JPH06187326A - Method for communication network line allocation by decentralization restriction sufficiency - Google Patents

Abstract

PURPOSE:To decentralize a load and evade the down state of a system in the case of fault occurrence when communication resources are allocated at plural line allocation requests which are generated at the same time. CONSTITUTION:When plural line allocation requests for a line in a communication network 35 are made at the same time, the respective line allocation requests are formulated as decentralization restriction sufficiency and put in charge of respective computers 31-34 on condition that communication resources on the same path are not allocated repeatedly to plural line allocation requests; and the line is allocated by decentralization restriction sufficiency algorithm in decentralized computer environment.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a communication network line allocation method for allocating communication resources to a plurality of line allocation requests when there are a plurality of line allocation requests at the same time.

[0002]

2. Description of the Related Art Conventionally, when allocating a communication path to a plurality of line allocation requests, it has hitherto been mainly decided on a single computer how to allocate the communication path based on various search methods.

[0003]

In the conventional method of allocating communication paths for a plurality of line allocation requests on a single computer, the load is excessively concentrated on a single system and the database becomes huge. However, there was a drawback that the entire system went down when a failure occurred.

An object of the present invention is to provide a communication network line allocating method in which the load can be distributed, the database does not become huge, and the entire system does not go down when a failure occurs.

[0005]

According to the communication network line allocation method of the present invention, when a plurality of line allocation requests are simultaneously made to the lines in the communication network, the communication resources on the same route are spread over the plurality of line allocation requests. Under the condition that duplicate allocation is not done,
Each of the line allocation requests is formulated as distribution constraint satisfaction and shared by one computer, and circuit allocation is performed in a distributed computer environment by a distribution constraint satisfaction algorithm.

[0006]

The present invention is characterized in that a communication network line allocation problem for allocating communication resources to a plurality of line allocation requests is formulated as a distributed constraint satisfaction problem and solved by a distributed constraint satisfaction algorithm. Distributed Constraint Satisfaction Problem constraint satisfaction problems m variables x _1, x _2, ..., are defined and x _m, a set D ₁ of the possible values of that variable, D _2, ..., by a set of D _m and constraints . The constraint is represented by a set of possible values for each variable. In the following, the predicate P that takes a set of variables as an argument
The constraint is represented by (x _i , x _j ). That is, the constraint P (x
_i , x _j ) is defined for the Cartesian product D _i × D _j .

Solving the constraint satisfaction problem is to find a combination of values satisfying all the constraints given the above variables, regions and constraints.

The distributed constraint satisfaction problem is a constraint satisfaction problem in a distributed environment, and the variables, areas, constraints, etc. in question are distributed and managed by a plurality of agents (computers) that operate autonomously.

Under this condition, the fact that the distributed constraint satisfaction problem can be solved means that the value of the variable x _i is d _i in all agents.
(D _i εD _i ), all constraints are x ₁ = d ₁ ,
It is true under the condition of x ₂ = d ₂ , x ₃ = d ₃ , ..., X _n = d _n . Formalization of distributed network allocation by satisfying distributed constraints The distributed circuit allocation problem is that multiple agents bring requests for circuit allocation and allocate resources on the route so that resource competition does not occur between allocation requests. Is a problem that constitutes. The agent needs to have information on the entire network.
Further, the constraint violation here is a state in which resources on the same route are redundantly allocated. This is shown in FIG.

According to the present invention, the distributed line allocation problem is distributed constraint
It is formulated as a satisfaction problem. Communications subject to this issue
The net is represented by an undirected graph.・ Variable Here, the variable is the line allocation request that connects any two nodes.
is there. For example, in FIG.₁₁, N₄₄Connect between
This is the line allocation request. -The area value is a set of multiple errors necessary to satisfy the circuit allocation request.
It is a route composed of N₁₁, N₄₄Line connecting between
The value for the allocation request is {N₁₁-N_{twenty one}-N_{twenty two}-N ₃₂-N₃₃
-N₃₄-N₄₄} Edge group. This line allocation request
There are multiple satisfied routes on the graph, and the route group
It becomes an area.・ Constraints The constraint here is that the same edge is assigned to multiple line assignment requests.
It means that you must not end. In the state of FIG. 1, N
₁₁, N₄₄Between and N₄₁, N₄₁Two times to connect between
N for line allocation request_{twenty two}-N₃₂Assigned the edge of
It is a constraint violation.・ Agent One variable (line allocation request) is allocated to each agent
Be taken. Also, all agents are assigned to variables that belong to themselves.
Know the area (set of routes that satisfy the circuit allocation request)
Suppose

As an example of the state where the solution is obtained,
Agent 1 is {N₁₁-N_{twenty one}-N _{twenty two}-N₃₂-N₃₃-N
₃₄-N₄₄} A route consisting of edges
Event 2 is {N₁₄-N₁₃-N_{twenty three}-N₃₃-N₄₃-N₄₂−
N₄₁} The state of assigning a route composed of edges
can give.

The communication network line assignment problem thus formulated can be solved by a distributed constraint satisfaction algorithm by a plurality of agents. Asynchronous backtracking (references: Makoto Yokoo, Edmond H. Durphy, Toru Ishida, Kazuhiro Kuwahara: Formulation and solution of distributed cooperative problem by satisfying distributed constraints, electronic information communication) Academic journal J75-D-1,
8, pp. 704-713 (1992). )It has been known. Subsequently, this algorithm will be briefly described with reference to FIG. Asynchronous backtracking This algorithm is characterized by backtracking when viewed globally by an agent group that operates asynchronously and in parallel.

This algorithm is premised on that all agents have a total order by ID.

Each agent assigns a value asynchronously and in parallel, and sends the value to the agent bound by the constraint (step 11). The agent then waits for a message (step 12). One is OK? The message is sent from the upper agent to the lower agent and asks whether the assigned value violates the constraint. The other is nogo
In the od message, this message is sent to the agent one level above, and indicates that a constraint violation has been found.

Agents view the set of variable value assignments sent by other agents as agent-view.
Manage as. Is each agent OK? When a message is received, assign the sent value to the agent_v
It is added to the iew (step 13), and it is checked whether the current assignment of the own value is consistent with the agent_view (step 14). Assigning your own variables is agent
Consistent with _view means that the constraints evaluated by the agent are all true under their own value assignments and the value assignments shown in agent_view, and that nogoods communicated from other agents described below are This is the case when there is a match. If your variable's value is agent_
If it conflicts with the view, set your value to agent_v
Change it so that it does not conflict with the iew (steps 16 to 1)
9).

All the values of the variables of the agent are a
If it conflicts with a subset of gent_view, its a
The subset gent_view is nogood. If an agent finds that a subset of his agent_view is nogood, then the value assignments of the other agents need to be changed, causing backtracking and n
Send the good message (steps 16 and 1)
7, 20, 21). This corresponds to the back track.
Since nogood is monotonically stored in the agent, it does not fall into an infinite loop even if the agents operate asynchronously. If you derive one solution, it will be nogood
Is not sent and the algorithm converges.

[0017]

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

FIG. 3 is a conceptual diagram showing an example of a communication network having a plurality of computers, FIG. 4 is a diagram showing an example of the communication network 35, and FIG. 5 is a diagram showing an example of line allocation in the communication network 35 of FIG. is there. The communication network 35 shown in FIG. 3 is managed by four computers 31 to 34. Each of the computers 31 to 34 includes an independent storage medium (memory, hard disk, etc.), an input / output device, an arithmetic processor, and a communication network connection module for connecting to the communication network 35.

[0019] Now, in the communication network shown in FIG. 4, between the nodes a ₁ -e _1, between b ₁ -e _1, between _{c 1 -e 1, a 1 -d} 1
It is assumed that there is a line allocation request to connect the two.

Consider this as a communication network distributed allocation problem
And the variable x₁ = A₁ -E₁ While the variable x ₂ = B₁ -E₁ 
While the variable x₃ = C₁ -E₁ While the variable x_Four = A₁ -D₁ while
Defined as the line connection request of each variable x₁ , X₂ , X
₃ , X_Four Computer 3 as four agents in charge of
1, 32, 33, 34 will be defined. Table 1, table
Variables and areas are defined in 2 and the constraint is to duplicate the same communication path.
It will not be assigned to a number variable.

[0021]

[Table 1]

[0022]

[Table 2] Here, the constraint P (x ₁ , x ₂ ) between the variables x ₁ and x ₂
Can be defined as:

P (x ₁ , x ₂ ) = (x ₁ , x ₂ ) ε {(d ₁₁ ,, d ₂₁ ) (d
₁₁ , d ₂₂ ) (d ₁₁ , d ₂₃ ) (d ₁₂ , d ₂₁ ) (d ₁₂ , d ₂₂ ) (d ₁₂ , d ₂₂ ) (d ₁₂ , d
₂₃ ) (d ₁₂ , d ₂₄ ) (d ₁₂ , d ₂₅ ) (d ₁₃ , d ₂₃ ) (d ₁₃ , d ₂₄ ) (d ₁₄ , d ₂₃ )} Indicate whether processing is performed, where each agent 3
The total order of _{1 to} 34 is x ₁ > x ₂ > x ₃ > x ₄ .

Each of the agents 31 to 34 independently allocates itself and becomes in the state of x ₁ = d ₁₁ , x ₂ = d ₂₁ , x ₃ = d ₃₁ , x ₄ = d ₄₁ and is OK? A message is sent to each subordinate agent (step 11). In this state, the agents 33, 34 in charge of the variables x ₃ , x ₄
Was received from the upper agent? Since an inconsistent assignment is made to the agent_view by the message, the value assignment is changed (step 12).
~ 16). However, in this state, a ratio of consistent values to the variable x ₃ is impossible, and {(x ₁ , d ₁₁ )
(X ₂ , d ₂₁ )} becomes nogood (step 1
7, 20), the agent 32 in charge of the variable x ₂ receives n
The good message is sent (step 21).

Variable that received the nogood message
x₂ The agent 32 in charge of the
Values are reassigned so as to be consistent with. here,
Variable x₂ = D_{twenty two}Becomes However, even in this state
Number x₃ , X_Four Agents 33 and 34 in charge are restricted
It is impossible to assign a value that satisfies₂ In charge of
Sending a nood message to the agent 32
Believed and backtracked. Variable x₂ In charge of
The agent 32 that finally executes {(x₁ , D ₁₁) (X
₂ , D_{twenty one})} {(X₁ , D₁₁) (X₂ , D_{twenty two})} {(X
₁ , D₁₁) (X ₂ , D_{twenty three})} {(X₁ , D₁₁) (X₂ ，
d_{twenty four})} {(X₁ , D₁₁) (X₂ , D_{twenty five})} All n
finds that it is good, and ends up with the variable x₁ In charge of
The agent 31 with {(x₁ , D₁₁)} N
Send the good message. Where the variable x₁ To
The agent 31 in charge changes its allocation, and x₁ =
d ₁₂And

Under this condition, each agent 31-34
Repeats the above-mentioned processing and ends up in the assignment of x ₁ = d ₁₂ , x ₂ = d ₂₃ , x ₃ = d ₃₂ , x ₄ = d ₄₅ . The allocation for each line allocation request is illustrated in FIG.

FIG. 6 is a diagram showing an example of a line allocation request when the present invention is applied to an actual communication network (FIG. 1 (1) and an example of communication resource allocation for it (FIG. 2)). Node allocation requests (white circles and black circles in the figure) to the communication network connected by 391 routes, 10 line allocation requests R1 to R
10 is set as shown in FIG. 6 (1), and 10 agents are connected by 10 agents as shown in FIG. 6 (2).
1 to R10 are assigned. The calculation process will be omitted because the number of nodes and the number of routes are enormous.

[0028]

As described above, according to the present invention, the communication network line allocation problem of allocating communication resources to a plurality of line allocation requests is formulated as a distributed constraint satisfaction problem, and each computer shares the problem with the distributed constraint satisfaction algorithm. The solution has the following effects. (1) It is faster than the case where the line allocation is processed by one computer. Actually, the processing time of one computer ≈ the internal processing of one agent + communication time + communication waiting time.
At this time, if the communication time is so small that it can be ignored for internal processing, the processing will be completed in 1 / N (N is the number of computers) of the processing time required for allocation in the centralized system. Become. (2) Also, since the line allocation request is processed by multiple computers, the system down of the entire system does not occur even if one computer fails, and there is no problem if N-1 units process the line allocation request. become. (3) The load can be distributed. (4) The database does not become huge.

[Brief description of drawings]

FIG. 1 is a schematic diagram of a communication network distributed line allocation problem.

FIG. 2 is a flow chart of an algorithm / asynchronous backtracking for solving the communication network distributed line allocation problem.

FIG. 3 is a system configuration diagram in one embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of a communication network according to an embodiment.

5 is a diagram showing an example of line allocation in the communication network of FIG.

FIG. 6 is a diagram showing a line allocation request example and a communication resource allocation example for the line allocation request when the present invention is applied to an actual communication network.

[Explanation of symbols]

 1, 2 Agents 11-21 Steps 31-34 Computer 35 Communication network

Claims (1)

[Claims] 1. When a plurality of line allocation requests are simultaneously made to lines in a communication network, each line allocation request is provided under the condition that communication resources on the same route are not overlapped over a plurality of line allocation requests. Is formulated as distributed constraint satisfaction, and each computer is assigned to each of them, and the distributed constraint satisfaction algorithm is used to perform circuit allocation in a distributed computer environment.