JPH03253147A - Distributed setting method for plural routes - Google Patents

Abstract

PURPOSE:To improve resistance against the increment of failure/traffic by constituting routes while guaranteeing the prevention of generation of closed routes in the net routes by using all branches in the network. CONSTITUTION:This distributed setting method consists of a 1st step for successively selecting the route of adjacent nodes along a certain purpose to select a normal route and a 2nd step for finally setting up the route from the 1st node up to the 2nd node as a by-pass other than the normal route. Whether all unselected branches can be selected as a part of the routes or not is decided so that no closed route is generated and executes processing for entering branches to be combined as a part of a route. For instance the route b c a of a node (b) and the rote d b a, d b c a of a node (d) can be used as new routes. Since a substitutive route can be immediately utilized even at the time of increment of failure/traffic in the branches, the resistance can be improved.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides a network in which a packet-switched subnetwork in which packet communication terminals are connected is used as a node Cm point), and each node is connected by a communication line (branch). In a communications network (for example, a LAN interconnected by routers) configured by After selecting a normal route for a purpose such as selecting a normal route, another route from a first node to a second node may be selected for other purposes such as load balancing. The present invention relates to a distributed setting method for a plurality of routes in which the settings are determined in a distributed manner through selection at each node along the route.

Of course, communication actually takes place between the packet communication terminal connected to the first node and the packet communication terminal connected to the second node. However, here, we are not concerned with the connection route of the packet communication terminal within the first node (or the second node), but with the subsequent determination (setting) of the connection route between the nodes. That's why there is.

In addition, in accordance with the above-mentioned purpose of selecting the route with the minimum cost, the present invention is based on the standpoint that the technology of first selecting the normal route exists as a prior art, and the subsequent load distribution This technology is aimed at selecting and determining alternative routes for purposes such as:

[Prior Art] Hereinafter, a packet-switched network will be logically treated as a collection of nodes representing exchanges and technologies representing communication lines. Packets sent and received between packet terminals are assumed to be sent and received between nodes in a pseudo manner.

The information used for route calculation and its notation are defined below.

Every node is given an identifier.

The identifier should be one that can be compared in size (for example, in alphabetical order, and the size is determined by the order). The technique between node i and node j is written as (i, j). branch(
The cost of i, j) is expressed as Wij (=W j= ). Also, let δ be the sum of the branch costs on the line from node i to destination node, and let it be the route cost from node i to destination node.

Here, cost may represent the time required to transmit a packet via a corresponding line or route, or may represent expenses, and in the following description, it represents both.

Now, in the conventional technology, based on the total cost of branches on the route from the originating node to the destination node, a route is configured so that this value is minimized (or minimized). The route constructed at this time is required to be constructed so that it does not include a closed circuit.

When a route is constructed, it is necessary for each node to memorize all nodes on the route, which requires a large storage capacity. Only one node is memorized and a route is constructed by sequentially tracing this node.

From this, when focusing on a specific node on the network and assuming multiple nodes as the originating nodes on the network, all routes from multiple originating nodes to the node are trees with the node as the root. becomes. Hereinafter, when the next node stored by each node i to be relayed to is p along the branches of this tree, p will be called i's parent node and i will be called p's child node.

A typical conventional technique for configuring such a route is the method of Merlin et al. (JEEE TRAN
SACTIONS ON COMMUNICATI
ONS, VOL, C0M-27, NO19゜SEPTE
MBER1979 “A FailsafeDist
“ributed RoutingProtocol”
) and the method of \S 1os+an et al.

The method of Merlin et al. will be described below. As will be described later, since the method of generating a route for the first time can be treated as a special case of updating a route, an updating method of finding a route with a lower cost from an existing route will be described.

Each node exchanges control messages with adjacent nodes while proceeding with the process for setting a route. Generally, the control message MSG sent by node i includes the route cost value from node i to node i as information.

From this control message, the adjacent nodes of node i
We can calculate the cost of the route to the node via .

Although FIG. 1 shows a processing flow according to the present invention, it also shows a conventional processing flow for configuring a route, so the conventional technology will be explained below with reference to FIG.

The conventional method consists of two phases, as seen in FIG. Phase A is a phase in which the weight of the route to the node is recalculated, and is started by the node sending a control message to the Kaneko node on the mainland. The processing in phase A is propagated along the existing tree to the end points (nodes with no child nodes).

In phase B, among the route candidates from each node to D,
The adjacent node that gives the lowest cost is selected as the next node on the route, and the end of the process is notified to the parent node. Phase B processing starts from the endpoint and propagates to nodes closer to the node.

When phase B ends at the node, the tree update ends. In other words, the route has been set.

This updating process improves the cost of the route (this means that the route with the minimum cost has been set).

Furthermore, each node i holds the route cost δ1(k) leading to the node via an adjacent node k during calculation.

Details of each phase are shown below with reference to FIG.

At the Yu-Tama-Ni Engineering Node: (1) δ. ← Set O, that is, route cost δD is set to 0,
This route cost δゎ is sent as a control message to all adjacent nodes to start processing.

At node i (≠D): (1) If a control message is received from adjacent node j,
The cost of the route to node j via adjacent node j is calculated and retained. In other words, the following equation is processed.

δ1(j)←δj +Wij (2) If j is the parent node p of i (a node that is one hop closer to the node on the tree): (2, 1) After the above processing is completed, the calculated The minimum value of the route costs δ1(k) (k is an adjacent node of i) is selected and set as δ-factor.

(2,2) Send the route cost δ5 as a control message to all adjacent nodes except p.

(2, 3) Transition to phase B.

Phase B At node: (1) When a control message is received from adjacent node j,
It remembers that a control message has been received from the node.

(2) If control messages are received from all adjacent nodes, the process ends.

At node i: (1) When a control message is received from adjacent node j, calculate and hold the cost of the route to node i via adjacent node j. In other words, the following equation is processed.

δ+(j)←δj +Wij (2) When control messages are received from all adjacent nodes: (2, 1) Send route cost δ to the (old) parent node P as a control message.

(2, 2) Let the adjacent node that gives the minimum path cost δ be the new parent node p'.

From the calculation formula of process (1), the route cost value decreases each time the route is advanced. From this, it can be seen that there is no closed path on the path. Furthermore, each node has only one path to the node.
(selecting one node as the parent node p),
The paths leading to nodes on the network constitute a tree.

In the process (2) in phase A of node i, the cost calculation of the new route is performed based on the reception of the control message from the parent node, and does not take into account the case of passing through all adjacent nodes. Therefore, the route cost δi is generally not the minimum.

In addition, when a route is configured for the first time (that is, node i has not determined a parent node p), when a control message is received from an adjacent node k for the first time, k is regarded as the parent node p, and the process immediately shifts to phase B. do it.

A specific example of route setting according to the prior art will be described below, targeting the network shown in FIG.

In FIG. 2, the nodes are a and b.

There are four nodes c and d, and the time required for a message to travel between nodes a and b is 3, and the branch cost between them is also 3. Similarly, the number between nodes a and C is 2, the number between nodes and C is 4, the number between nodes and d is 1, and the number between nodes d and C is 1. The process of generating a route to node a is shown in FIGS. 2-1 to 2-12.

In Figures 2-1 to 2-12, t indicates time, N indicates the node where the event occurred, and fro
m indicates the node that sent the message, and is indicated by a thin, short arrow (→
) indicates the message, and the thick dotted arrow indicates the phase A
The branch that gives the minimum route cost at the time of completion is shown, and the thick solid arrow indicates the generated route. Further, nodes marked with a circle indicate nodes in phase B, and nodes marked with a delta mark indicate nodes that have completed processing.

Also, the arrival of control messages to each node at this time.

The details of the cost calculation 2 phase transition are shown in the explanatory diagram of FIG.

First, in Fig. 2-1, node D (=a
) A control message M S G,=0 is sent to crow, c. This control message arrives at node C at time t=2. Node C has a as its parent node, and has a path cost (=2)
After calculating the control message MSGc other than a,
d (Figure 2-2).

Thereafter, the process is similarly propagated from nodes close to node a to nodes far away (FIGS. 2-3 and 2-4). Time t
At =4, node d receives a control message from node S (Fig. 2-5). Node d calculates δd(b) = 4, receives control messages from all neighboring nodes, and selects neighboring node C, which provides the path with the lowest cost, as a new neighboring node (in this case, the old parent node ) and send a control message to the parent node.

=3 is returned and the processing at node d is terminated (Figure 2-6,
Figures 2-7).

Similar processing is performed on other nodes by receiving control messages from all adjacent nodes, and the process ends (
Figures 2-8 to 2-IO). Finally, at t=7, the node receives a control message from the previous adjacent node, and the processing for the entire network ends (Figure 2-11).

Figure 2-12 shows the obtained route, and Figures 4-1 to 4-11 show the process by which the route is updated when branches (c, d) fail. Shown below.

In FIG. 4-1, nodes C and d at both ends discover a failure in branches (c, d) and attempt to notify the nodes. At this time,
Since the route to node d is cut off, only notifications from C reach D (Figure 4-2), 1! The fj point receives the technique failure notification from the node C and starts updating the route.

The details of the arrival of the control message at each node in this series of processing, cost calculation, and phase transition are shown as an explanatory diagram in FIG.

As a result of the processing, a new route (d, b) is selected as the route for the node d, as shown in FIG. 4-11.

[Problem to be solved by the invention]

In the conventional technology described in detail above, only one communication path is set between the originating and terminating nodes. Therefore, when a branch failure occurs, the route must be reconfigured immediately. Also,
When comparing loads, it is not possible to detour the packet flow or distribute the load using multiple routes.

An object of the present invention is to construct a route between originating and destination nodes using all the techniques in the network while ensuring that no closed circuits occur in the route in the network. Another object of the present invention is to provide a distributed setting method for a plurality of routes that can provide other routes in addition to the above.

(Means for Solving the Problems) In order to achieve the above object, in the present invention, packet communication terminals are connected to packet-switched subnetworks as nodes, and each node is connected by a communication line. In line with the purpose of selecting a route with the minimum cost from an arbitrarily selected first node to an arbitrarily selected second node in a communication network configured by After going through the first step according to the prior art of sequentially selecting a route from adjacent node to adjacent node and selecting it as a normal path, the first node first selects the normal path among its adjacent nodes. For each of the remaining adjacent nodes that were not selected as and the cost of the detour route leading to the second node, and based on the results, determine the traveling direction of the communication line route between the first node and the adjacent node, and the following: , by repeating the same procedure for the next adjacent node along the determined traveling direction, the route is determined in a distributed manner for each node, and finally the first node is connected to the second node. We have decided to implement the second step of setting a route from 1 to 3 as a detour route other than the regular route as a new technology.

The present invention targets costs that are equal in both directions of a technique. In addition, in the present invention, this cost is assumed to be a fixed value, but a cost that changes sufficiently slowly compared to the time to which the present invention is applied (such as measuring the transmission delay time at regular intervals and using it as a cost) is It can be treated as a fixed value.

[Effect]

In the present invention, after the route setting/updating (phase A, phase B) according to the conventional technology is completed, as phase C,
At each node, for all branches that were not selected as tree branches, it is determined whether they can be selected as part of the route so that the route does not become a closed path, and techniques that can be incorporated are incorporated as part of the route. Do the processing to go.

As already mentioned, FIG. 1 shows the process flow in the prior art and the present invention.

As shown below, in Phase A and Phase B, the processing is propagated between adjacent nodes, whereas in Phase C, each node performs processing independently without any special communication.

The details of the processing are as follows.

The explanation will be given on the assumption that each node holds the following information as a result of executing phases A and B.

(a) Parent node p of node i on the tree.

(b) Child node CI + CZ r of node i on the tree
... c7 (C) Cost δ of the route from node i to node ri
, (=δp + W 1 p ) + all adjacent nodes k
(d) Cost W ij+ of edge (i, k) and (e) Path cost δ1(k) (=δp
+ Wik) In the present invention, processing is performed at each node after a new tree is constructed by applying an existing algorithm. Node i
For a technique (i, j) that is not on the tree, the cost of adopting (i, j) as a new path for i is greater than the cost of adopting it as a new path for j (if they are equal, then , following the alphabetical order, i
<j), select technique (i, j) as part of the route to D.

In other words, a node i has an edge (
i, j) as the path from i to D.

δj(i) (=δi + W = j) 〉δt(j)
......(a) or δJ(i)=δ, (i) and i>j...
(b) Hereafter, this process will be referred to as node i adopting technique (i, j) as an additional technique.

Since the present invention is characterized by the dispersibility of processing, this point will be explained. In other words, processing at each node can be performed independently of processing at other nodes. Further, since the present invention does not make any assumptions before the network configuration, it can operate on a network with any connection type.

Furthermore, since the present invention is characterized by the non-circular nature of the path, this point will also be explained. The path constructed according to the present invention does not include a closed circuit. A simple proof is shown below.

From the definition of a path on a tree, the following formula holds true for any node i and its parent node P.

δ1〉　δ.

Furthermore, from the above conditional expressions (a) and (b), it is generally understood that nodes i and j do not both select the branch (i, j), and that the technique (i, j) is in the direction from i to j. is selected as the route, δ1≧δj (the equal sign is when i > j) holds true. Therefore, there are no closed circuits on the route.

A simple proof using the contrarian method is as follows.

If the cycle 11 + tz + ...jl'l+11
If δ exists, then from the above conditional expressions (a) and (b), δ
, ≧δe2≧...≧δil + that is, δ1=δ1□=
...=δ□1, and from the condition of equality, 1+>1g>...>ill>i, and so on.

il>i, which is a contradiction.

This completes the proof.

〔Example〕

FIG. 6 shows the route and route cost of a tree structure whose root is node D (a in the figure) obtained by applying the conventional technique (phases A and B). The result of applying the present invention (phase C) to this is shown in FIG.

Further, condition determination at each node is illustrated in FIG. 8 as an explanatory diagram.

Several cases representing the characteristics of the processing will be explained in detail. For example, node S is adjacent nodes c and d other than parent node a.
For the branches (b, c) and (b, d) connected to , it is determined whether these should be selected as additional techniques. Branch (b
, c), δ, (b) = δb + W bc = 3 + 4 >
Since δb(c)=6, the node Suha branch (b, c) is adopted as an additional technique. On the other hand, for branches (b, d), δa(b) = δb+ Wbd = 3 + 1 = δd(
b)=4°bad.

Therefore, the node suma techniques (b, d) are not adopted as additional techniques.

Furthermore, in the larger network shown in Figure 9, Figure 10 shows the results of applying the prior art, which is the first stage of implementing the present invention, and then applying the second stage of implementing the present invention. The results are shown in FIG. 11, and the process of condition determination in the present invention is shown as an explanatory diagram in FIG. 12, so please refer to it.

Next, the features of the present invention will be summarized and explained.

Explain the comprehensiveness of branch selection. The above conditional expression (a)
, (b), all non-tree branches (1+j) are nodes i
, j is always selected as the path on the route. Therefore, all branches on the train are used as routes.

Furthermore, we will discuss the flexibility of cost calculation formulas.

The conditional expression for the cost on the tree is 61= between node i and parent node p.
62+W.

Even if not, the following conditions should be met.

δ1≧δ, +W.

If , the acyclic nature of the path is preserved.

From this, the route cost does not necessarily have to be the minimum cost. From this, δ□ ≧ δ, +wi, +x.

It is possible to reflect a new cost X such that

Next, reduction of the information to be used will be explained.

Between node i and its parent node C, δ, 〉 δ. stands firm. Therefore, node i can construct a path that does not include a cycle even if it does not have child node information.

This is because, from the above equation, node i will never select technique (i, c) (in the opposite direction to the tree). However, since node i needs to perform calculations on all nodes other than the parent node, processing increases for the entire network.

Next, combinations with other conventional techniques will be explained.

Phase C can be combined with other conventional algorithms to replace Phase A and Phase B.

The processing methods of Phase A and Phase B themselves are not essential to Phase C.

〔Effect of the invention〕

According to the present invention, it is possible to improve resistance to failures/increased traffic in a communication network. In the case of FIG. 7, in addition to the three branches on the tree, the other two branches are all selected so that they are not closed paths.

Here, for example, the node S is the path b-+C-+a, and the node d
-+b→a and d→b→C→a become available as new routes. As a result, an alternative route can be immediately used even in the event of a branch failure/increase in traffic, thereby improving durability.

Furthermore, in large-scale nets, the ratio of the number of techniques to the number of nodes generally increases, so it is expected that the effect of improving strength will increase. For example, in FIG. 7, the number of branches on the route increases from three to five, while in FIG. 11, the number increases from nine to 15.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing both the processing flow according to the present invention and the processing flow according to the prior art, FIG. 2 is an explanatory diagram showing an example of a network used to explain a specific example of route setting according to the prior art, and FIG. 2-1 2-12 are explanatory diagrams showing the route setting process according to the prior art for the network shown in FIG. 2, FIG. Figures 1 to 4-11 are explanatory diagrams showing the process by which a route is updated when a failure occurs in a branch within the route, and Figure 5 summarizes the route update process in a table. An explanatory diagram, FIG. 6 is an explanatory diagram showing the path and route cost of a tree structure obtained by executing the first stage according to the prior art, and FIG. 7 is an explanatory diagram showing the second stage (C phase) according to the present invention. Fig. 8 is an explanatory diagram showing the results of applying the method, Fig. 8 is an explanatory diagram summarizing the condition judgments at each node at this time and shown in a table, Fig. 9 is an explanatory diagram showing an example of a large-scale network, and Fig. 10 is an explanatory diagram showing the result of applying the prior art which is the first stage of implementing the present invention to this, Fig. 11 is an explanatory diagram showing the result of applying the second stage of implementing the present invention to this, and Fig. 12 is an explanatory diagram of the result. FIG. 4 is an explanatory diagram illustrating condition determinations at each node at a time in a table. Explanation of symbols t...Time, N...Node where the event occurred, from
...Node that sent the message

Claims (1)

[Scope of Claims] 1) In a communication network configured by connecting packet-switched sub-networks with packet communication terminals connected therein as nodes, and connecting each node with a communication line, A route from one node to an arbitrarily selected second node is sequentially selected from adjacent node to adjacent node for a certain purpose, such as selecting the route with the minimum cost. and then the first node first selects among its neighboring nodes as a normal route.
For each of the remaining adjacent nodes that were not selected as the normal route, the cost of a detour route from the first node to the second node via the adjacent node, and the cost of the detour route from the adjacent node to the first node. and the cost of a detour route leading to the second node via the node, and based on the result, determine the traveling direction of the communication line route between the first node and the adjacent node. death,
Thereafter, by repeating the same procedure for the next adjacent node along the determined traveling direction, the route is determined in a distributed manner for each node, and finally the route is determined from the first node to the second node. A method for dispersively setting multiple routes, comprising: a second step of setting a route from a node as a detour route other than the regular route.