CN101166108B

CN101166108B - Method for a distributed task dispatching soft bus with dynamic scalability

Info

Publication number: CN101166108B
Application number: CN2006100978151A
Authority: CN
Inventors: 于耀; 周余; 都思丹
Original assignee: Nanjing University
Current assignee: Nanjing University
Priority date: 2006-11-15
Filing date: 2006-11-15
Publication date: 2011-06-01
Anticipated expiration: 2026-11-15
Also published as: CN101166108A

Abstract

This invention relates to a method for dispatching soft bus with dynamic expansibility and distributed task including: setting up SRB, IRB and FRB in a distributed network, in which, SRB manages all IRB and FRB, and service nodes carry out data interaction with the network center by modes of service-request-response, state and event-trap, which can manage data and information of device nodes dynamically in a distributed network.

Description

A kind of distributed task dispatching flexible bus method that has dynamic scalability

Technical field

The present invention relates to service-oriented remote control field, this method for designing can increase equipment and dynamically increase and decrease ability.For possessing the system that network management is supported, the functional node that each device node of network is carried sometimes or each subset often require to possess certain allocative abilities on function, therefore for the increase and decrease of single equipment functional node, need accomplish the dynamically ability of increase and decrease.Specifically, the present invention proposes the method for a kind of " flexible bus SoftSBus ", to overcome the above problems.

Background technology

The introducing of dynamic scalability certainly will bring two technological difficulties, and the one, the stability of entire system and equipment itself; The 2nd, the inconsistent region dynamic of bringing a certain function of each functions of the equipments, this has introduced again for possessing weak coupling between function service providers and the requestor.

In the in the industry cycle existing distributed middleware, processing modes such as SunRPC, Door need service providers and caller to interrelate with some similar mode usually, to realize calling of service, see Fig. 1.

Because more so intrinsic dependences, in case service itself by a certain node, is transferred to another node, its corresponding realization also must be made change, as Fig. 2.And a kind of like this demand of serving change is again quotidian in containing the distributed apparatus of network management.And more serious problem is, if in a distributed system that only has a small number of nodes, though change code or configuration remains inevitable situation, such change is not to stand; But in a system that has up to a hundred nodes, such change will make system be tending towards paralysis.

In the description of dynamic scalability, mentioned the introducing of dynamic, might cause influence to stability.Really, the proposition of flexible bus can not bring the basic lifting of stability on bottom, and still, it has improved the exploitation feasibility that stability is guaranteed greatly and has safeguarded unicity.

Flexible bus has been introduced uniformity at least aspect two, and improves the assurance to software quality:

1) the demand focus of concentrating stability to be proposed.

2) unified other demand except that service-request pattern.

In the assurance for stability, because the flexible bus dependence that each module is mutual, transfer on the dependence of each module and flexible bus, for on arbitrary node, increasing and decreasing function, only need to file an application to flexible bus, and the request server, also only need flexible bus is filed a request, thus, if the operation that flexible bus can be stable, the operation of each module on whole distributed system, just do not have any influence, therefore the stability of a system also by being dispersed in the requirement of each service with request, is summed up in the point that the unitary request to flexible bus.

Summary of the invention

1. flexible bus, this is a kind of distributed task dispatching middleware that has dynamic scalability, in object-oriented design patterns, contain a pattern that is called the tertium quid, its effect is the dependence that will have each other between the dependent object, shift the dependence for each object and tertium quid's object, and contain the object of coupling originally, mutual dependence is removed thus.

The proposition of the distributed middleware of flexible bus, just in order to remove the dependence between each device node and way to manage originally, and take dependence mode as Fig. 3, in a kind of like this mode, ISP's geographical attribute is transparent for caller, vice versa, therefore, any change of any one node is for whole system and various service, requestor as software upgrading, state are uploaded etc. has nothing to do.

For network management system, remove outside the service request pattern of issuing an order such network equipment node and each nodal function, also need some extra demands, as the condition managing pattern---safeguard the quantity of state of each equipment subtree, the event response pattern---when a certain subset state point exceeds the appointment thresholding, its superior node must be circulated a notice of by the time, to make corresponding countermeasure.In the design, subset sends incident to superior node, and superior node can pass through the trap capturing events, and the same unification of this data flow is finished by flexible bus, as Fig. 4.

Fig. 4 has provided following several quasi-modes that flexible bus is finished in the native system:

Service-request-response modes, acting on behalf of SRB or equipment operation phase by the upper level device node server runtime acts on behalf of IRB and acts on behalf of FRB to subtree IRB and function runtime and propose, corresponding response is returned on the way, when being served by request, the path dynamically sets up route, masked the ability of subtree node, promptly under distributed way, added the sense of hierarchy to the superior node request.

State model is sent to higher level IRB or SRB by subtree IRB, so that the upper management node is possessed the state information to subtree.

Incident-trap pattern, subtree IRB or FRB send incident to superior node, at source node all nodes to the path of SRB, all can tackle incident with trapping mode.

Simultaneously, uniformity can also avoid different management mode for the taking of bottom channel, thereby reduces the complexity that network etc. is safeguarded.

2. differentiated control

For big-and-middle-sized network management project, the classification type of management is quite necessary, and each subtree is by each minute station administration, and root node then can manage maintenance to the entire equipment tree, and therefore, the demand to differentiated control has also been considered in the design of native system.

In flexible bus, IRB need not be directly connected to the SRB node, and allows to be connected to the IRB node, and thus, the latter promptly becomes the former subtree root node, and the equipment with the SRB place is the same thus, can its subtree node of identical management.As Fig. 5, IRB1 contains two sub-tree nodes and self totally three nodes, and IRB4 only contains himself, and SRB then manages four sub-device nodes of FRB 1-4.

The subtree of IRB1 representative in system, can be single manage, its effect is identical with SRB management, and is only different on the interstitial content of management.

Description of drawings

Fig. 1 is the dependence that Door, SunRPC distributed are handled

Fig. 2 is that service A is transferred to after the node 2 by node 1, the schematic diagram that corresponding realization also must change thereupon

Fig. 3 software bus region transparency key diagram: provide after the node change of service, requesting node need not to do any change.

Each network management framework under Fig. 4 the inventive method and relevant flow graph

Fig. 5 SoftSBus hierarchical management mechanism

Embodiment

Each network management framework of Fig. 4 the inventive method and relevant flow graph.The following describes a specific implementation process of the present invention:

1. the runtime server runtime is acted on behalf of SRB on central server, but act on behalf of IRB and the function runtime is acted on behalf of FRB by each equipment operation phase of SRB unified management, when the geographical position of IRB and FRB changed, service end was acted on behalf of SRB and can be managed it automatically, did not need manual maintenance.

2. each branch center server moves IRB, and IRB initiates register requirement and obtains major device number from trend SRB, and accepts the register requirement of IRB of subordinate and FRB simultaneously, makes that IRB can unified management all IRB of subordinate and FRB.When the physical location of server operation IRB changes, can re-register automatically to higher level IRB or SRB, IRB of subordinate and FRB also can register automatically from this IRB of trend.

3. the service on each equipment, also can re-register service and obtain new dynamic Service number to the registration service of whole piece software bus and automatically obtain dynamic Service number by FRB when the physical location of FRB changes.

When central server need be when node proposes service request, propose to subtree IRB and FRB, IRB at different levels propose to next stage subtree IRB and FRB, corresponding response is returned on the way, when being served by request, the path dynamically sets up route, masked the ability of subtree node, promptly under distributed way, added the sense of hierarchy to the superior node request.Service requester on each equipment is also served to the software bus request by FRB, and software bus returns corresponding response.This pattern is called service-request-response modes.

5. central server need obtain the running status of each node, and implementation method is by the automatic maintenance state of IRB pond, and corresponding IRB is according to the IRB runtime parameter, and the timed sending state flow is given higher level IRB, and first order IRB sends to SRB.Central server just obtains the running status of each node like this, and this pattern is called state model.

6. subtree IRB or FRB send incident to superior node, at source node all nodes to the path of SRB, all with trapping mode interception incident, each FRB node can be submitted the incident trap to the higher level, in case corresponding event occurs, this FRB enters incident trap responder.This pattern is called incident one trap pattern.

Claims

1. distributed task dispatching flexible bus method that has dynamic scalability, it comprises following steps:

1) the runtime server runtime is acted on behalf of SRB on central server, but acts on behalf of IRB and the function runtime is acted on behalf of FRB by each equipment operation phase of SRB unified management;

2) each branch center server operation IRB, IRB initiates register requirement and obtains major device number from trend SRB, accepts the register requirement of IRB of subordinate and FRB simultaneously, makes that IRB can unified management all IRB of subordinate and FRB;

3) service on each equipment by FRB to the registration service of whole piece software bus and automatically obtain dynamic Service number;

4) when central server need be when node proposes service request, propose to subtree IRB and FRB, IRB at different levels propose to next stage subtree IRB and FRB, corresponding response is returned on the way, dynamically set up route when the path is served by request, masked the ability of subtree node, promptly under distributed way, added the sense of hierarchy to the superior node request, service requester on each equipment is also served to the software bus request by FRB, and software bus returns corresponding response;

5) central server need obtain the running status of each node, implementation method is by the automatic maintenance state of IRB pond, and corresponding IRB is according to the IRB runtime parameter, and the timed sending state flow is given higher level IRB, first order IRB sends to SRB, and central server just obtains the running status of each node like this;

6) subtree IRB or FRB send incident to superior node, at source node all nodes to the path of SRB, all can tackle incident with trapping mode, each FRB node can be submitted the incident trap to the higher level, in case corresponding event occurs, this FRB enters incident trap responder.

2. a kind of distributed task dispatching flexible bus method that has dynamic scalability according to claim 1 is characterized in that: when the physics region of IRB changes, can re-register and obtain major device number automatically in the step (2), not need manual intervention.

3. a kind of distributed task dispatching flexible bus method that has dynamic scalability according to claim 1 is characterized in that: when the physics region of FRB changes, can re-register and obtain major device number automatically in the step (3), not need manual intervention.

4. a kind of distributed task dispatching flexible bus method that has dynamic scalability according to claim 1 is characterized in that: step (4), (5), (6) can be passed through service-request-response modes, state model and incident-trap pattern respectively and realize the data interaction between each node and central server in the distributed network.