JP2001067257A - Inter-distributed node exclusive updating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To an inter-distributed node exclusive updating device easily integrable, and usable a data region on-line simultaneously while making data coincident with each other among nodes to improve the usability. SOLUTION: This inter-distributed node exclusive updating device is provided on the nodes of a network so that the same shared data can be held on all nodes in the exclusive state. In this case, all nodes A-D for holding shared data 11 are used as objects, and the shared data are exclusively updated for every region of virtual regions c1-cn by an exclusive managing part 14.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an exclusive update apparatus between distributed nodes for exclusively updating common data between distributed nodes, and more particularly to an exclusive update between distributed nodes which can use a data area even during exclusive update. It relates to an updating device.

[0002]

2. Description of the Related Art Conventionally, a plurality of computers have been connected to duplicate (multiplex) a system or duplicate stored data.

[0003] A data update process for securing the same data in different computers is called an exclusive update process. When common data is used in different programs, exclusive update must be performed. Generally, an OS (Operating System) provides such a function in the same computer.

However, in order to perform exclusive update between a plurality of computers, it is necessary to make the entire computer system a mechanism dedicated to duplexing (multiplexing), and such a mechanism is incorporated into a general LAN or network system. It is difficult.

[0005] In addition, in the exclusive update processing, the common data between its own computer (own node) and another computer (other node) are made to coincide with each other, that is,
Until the software configuration control is completed, it is impossible to use the data area online.

[0006]

As described above, in the conventional exclusive update processing, it is difficult to incorporate the exclusive update processing into a general system, and the data area becomes unusable until the processing is completed. For this reason, the exclusive update process is actually in operation but cannot be used by the user, so that the usability (availability) seen from the user is reduced.

The present invention has been made in consideration of the above-mentioned circumstances, and can be easily incorporated. The data area can be used at the same time while matching data with other nodes, and the data area can be used online, so that the availability can be improved. An object of the present invention is to provide an exclusive update device between distributed nodes.

[0008]

According to a first aspect of the present invention, there is provided an exclusive update between distributed nodes for holding the same shared data on all nodes in an exclusive state provided on nodes of a network. An apparatus for exclusive distribution between distributed nodes, comprising: an exclusive management unit that exclusively updates the shared data for each virtual partial area for all nodes holding the shared data.

According to a second aspect of the present invention, there is provided an exclusive update apparatus between distributed nodes provided on a node of a network and holding the same shared data on all nodes in an exclusive state. An update management information storage unit in which each of the partial areas, which is obtained by virtually dividing the shared data, is set as a management target, and updated or unupdated is stored for each of the partial areas; An update processing unit that updates shared data of a corresponding part for each unupdated partial area in the update management information storage unit with the shared data of the node as an update target. is there.

[0010] Furthermore, an invention corresponding to claim 3 is a distributed inter-node exclusive update apparatus provided on a node of a network and for holding the same shared data on all nodes in an exclusive state. An update management information storage unit that manages a plurality of large areas indicating virtual division of the shared data and stores updated or unupdated data for each of the large areas; The shared data of the node is set as an update target, the shared data of a corresponding portion is updated for each unupdated large area in the update management information storage means, and when a certain large area is set as an update target, the exclusive When the state cannot be shifted to, the small area virtually divided from the large area is set as a management target of the update management information storage unit,
The exclusive inter-distribution node updating apparatus includes update processing means for updating shared data of a corresponding small part for each of the non-updated small part areas.

(Operation) Therefore, the invention corresponding to claim 1 takes the above-described means, so that the exclusive management means targets all nodes holding the shared data, and transfers the shared data to the virtual nodes. Since exclusive update is performed for each partial area, it can be easily incorporated. For the partial area, the data area can be used online while the data other than the partial area is matched while data is matched with other nodes. Therefore, availability can be improved.

According to a second aspect of the present invention, there is provided update management information in which each partial area indicating virtual division of shared data is to be managed and updated or not updated is stored for each partial area. A storage unit is provided, and the update processing unit updates shared data of all nodes in the exclusive state as update targets, and updates, for each unupdated partial area in the update management information storage unit, shared data of a corresponding portion. Therefore, the same operation as the operation according to the first aspect can be easily and reliably performed.

Further, according to a third aspect of the present invention, there is provided an update management system for managing a plurality of large areas which virtually divide shared data and indicating whether each large area has been updated or not updated. An information storage unit is provided, and the update processing unit sets shared data of all nodes in the exclusive state as an update target, and for each large area not updated in the update management information storage unit, updates the shared data of a corresponding portion. When a large area is updated and a transition to the exclusive state cannot be made when the large area is updated, a small area obtained by virtually dividing the large area is set as a management target of the update management information storage unit, and the unupdated small partial area is updated. Since the shared data of the corresponding small portion is updated every time, in addition to the effect similar to that of the second aspect, the update is executed in the large region, and the update is performed in the small region when the large region cannot be updated. From, share according to the size of the large area The exclusive update processing of the entire over data can be executed at high speed.

[0014]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a conceptual configuration diagram showing an example of an inter-distributed node exclusive update apparatus according to a first embodiment of the present invention. Exclusive update device 1 between distributed nodes of this embodiment
Is incorporated in the computer 3 (node A, node B,...) Which is a node of the network 2. As shown in the figure, when, for example, the application program 4 of the node A among the nodes A and B in the exclusive securing state outputs an exclusive update request, the exclusive updating device 1 notifies the fact via the network 2 and After the shared data 11 of the other node is updated by the exclusive update device 1 of the other node, the shared data 11 of the own node is updated. The shared data 11 is virtually divided into areas, and is exclusively updated for each area, so that data areas other than the area being exclusively updated can be used online.

Here, the network 2 may be a large-scale network system such as a WAN using a dedicated line or a public line, or a LAN such as an Ethernet. In the following description, a case where a LAN is used as the network 2 will be described.

FIG. 2 is a block diagram showing a configuration example of a computer in a LAN to which the exclusive update device between distributed nodes according to the present embodiment is applied.

As shown in FIG. 1, the network system includes nodes A, B, C, D,. . Calculator 3 as
A, 3B, 3C, 3D,. . It is composed of In the following description, when a node is not specified, the computer 3 and the exclusive update device 1 are indicated, and when a node is specified, the computer 3A and the exclusive update device 1A are indicated.

The computer 3 includes at least an application program 4 for realizing the main function of the node, an exclusive update device 1, a storage device 5, and a communication device 6 for communicating with another node via a network. I have.
Each of the devices 1, 4, 5, and 6 is a software unit such as data or a program and a hardware unit such as a memory or a CPU or a function realizing unit formed by combining these.

The storage device 5 includes storage means such as a main storage (memory) and a hard disk, and stores shared data 11 and non-shared data 12 that are shared by applications in an exclusive state. The shared data 11 is, for example, MM
It is saved as an AP or a file. MMAP is
It maps memory to a file, which looks the same as memory when viewed from a program, but the entity exists as a file.

When the application program 4 is in the exclusive state, the application program 4 stores it in the storage device 5 via the exclusive update device 1.
Access to the shared data 11 in the storage device 5 without passing through the exclusive update device 1 irrespective of whether the storage device 5 is in the exclusive state or not.
Access the non-shared data 12 in the. Therefore, the application program 4 cannot access the shared data 11 composed of MMAP or the like without passing through the exclusive update device 1. The means for accessing the non-shared data 12 is not shown.

The exclusive update device 1 includes an exclusive update request receiving unit 1
3, an exclusive management unit 14, a data management unit 15 that updates the shared data 11 for each area, and a failure detection unit 16. Here, first, the exclusive update request receiving unit 13
Accepts an exclusive update request from the (OS or) application program 4 and passes the request to the exclusive management unit 14. When an exclusive update request is issued from another node, the exclusive update request is notified to the application program 4. The failure detection unit 16 communicates with the other nodes at regular intervals, and detects a failure at the other node by being unable to communicate with the other nodes at regular intervals. Further, when a failure occurs in another node, the exclusive management unit 14 is notified that the failure has occurred.

The exclusion manager 14 communicates with another node and actually performs exclusion updating. Further, when a failure occurs in another node during communication for exclusion request processing, the exclusion manager 14 disconnects that node. here,
The exclusion update is divided into an exclusion securing phase, a distributed update phase, and an exclusion release phase.
The processing in each of these phases is performed.

FIG. 3 is a schematic diagram showing the detailed configuration of the exclusive management unit 14 and the relationship with other components. The exclusion management unit 14 includes an exclusion securing processing unit 21 and a distributed update processing unit 2
2, an exclusive release processing unit 23, a management information holding unit 24,
A timer 25 is provided.

According to the request from the exclusion update request receiving unit 13, the exclusion securing processing unit 21 communicates with the exclusion management unit 14 of another node to which data is to be replicated via the communication device 6, and performs exclusion in the other node. To secure. Further, the timer 25 is started, and when a packet to another node times out, processing such as retransmission is performed. In addition,
When an exclusion securing request is received from another node, the exclusion securing process is executed, and a response to the effect of securing exclusion is returned to the other node.

The exclusion securing processing unit 21 determines which storage area shared data 11 in the storage device 5 when securing exclusion between distributed nodes, whether it is based on a request from its own node or a request from another node. Management information such as whether to store the information is created and stored in the management information holding unit 24.

When a request for updating shared node shared data is received via the exclusive update request receiving unit 13 after the exclusion is successfully completed by the exclusion securing processing unit 21, the distributed update processing unit 22 updates the update data together with the request. To other nodes.
After the update process in the other node is confirmed by the response, the shared data 11 is updated via the data management unit 15 and the management information in the management information holding unit 24 is corrected. When a distributed update request is received from another node, the management information in the management information holding unit 24 is modified while
The shared data 11 is updated, and a response is returned to the other node.
Distributed updating is realized by these functions.

The distributed updating process is performed by the timer 25
Is used to perform time management for retransmission and the like.

When the exclusion update is completed, the exclusion release processing unit 23 communicates with another node to release the exclusion, and notifies the exclusion update request reception unit 13 of the exclusion completion. In addition, an exclusive release request is executed from another node, and a response is returned. When the exclusion fails in the middle, the exclusive update request receiving unit 13 is notified of the exclusion update failure. Further, with the release of the exclusion, the management information in the management information holding unit 24 is changed to information indicating the release of the exclusion. The exclusion release processing unit 23 uses the timer 25 to manage time for retransmission and the like.

The management information holding unit 24 can identify the shared data 11 in the storage device 5 while the own node is in the exclusive state, and also stores management information 26 such as a memory / file address for management and shared data. Update management bitmap table 27 for exclusively updating 11 for each virtual area
And hold.

The management information 26 corresponds to each shared portion stored as shared data.
Specific information (each shared part) in the storage device 5 is shared between the distributed nodes and is identified as being in an exclusive state.

The update management bitmap table 27 is
As schematically shown in FIG. 4, bitmap areas b1 to b corresponding to virtual areas c1 to cn of shared data 11 are provided.
n, all areas b1 to bn are cleared at the start of the exclusive update, and each time an area ci (i is an arbitrary number) of the shared data 11 is updated by the distributed update processing unit 22, the area is updated. bitmap area bi (i
(The same number as i of ci) is checked (the updated flag is set), and at the end of the exclusive update, all the bitmap areas b1 to bn are eventually checked. In other words, the update management bitmap table 27
Indicates, for each of the bitmap areas b1 to bn, whether the shared data 11 of the corresponding area c1 to cn has been updated or not updated, based on the presence or absence of the updated flag.

That is, the exclusion manager 14 sets each area ci
By updating the shared data 11 each time, the shared data areas c1 to ci-1 and ci + 1 to c other than the area ci being updated are updated.
n can be used online.

Next, the operation of the exclusive update apparatus between distributed nodes according to the present embodiment configured as described above will be described.

In the present embodiment, there are at least four node nodes A, B, C and D as the computer 3 connected to the network 2, and the one area ci of the shared data 11 is replicated among these four nodes. Is implemented.

As described above, the exclusive update is divided into an exclusive securing phase, a distributed update phase, and an exclusive release phase. The communication protocol sequence in each phase will be described. The areas other than the one area ci can be used online even during the exclusive update. In addition, a series of exclusive updates of one area ci described below are repeatedly executed until the update of all the areas c1 to cn of the shared data 11 is completed, as shown in FIG.
The following description is based on the flowchart of FIG. 5 and FIGS.
This is performed according to both the sequence diagram of FIG.

FIG. 6 is a diagram showing a basic process according to the exclusion securing protocol. The exclusion securing is mainly realized by the exclusion securing processing unit 21 of the exclusion management unit 14. First, after the update management bitmap table 27 is cleared at the start (ST
1), exclusion securing processing unit 2 in each node (computer 3)
1 executes basically the following sequence of securing exclusion.

First, the exclusion securing request node transmits an exclusion securing request to another node (s1: ST2).

Next, OK or NG is responded to the exclusion securing node from the exclusion securing request receiving node (s2). FIG. 6 shows a case where all OKs are returned (ST
3: OK).

Then, the exclusion permission response (O
Upon receiving K), the exclusion securing request node transmits a success (exclusion start) / failure protocol to all nodes (s3). When an NG is received, an exclusion failure is transmitted (ST3: NG). In this case, if any node fails in exclusion because the area ci of the shared data 11 is being used (ST4: Y), the next area ci + 1
(ST5) and returns to step ST2, and if the failure is due to packet loss or exclusion conflict described later (ST5).
4: N), a retry is performed for the same area ci.

This protocol has the advantage that all nodes can clearly recognize which node is currently exclusive. The exclusive control is performed by the exclusive management unit 14, and the failure detection is separately performed by the failure detection unit 16. Since the two are distinguished from each other, it is only necessary to perform a retransmission process for packet loss. Good. In addition, failure detection
Regardless of each phase of exclusion securing, distributed update, and exclusion release, the following is continuously performed by the failure detection unit 16, and when a failure is detected, the fact is notified to the exclusion management unit 14, and the exclusion management unit 14 generates a failure. Processing for separating the node is executed.

Next, FIG. 7 shows a protocol when a packet is lost. FIG. 7 is a diagram showing processing by an exclusion securing protocol when a packet is lost. As shown in the figure, when the exclusion securing requester (node A) transmits the exclusion securing request protocol (t1), the timer 25 for detecting a timeout is started. For example, as shown in FIG. 7, the exclusion securing request is lost (transmitted to the node B),
When either the response to the exclusion securing request (response from the node D) is lost, packet loss is detected by this timeout (t2), and the exclusion securing request is retransmitted (t3). This is repeated until exclusion securing responses are received from all nodes.

After receiving the exclusion securing response from all nodes (t4), the exclusion securing requesting node (node A) transmits an exclusion start declaration (t5). If this packet is lost, the response node detects a timeout (t6).

For example, as shown in FIG. 7, when the exclusion start declaration to the node C is lost, the timeout detection timer 25 is started in the node C from the time of receiving the exclusion securing request. A response is transmitted (t7).

Since the exclusion securing request node (node A) has already started exclusion, an exclusion start declaration is simply retransmitted for this packet. That is, an exclusion start request is transmitted to the node C (t8). In the above case, the exclusion start time differs among the four nodes in the protocol, but no problem occurs due to the following two points. The purpose of securing exclusion is distributed update, and there is no problem if exclusion is completed before distributed update occurs. During this protocol, no exclusion request is issued from another (Once a response to the exclusion reservation request from another node is OK, the exclusion reservation request is not issued from the own node). Next, a case in which a conflict between exclusive requests occurs will be described with reference to FIG. FIG. 8 is a diagram showing processing by an exclusion securing protocol when an exclusion conflict occurs. As shown in the figure, when an exclusion securing request is issued from the nodes A and D (u1, v1), OK is responded to the exclusion securing request first received from the responding node, and the exclusion securing received later is received. NG is responded to the request. In the example of FIG. 8, the node A is NG from the node D, and is O from the node B / C.
K is received (u2). Since the OK response could not be received from all the nodes, the exclusive exclusion failure is notified from the node A to all the nodes (u3). On the other hand, since the node D receives NG from all the nodes (v2), the node D also notifies all nodes of the failure in securing the exclusion (v3).

The nodes A and D sleep for a random time, and after the lapse of the time, the exclusion securing request is transmitted from the nodes A and D again (u4, v4).

When an exclusion conflict occurs, retry is performed until an OK response to the exclusion securing request is received from all nodes. Further, once OK responses are received from all nodes, exclusion always succeeds unless a node failure of the exclusion requesting node occurs.

2) Distributed Update Protocol Next, the distributed update protocol will be described. Distributed updates are
This is mainly realized by the distributed update processing unit 22 of the exclusive management unit 14. In the distributed update protocol, since all nodes have already recognized the exclusive node, only a response to the update need be returned. Further, since no contention occurs, if a packet is lost, it is sufficient to retry the update request node.

FIG. 9 is a diagram showing processing according to the distributed update protocol. In the distributed update request node (node A), a time-out timer is started (w2) at the time when the update request is transmitted (w1: ST6). This update request is transmitted together with the update content information to the node in the exclusive state when the application program 4 requests the exclusive update device 1 to change the content of the shared data 4 (distributed update).

As shown in FIG. 9, when a request packet to the node C and a response packet from the node B are lost, an update request is retransmitted after a timeout is detected (w
3). The response is returned as it is from node C, and node B
Has already completed the update, but returns a response again.
From the distributed update request node, this is repeated until responses are received from all nodes. When responses are received from all nodes, one area ci of the shared data 11 of the distributed update request node
Is updated to complete the update (w4), and the corresponding area bn of the update management bitmap table 27 is updated.
Is checked (ST10).

3) Exclusive Release Protocol Finally, the exclusive release protocol sequence will be described.

The exclusion release is realized mainly by the exclusion release processing unit 22 of the exclusion management unit 14. FIG. 10 is a diagram showing processing according to the exclusive release protocol. In the exclusion release protocol, processing similar to that in the distributed update protocol is performed.

That is, first, a release request is transmitted from the exclusive release request node (node A) (x1: ST8), and an OK response is returned from each node. Note that the node receiving the release request releases the exclusion. On the other hand, when a request packet to the node C or a response packet from the node B is lost, an exclusive release request is retransmitted (x3) after a timeout is detected (x2). The node C releases the exclusion and returns a response. The node B has already completed the exclusion release, but returns a response again. This is repeated from the exclusive release request node until responses are received from all nodes. When responses are received from all nodes, the exclusive release request node itself is exclusively released and exclusive release is completed (x
4: ST9).

In the case of FIG. 10, the next exclusion request may be issued from, for example, the node D which has already been released. However, this request does not succeed because the node A does not release the exclusion until receiving a response indicating completion of exclusion from all nodes. Therefore, there is no problem due to a time lag of release completion between the nodes.

As described above, one area ci of the shared data 11
, The exclusive update process is completed. Note that such exclusion securing processing, distributed update processing, and exclusion release processing are continuously executed until all the areas b1 to bn of the update management bitmap table 27 are checked (ST1).
0). For example, even if there is an area ck skipped because exclusive use cannot be ensured because it is in use at the time of requesting exclusive control, the corresponding bitmap area ck has not been checked, so
An exclusive update is performed.

That is, the exclusive update process is performed for the shared data 1
The processing is continuously executed until the update is completed in all the areas c1 to cn.

Next, an application example of the exclusive update apparatus between distributed nodes according to the present embodiment will be described.

FIG. 11 is a schematic diagram showing an application example of the inter-distributed node exclusive update apparatus of the present embodiment. In FIG.
The figure shows a case where an exclusive update device between distributed nodes is applied to a long-distance data backup system.

As shown in the figure, if the exclusive update device between distributed nodes according to the present embodiment is incorporated in a data backup system, even if each file system is far away, the data of both systems can be used using a general network. Replication becomes possible. Therefore, even if an earthquake such as the Great Hanshin Earthquake causes data loss in the Osaka system, necessary information is retained by the Tokyo system.

FIG. 11B shows an example in which the inter-distributed node exclusive update device is applied to a distributed processing system.

When a program requiring a large amount of data processing is executed, for example, partial programs (applications A and B) are distributed and mounted on nodes A and B,
Processing can be performed in parallel using each computer resource. As shown in the figure, by constructing a distributed processing system via the inter-distributed node exclusive update device 1 of the present embodiment, a kind of parallel processing computer can be simply and reliably configured on a network.

As described above, according to the present embodiment, the exclusion manager 14 targets all nodes A to D holding the shared data 11 and stores the shared data 11 in the virtual area c1.
Since exclusive update is performed every ~ cn, it can be easily incorporated,
Regarding the area ci, the data is matched with the other nodes, and at the same time, the areas c1 to ci-1, cc other than the area ci are set.
For i + 1 to cn, the data area can be used online, so that the availability can be improved.

More specifically, each area b1 to bn, which is a virtual division of the shared data 11, is managed, and each area b1
, An update management bitmap table 27 in which updated or unupdated data is stored for each.
Updates the shared data 11 of the corresponding area ci for each unupdated area bi in the update management bitmap table 27, with the shared data 11 of all nodes in the exclusive state as update targets. The effect described above can be easily and reliably achieved.

This embodiment has the advantages described above.
The following effects are obtained. That is, when the plurality of application programs 4 on different nodes share data and update the shared data 11, the exclusive update is received via the exclusive update request receiving unit 13, and the exclusive management unit 14 communicates with another node. Data sharing and changing the shared data after the response, it is possible to reliably ensure exclusion for all nodes that are to be placed in an exclusive state, and to replicate data replicated between distributed nodes. It is possible to freely access and execute exclusive update at each node while ensuring that the nodes are updated.

Here, at the time of exclusion securing, the exclusion securing processing unit 21 requests one or more designated nodes via the network to shift to the exclusive state. A response such as an exclusion permission response (OK) or a rejection notification (NG) is returned from each node in response to the exclusion securing request. When the exclusion permission response is received, the node has already shifted to the exclusion state. Is finished or
Or the transition is in a definite state.

Then, when the exclusion permission response (OK) is received from all nodes that have transmitted the exclusion securing request, the exclusion securing processing unit 21 causes the own node to shift to the exclusive state.

As described above, since the self-node is shifted to the exclusive state after confirming the exclusive shift of the other node, it is possible to reliably ensure the exclusion of all the nodes to be set to the exclusive state.

Next, in the distributed update, the designated 1
When the above nodes and the own node are in the exclusive state,
An update request for shared data is accepted.

The distributed update processing unit 22 attaches the update data received by the exclusive update request reception unit 13 and updates the shared data for each area ci to one or more designated nodes via the network. Is required.
In response to this update request, the same shared data is updated by the update data for each area ci in each node, and a response to the request is returned after the update or after the update is in a definite state. Will be.

When the update notice of the shared data is received from all the nodes that have transmitted the update request, the distributed update processing unit 22 and the data management unit 15 update the shared data in the own node for each area ci. .

As described above, since the shared data 11 of the own node is updated after confirming the update processing at the other node, it is ensured that the data replicated between the distributed nodes is replicated. However, exclusive update can be freely and reliably performed at each node.

When there is a node that does not respond after a lapse of a predetermined time from the transmission of the exclusion securing request, the exclusion securing processing unit 21 transmits the exclusion securing request to the node again. Even when a packet such as an exclusion securing request or an exclusion permission response is lost, a re-request is transmitted, and more secure exclusion transfer can be secured.

When the exclusive exclusion unit 21 receives a transfer rejection notification from any of the nodes that transmitted the exclusion securing request, the exclusion securing unit 21 waits for a randomly set time to elapse.
Since the exclusion securing request is transmitted to one or more designated nodes again, even if a migration request is sent from each of the distributed node exclusion update devices in a plurality of nodes on the same network, exclusion conflict occurs. After the time set at random has elapsed, the transfer request is transmitted again to one or more designated nodes, so that the race condition can be released.

When there is a node that does not respond even after a predetermined time has passed since the transmission of the update request, the distributed update processing unit 22 transmits an update request to which the update data is attached again to the node. Therefore, even when a packet such as an update request or an update completion notification is lost on the network, a re-request is transmitted, and a more secure exclusive transfer can be ensured.

(Second Embodiment) Next, the second embodiment of the present invention will be described with reference to the drawings.
Here, different parts will be mainly described.

That is, the present embodiment is a modification of the first embodiment, and aims at speeding up the exclusive update process. Specifically, as shown in FIG. Instead of the bitmap table 27, for example, four areas b1 to b of the update management bitmap table 27
The management information holding unit 26 includes a large area bitmap table 28 having areas G1, G2,..., Gn / 4 having sizes corresponding to 4, b5 to b8,.

Here, in addition to the functions described above, the exclusive management unit 14 has a function of clearing the large area bitmap table 28 at the start of the exclusive update, and a large area bitmap table ., And the area Gk for which exclusion was not secured in the second cycle, this area Gk is replaced with, for example, four small areas gk1 to gk4 as shown in FIG. And the exclusive update of these small areas gk1 to gk4 is sequentially performed.

Next, the operation of the inter-distributed node exclusive update device having the above configuration will be described with reference to the flowchart of FIG. Now, the exclusion management apparatus 14 uses the large area bitmap table 28 (ST1a) and executes exclusive update for each large area Gi in the same manner as described above (ST2 to ST2).
ST6) After the update is completed, the area Gi is checked (ST7a).

Here, for example, when the exclusive securing of the area Gk fails, the area Gk is ignored as described above, and the next area Gk +
With respect to 1, the processes from securing exclusion to updating and exclusion release of the corresponding area Ck + 1 of the shared data 11 are executed (ST2).
To ST11). Similarly, in the subsequent areas Gk + 2 to Gn / 4, the corresponding areas Ck +
Data update and exclusion release are executed from exclusion securing of 1 to Cn / 4.

In this embodiment, the following processing is performed when there is an unchecked area Gk even after the second cycle. That is, in the second round, if the attempt to secure the exclusion of the area Gk again fails, the area Gk is divided into four small areas gk1 to gk4, as shown in FIG.
To gk4 are targeted for exclusive update processing (ST12). Thereafter, similarly, for the small area gk1, data update and exclusive release are performed from the exclusion securing of the small area ck1 of the corresponding shared data. Similarly, for the subsequent areas gk2 to gk4, data update and exclusion release are performed individually from securing exclusion of the corresponding areas Ck2 to Ck4 of the shared data 11.

Thus, the number of exclusive updates between nodes can be reduced, and the exclusive update process can be speeded up. In this case, one area G corresponding to the four areas b1 to b4
Since 1 is provided, the exclusive update process can be completed in about 1/4 time compared to the first embodiment.

As described above, according to the present embodiment, a plurality of large areas G 1 to G
Gn / 4 is to be managed, and a large area bitmap table 28 in which updated or not updated is stored for each of the large areas G1 to Gn / 4 is provided. Is set as an update target, and for each unupdated large area Gi in the large area bitmap table 28,
The shared data 11 of the corresponding area ci is stored in the data management unit 15
When a large area Gk cannot be shifted to an exclusive state when the large area Gk is to be updated, the small areas gk1 to gk4 obtained by virtually dividing the large area Gk are set as management targets of the large area bitmap table 28. , The unupdated small area gk
Since the shared data 11 of the corresponding areas ck1 to ck4 is updated every 1 to gk4, in addition to the effect of the first embodiment, the update is executed in the large area, and when the update is not possible in the large area Gk, the small area is updated. Since the update is performed with gk1 to gk4, the exclusive update process of the entire shared data can be executed at high speed according to the size of the large area.

The present invention is not limited to the above embodiments, but can be variously modified without departing from the gist thereof. For example, in each of the above embodiments, the bitmap tables 27 and 2 are used as elemental technologies for managing updated or unupdated for each area of the shared data 11.
Although 8 is used, the present invention is not limited to this. For example, updated or unupdated data may be managed for each part (each file or the like) of the shared data 11 using a list of files or the like.

The method described in the embodiment can be implemented by a computer (computer) as a program (software means) such as a magnetic disk (floppy disk, hard disk, etc.), an optical disk (CD-ROM, DVD, etc.). ), Stored in a storage medium such as a semiconductor memory, or transmitted and distributed via a communication medium. The programs stored on the medium side include
A software program (including not only an execution program but also a table and a data structure) to be executed by the computer includes a setting program for configuring the computer. A computer that realizes the present apparatus reads a program recorded in a storage medium, and in some cases, constructs software means by using a setting program, and executes the above-described processing by controlling the operation of the software means.

In addition, the present invention can be variously modified and implemented without departing from the gist thereof.

[0085]

As described above, according to the present invention, a distributed node which can be easily incorporated, can use a data area online at the same time while matching data with another node, and can improve availability. An exclusive update device can be provided.

[Brief description of the drawings]

FIG. 1 is a conceptual configuration diagram illustrating an example of an inter-distributed node exclusive update device according to an embodiment of the present invention.

FIG. 2 is an exemplary block diagram illustrating a configuration example of a computer in a LAN to which the distributed inter-node exclusive update device according to the embodiment is applied.

FIG. 3 is an exemplary diagram showing a detailed configuration of an exclusive management unit and a relationship with other components according to the embodiment;

FIG. 4 is a schematic diagram showing the configuration of an update management bitmap table according to the embodiment;

FIG. 5 is a flowchart for explaining the operation in the embodiment;

FIG. 6 is an exemplary sequence diagram showing basic processing by an exclusion securing protocol in the embodiment;

FIG. 7 is an exemplary sequence diagram showing processing by an exclusion securing protocol when a packet is lost according to the embodiment;

FIG. 8 is an exemplary sequence diagram illustrating processing by an exclusion securing protocol when an exclusion conflict occurs in the embodiment;

FIG. 9 is an exemplary sequence diagram illustrating processing by the distributed update protocol according to the embodiment;

FIG. 10 is an exemplary sequence diagram illustrating processing by an exclusive release protocol according to the embodiment;

FIG. 11 is a schematic diagram showing an application example of the inter-distributed node exclusive update device in the embodiment.

FIG. 12 is a schematic diagram showing a configuration of a large area bitmap table applied to an inter-distributed node exclusive update device according to the second embodiment of the present invention;

FIG. 13 is a schematic diagram of a large area bitmap table for explaining a function of an exclusive management unit in the embodiment.

FIG. 14 is a schematic diagram for explaining the operation in the embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Exclusive update apparatus between distributed nodes 2 ... Network 3 ... Computer 4 ... Application program 5 ... Storage device 6 ... Communication device 11 ... Shared data 12 ... Non-shared data 13 ... Exclusive update request reception part 14 ... Exclusive management part 15 ... Data Management unit 16: Failure detection unit 21: Exclusive securing processing unit 22: Distributed update processing unit 23: Exclusion release processing unit 24: Management information holding unit 25: Timer 26: Management information 27: Update management bitmap table 28: Large area Bitmap table

Claims (3)

[Claims] 1. An exclusive update apparatus between distributed nodes for storing the same shared data on all nodes in an exclusive state provided on nodes of a network, wherein all of the shared data holding An exclusive update apparatus between distributed nodes, characterized in that the exclusive update means is provided for exclusively updating the shared data for each virtual partial area for nodes. 2. An inter-distribution exclusive update device provided on a node of a network and holding the same shared data on all nodes in an exclusive state, wherein the shared data is virtually divided. Update management information storage means in which updated or not updated is stored for each of the partial areas, and the shared data of all nodes in the exclusive state are to be updated, Update processing means for updating, for each unupdated partial area in the update management information storage means, shared data of a corresponding part, the inter-distributed node exclusive update apparatus. 3. An exclusive update device between distributed nodes, provided on a node of a network and for holding the same shared data on all nodes in an exclusive state, wherein the shared data is virtually divided A plurality of large areas shown as management targets, an update management information storage unit in which updated or unupdated is stored for each large area, and shared data of all nodes in the exclusive state are set as update targets. In each of the large areas that have not been updated in the update management information storage means, the corresponding part of the shared data is updated, and when a certain large area is to be updated, the large area cannot be shifted to the exclusive state. And updating processing means for updating the shared data of the corresponding small part for each of the non-updated small part areas, with the small area virtually divided into the management target of the update management information storage means. Features and Distributed nodes between the exclusive update apparatus that.