management takeover instruction and reads the ticket management information from the magnetic disk device 24 (the highest numbered magnetic disk device in which the server 11 is in charge of the band management). Update the bandwidth management server table and take over the bandwidth management. Also,
The server 11 and the server 12 are notified that the band management of the magnetic disk device 24 has been taken over by the server 14.

Upon receiving the notification, the bandwidth reservation application units 2 of the servers 11 and 12 change the bandwidth management server of the magnetic disk device 24 in the bandwidth management server table to the server 14 (FIG. 14).

According to the second embodiment, it is possible to cope with the downtime of the server that manages the bandwidth of the magnetic disk device as in the parallel video server system of the first embodiment, and to add a new server 14 to the network 30. , The load balancer 7a of the server 14 compares the load of the band management of the magnetic disk device in each server, and
By instructing the server 4 to take over the bandwidth management from the server 11 having a large load, the load of the bandwidth management can be distributed among the servers.

(Embodiment 3) FIG. 15 is a functional block diagram showing a configuration of a server 11 according to Embodiment 3 of the present invention. The configuration of the server 11 shown in FIG. 15 includes a down detection unit 1b, a load distribution unit 7b, a band management takeover unit 4
b, the configuration is the same as that of the server 11 shown in FIG.

It is assumed that servers 11 to 13 are connected to network 30 (for example, the state shown in FIGS. 1 and 3). Here, the servers 12 and 13
It is assumed that it has the same configuration as the server 11.

The down detection section 1b of each server periodically exchanges messages with the down detection section 1b of the server that is ordered immediately before, and detects the down of the server due to failure of the message exchange. For example, when the down detecting unit 1b of the server 11 detects that the server 13 is down,
The down detection unit 1b of the server 11 notifies the load distribution unit 7b of the server 11. The load balancing unit 7b of the server 11 refers to the bandwidth management server table to identify the server 12 having the minimum bandwidth management load except for the server 13, and to take over the bandwidth management of the down server 13 so as to take over the bandwidth management.
To the band management takeover unit 4b.

The band management takeover unit 4b of the server 12 receives the instruction of the band management takeover, reads the ticket management information from the magnetic disk device 23, updates the band management server table, and takes over the band management of the server 13. Thereafter, the server 13 is removed from the server ordering. Also, the server 11
To the bandwidth reservation application unit 2b that the server 12 has taken over the bandwidth management of the magnetic disk device 23. Upon receiving the notification, the bandwidth reservation application unit 2b of the server 11 changes the bandwidth management server of the magnetic disk device 23 in the bandwidth management server table to the server 12, and removes the server 13 from the server ordering (FIG. 9).

FIG. 16 shows the operation flow of the down detection unit 1b, the load distribution unit 7b, and the bandwidth reservation application unit 2b. FIG. 17 shows an operation flow of the band management takeover unit 4b.

According to the third embodiment, when the server 13 goes down, the load balancer 7b of the server 11 compares the load of the bandwidth management of the magnetic disk device in each server, and the load of the bandwidth management is minimized. Server (for example, server 1
2) is specified. The load distribution unit 7b of the server 11 instructs the bandwidth management takeover unit 4b of the server 12 to take over the bandwidth management from the down server 13. Thereby, the same effect as the effect obtained by the parallel video server system 10 of the first embodiment can be obtained.

(Embodiment 4) In Embodiment 4, the configuration and operation of the present invention will be described using a parallel video server system 40 as an example. However, the application of the present invention is not limited to the parallel video server system. The present invention can be applied to any resource sharing type parallel server system in which a plurality of servers share network resources. Where "network resources"
Is a network resource that can be accessed from a plurality of servers. Network resources are managed by one corresponding server among the plurality of servers.
The network resource is, for example, a network band. Resource management information for managing network resources is stored in a corresponding one of a plurality of storage devices connected to a plurality of servers via a network.

Each of the plurality of storage devices is, for example, a magnetic disk device. In the following description, it is assumed that each of the plurality of storage devices is a magnetic disk device. However, any storage device other than the magnetic disk device can be used as the storage device.

FIG. 18 shows a parallel video server system 40.
FIG. 3 is a block diagram showing the configuration of FIG. In the parallel video server system 40, the plurality of servers 11 to 13 share the bandwidth of the network 30 as resources of the network 30. The servers 11 to 13 are connected to a plurality of magnetic disk devices 21 to 24 via a network 30.

The servers 11 to 13 provide a video distribution service. The magnetic disk devices 21 to 24 store video data. The network 30 interconnects the servers 11 to 13 and the magnetic disk devices 21 to 24. The network 30 is, for example, an FC-AL (Fiber Ch).
Ann Arbitrated Loop).

In the example shown in FIG.
The bandwidth of 0 is managed by the server 11, and network bandwidth management information (hereinafter, sometimes referred to as ticket management information) for managing the bandwidth of the network 30 is stored in the magnetic disk device 21. I have. The servers 11 to 13 are cyclically ordered in the order of the server 11, the server 12, the server 13, and the server 11.

Each of the servers 11 to 13 accesses the magnetic disk devices 21 to 24 via the network 30 in response to a request from the user, and
To read the video data stored in.

FIG. 19 is a functional block diagram showing the configuration of the server 11. The configuration of the servers 12 and 13 is the server 1
1 is the same as that of FIG.

The server 11 has a server down detecting section 41
, A bandwidth reservation application unit 42, a bandwidth management unit 43, a bandwidth management handover unit 44, a user request processing unit 45, a video distribution unit 46, and a disk down detection unit 47.

The bandwidth reservation application section 42 sends a bandwidth reservation message to a server that manages the bandwidth of the network 30 when accessing the magnetic disk devices 21 to 24. The bandwidth reservation application unit 42 has bandwidth management server information and bandwidth management disk information. The bandwidth management server information is a server that manages the bandwidth of the network 30 (FIG. 18).
In the example shown in (1), information for specifying the server 11) is included. The bandwidth management disk information includes information for specifying a magnetic disk device (the magnetic disk device 21 in the example shown in FIG. 18) that stores network bandwidth management information for managing the bandwidth of the network 30. The bandwidth management server information and the disk information in the bandwidth reservation application unit 42
Updated in response to an update instruction.

The bandwidth management unit 43 discretely manages the bandwidth of the network 30 in the form of a ticket, receives a bandwidth reservation message from the server, and returns the ticket, if any. One ticket corresponds to a bandwidth of 30 Mbps, and a plurality of tickets are prepared according to the bandwidth of the network 30. The usage status of the ticket and the distribution status to the server are managed by the bandwidth management unit 43 as ticket management information. The ticket management information is written to the magnetic disk device each time it is updated.

The user request processing unit 45 receives a video distribution request from the user and instructs the bandwidth reservation application unit 42 to apply for a bandwidth reservation, and instructs the video distribution unit 46 to distribute a video if a ticket is obtained.

The video distribution unit 46 includes a user request processing unit 45
And distributes the video in response to the instruction from.

The server down detecting section 41 checks the operation status of the server by periodically transmitting / receiving a message to / from the server which is ordered immediately before, and the server goes down (normal stop or failure stop). Detect that it has done. If the downed server is a server that manages the bandwidth of the network 30, the server down detection unit 4
1 instructs the band management takeover unit 44 to take over the band management.

The band management takeover unit 44 receives the band management takeover instruction from the server down detection unit 41, takes over the band management of the network that was taken over by the server that went down, and sent the band reservation application unit 42 of another server. An instruction to update the bandwidth management server information is sent.

The disk down detecting section 47 detects that the magnetic disk device storing the ticket management information of the network 30 has gone down (normal stop or failure stop), and sends the ticket management information of the network 30 to another magnetic disk device. rewrite. Such rewriting is achieved, for example, by writing the ticket management information of the network 30 held in the memory of the server that manages the bandwidth of the network 30 to another magnetic disk device.

Hereinafter, an operation example of the parallel video server system 40 will be described.

When receiving the distribution request of the video title A from the user, the user request processing section 45 of the server 12 instructs the band reservation application section 42 to apply for a band reservation. The bandwidth reservation application unit 42 sends a bandwidth reservation message to the bandwidth management unit 43 of the server 11 that manages the bandwidth of the network 30. The bandwidth management unit 43 of the server 11
Check the ticket of 0, and if there is a ticket, send it to server 1
2 to the band reservation application unit 42 to update the ticket management information and write it to the magnetic disk device 21. When the ticket is obtained, the user request processing unit 45 sends a video distribution start instruction to the video distribution unit 46. If the ticket cannot be obtained, the user request processing unit 45 notifies the user that the service cannot be performed. Upon receiving the video distribution start instruction, the video distribution unit 46 periodically distributes the video data. as a result,
The video is played.

The server down detecting section 41 of each server
Is the down detection unit 41 of the server that is ordered immediately before
The message exchange is performed at predetermined time intervals (for example, 10 seconds) between the server and the server, and the server down due to the failure of the message exchange is detected. Down detection unit 4 of server 12
When the server 1 detects that the server 11 is down, the down detecting unit 41 of the server 12 transfers the band management takeover to the band management takeover unit 44 of the server 12 so as to take over the bandwidth management of the network 30 which was in charge of the down server 11. To instruct.

The bandwidth management handover unit 44 of the server 12
Upon receiving the instruction to take over the bandwidth management, the magnetic disk drive 21
, And updates the bandwidth management server information to take over the bandwidth management of the network 30. Thereafter, the server 11 is removed from the server ordering.

Further, it notifies the bandwidth reservation application section 42 of the server 13 that the server 11 has gone down and the server 12 has taken over the bandwidth management of the network 30. Server 1
In response to the notification, the third band reservation application unit 42 changes the band management server information.

Further, in this state, the disk down detecting section 47 of the server 12 which is in charge of the band management of the network detects whether the magnetic disk device 21 is down by checking whether the band management section 43 can access the disk. Disk down detector 4 of server 12
7 detects that the magnetic disk device 21 is down, the disk down detecting unit 47 of the server 12 updates the ticket management information by writing the ticket management information held in the memory of the server 12 to the magnetic disk device 22. . Further, the disk down detection unit 4 of the server 12
7 is a bandwidth reservation application unit 4 of the server 12 and the server 13
2 sends an instruction to update the bandwidth management disk information. FIG. 21 shows an operation flow of the disk down detection unit 47.

The bandwidth reservation application unit 42 of the server 12 or 13 updates the bandwidth management disk information in response to the update instruction (FIG. 20).

According to the fifth embodiment, the server down detecting section 41 of the server 12 detects that the server 11 is down and instructs the band management takeover section 44 to take over the band management of the network 30 from the server 11. This allows
Loss of the bandwidth management function of the network 30 due to the server 11 going down can be prevented.

When the disk down detecting unit 47 of the server 12 managing the bandwidth of the network 30 detects that the magnetic disk unit 21 storing the ticket management information has gone down, the ticket management information is transferred to the magnetic disk unit 22.
Rewrite to Thereby, even when the magnetic disk device 21 goes down, it is possible to take over the band management of the network 30 by stopping the server. As a result, the reliability of the parallel video server system 40 can be further improved.

[0092]

According to the present invention, the down detection unit detects that a specific server has gone down, and the resource management takeover unit takes over the management of resources managed by the particular server that went down. As described above, by dynamically moving the resource management function between a plurality of servers and coping with the server down, loss of the resource management function due to the server down can be prevented.

Further, the load distribution unit periodically or
When a server goes down or a server is added, by instructing a server with a large resource management load to a server with a small resource management load to take over the resource management,
The resource management load can be distributed to a plurality of servers.

Further, according to the present invention, the server down detecting section detects that a specific server has gone down, and if the specific server is a server in charge of network resource management, the resource management takeover section Takes over network resource management from the specific server that went down. This can prevent loss of the network resource management function due to the server going down.

When the disk down detecting unit of the server managing the network resources detects that the storage device storing the ticket management information has gone down,
The ticket management information is rewritten to another storage device. Thereby, even when the storage device goes down, the band management of the network 30 can be taken over by stopping the server. As a result, the reliability of the parallel server system can be further improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a parallel video server system 10 according to a first embodiment of the present invention.

FIG. 2 is a functional block diagram showing a configuration of a server 11 shown in FIG.

FIG. 3 is a diagram illustrating an example of a band management server table.

FIG. 4 is a parallel video server system 1 after the server goes down.
FIG. 3 is a block diagram showing a configuration of a 0.

FIG. 5 is a diagram illustrating an example of a bandwidth management server table after a server goes down.

FIG. 6 is a diagram illustrating an example of a bandwidth management server table after load distribution.

FIG. 7 is a diagram showing an operation flow of the load distribution unit 7 and the bandwidth reservation application unit 2.

FIG. 8 is a diagram showing an operation flow of the band management takeover unit 4.

FIG. 9 is a diagram illustrating an example of a bandwidth management server table after a server goes down.

FIG. 10 is a diagram showing an operation flow of the band management takeover unit 4;

FIG. 11 is a diagram showing an operation flow of the load distribution unit 7;

FIG. 12 is a functional block diagram illustrating a configuration of a server 11 according to Embodiment 2 of the present invention.

FIG. 13 is a parallel video server system 1 after adding a server.
FIG. 3 is a block diagram showing a configuration of a 0.

FIG. 14 is a diagram illustrating an example of a bandwidth management server table after adding a server;

FIG. 15 is a functional block diagram illustrating a configuration of a server 11 according to Embodiment 3 of the present invention.

FIG. 16 is a diagram showing an operation flow of a down detection unit 1b, a load distribution unit 7b, and a bandwidth reservation application unit 2b.

FIG. 17 is a diagram illustrating an operation flow of the band management takeover unit 4b.

FIG. 18 is a block diagram showing a configuration of a parallel video server system 40 according to Embodiment 4 of the present invention.

19 is a functional block diagram showing a configuration of a server 11 shown in FIG.

FIG. 20 is a block diagram showing a configuration of a parallel video server system 40 after server down and disk down.

FIG. 21 is a diagram showing an operation flow of a disk down detection unit 47.

FIG. 22 is a block diagram illustrating a configuration of a conventional disk-distributed parallel video server system.

FIG. 23 is a block diagram showing the configuration of a conventional disk-sharing type parallel video server system.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Down detection part 2 Band reservation application part 3 Band management part 4 Band management handover part 5 User request processing part 6 Video distribution part 7 Load distribution part 10 Parallel video server systems 11 to 13 Servers 21 to 24 Magnetic disk devices 30 Network 40 Parallel Video server system 41 Server down detection unit 42 Bandwidth reservation application unit 43 Bandwidth management unit 44 Bandwidth management takeover unit 45 User request processing unit 46 Video distribution unit 47 Disk down detection unit

Claims (6)

[Claims] 1. A parallel server system in which a plurality of servers share a plurality of resources via a network, wherein each of the plurality of resources is managed by a corresponding one of the plurality of servers. Each of the plurality of servers, a down detection unit that detects that a specific server among the plurality of servers has gone down, and a resource management takeover unit that takes over management of resources managed by the downed specific server A parallel server system comprising: a load distribution unit configured to distribute a management load of the plurality of resources. 2. The load distribution unit according to claim 1, wherein the first server having the largest resource management load and the second server having the least resource management load.
Server is specified at a fixed period, and when the difference between the resource management load of the first server and the resource management load of the second server is larger than a predetermined value, the resource management is performed from the first server. 2. The parallel server system according to claim 1, wherein an instruction is given to a resource management takeover unit of the second server to take over. 3. When the down detection unit detects that the specific server has gone down, the load distribution unit identifies a second server with a minimum resource management load, and identifies the downed server. 2. The parallel server system according to claim 1, wherein the parallel server system instructs the resource management takeover unit of the second server to take over resource management from the server. 4. When a new server including the down detection unit, the resource management takeover unit, and the load distribution unit is added to a parallel server system, the new server load distribution unit The first with the highest management load
2. The parallel server system according to claim 1, wherein a server is specified, and a resource management takeover unit of the new server is instructed to take over resource management from the first server. 5. The load management unit according to claim 1, wherein the load management unit controls the resource management takeover unit so as to take over the management of the resource managed by the downed specific server within a range in which a resource management load amount is smaller than a predetermined upper limit. Control,
The parallel server system according to claim 1. 6. A parallel server system in which a plurality of servers share network resources, wherein the plurality of servers are connected to a plurality of storage devices via the network, and wherein the resources of the network are Resource management information for managing resources of the network, which is managed by a corresponding one of the plurality of servers, is stored in a corresponding one of the plurality of storage devices. Each of the servers includes: a server down detection unit that detects that a specific server among the plurality of servers has gone down; and if the downed specific server manages resources of the network, A resource management takeover unit that takes over resource management, and stores the resource management information Have storage device detects that down, and a storage device down detection unit for rewriting the resource management information to another storage device,
Parallel server system.