KR100248075B1 - Multi-cast path setting method of atm network - Google Patents

Abstract

The present invention relates to a multicast path determination method in an asynchronous transmission mode network.

In the multicast path determination method in the asynchronous transfer mode (ATM) network based on the current virtual path (VP), the receiving node stops receiving the service during the service talk time or the node newly receives the service. The dynamic phenomena of the receiving node such as the case of participating in the service are not considered. Accordingly, if a method that does not take into account the dynamics of multicast connections is applied, some receiving nodes are unnecessary when routing services such as Video On Demand (VOD) and Video Conference. Using resources results in inefficient use of network resources.

In order to solve this problem, in the present invention, the asynchronous transmission that can efficiently use network resources in consideration of the probability that the node joins the service, the probability that the receiving node is in service, and the delay time between the transmitting and receiving nodes for real time communication. A multicast path determination method in a mode network is presented.

Description

How to Set Up Multicast Routes in Asynchronous Transfer Mode Networks

The present invention relates to a method for establishing a multicast path in an asynchronous transfer mode network, and more particularly, to a multicast path in an asynchronous transfer mode (hereinafter referred to as ATM) network based on a virtual path (hereinafter referred to as VP). Multicast path determination method in asynchronous transmission mode network that can be used to obtain the lowest cost spanning tree considering delay time between transmitting and receiving nodes in order to efficiently use network resources when determining the path to provide services It is about.

In the multicast routing scheme in the ATM network based on the VP currently being used, dynamic phenomena of the receiving node such as when the receiving node stops receiving the service or the node newly joins the service during the communication time of the service are Did not consider Therefore, if you apply a method that does not take into account the dynamics of multicast connections to route services such as Video On Demand (VOD) and Video Conference, some receiving nodes will be unnecessarily Because of the use of resources, there is a problem of inefficient use of network resources.

Accordingly, the present invention provides a virtual path switching method for constructing a spanning tree in consideration of a probability that a node joins a service, a probability that a receiving node of a service quits a service, and a delay time between transmitting and receiving nodes for real-time communication. VPX / Virtual Channel Exchange (hereinafter referred to as VCX) node creates a reduced network consisting of only VCX nodes from a given ATM network and satisfies the constraints of latency between transmit and receive nodes in this reduced network. The purpose of the present invention is to provide a multicast path determination method in an asynchronous transmission mode network that can efficiently use network resources.

Multicast path determination method in the asynchronous transmission mode network according to the present invention for achieving the above object is to obtain a reduced network consisting of only the virtual channel switching node from a given asynchronous transmission mode network consisting of a virtual path switching node and a virtual channel switching node. Obtaining a path having a minimum cost while satisfying a tolerance value of delay time for all transmission / reception node pairs for the reduced network; and selecting a path having the greatest number of relay nodes among the paths having the least cost. Defining a tree, selecting a receiving node having a minimum weight by calculating a weight of costs for all receiving nodes whose path is not determined, and receiving the node having the minimum weight and the tree on the tree determined so far. Determine nodes and use them to send and receive least cost Selecting a path, adding a minimum cost transmission / reception path to an initial tree, checking whether the new tree obtained by adding the transmission path to the initial tree includes all reception modes, and If the obtained tree does not include all the reception modes as a result of checking whether all the reception modes are included, selecting the reception node having the minimum weight by calculating the weight of the cost for all the reception nodes whose path is not determined. And if the tree includes all reception modes as a result of checking whether the tree includes all reception modes, converting a virtual link of the obtained tree into a physical link on a given circle graph. It features.

1 (a) and 1 (b) are functional diagrams for explaining a method for generating a reduced network from a given ATM network.

2 (a) and 2 (b) illustrate an example of a multicast routing tree.

3 is a flowchart illustrating a multicast path determination method in an ATM network according to the present invention.

Hereinafter, with reference to the accompanying drawings will be described in detail the present invention.

1 (a) and 1 (b) are functional diagrams illustrating a method of generating a reduced network from a given ATM network, and FIGS. 2 (a) and 2 (b) are exemplary diagrams of a multicast routing tree. Indicates. The cost function problem of the routing tree to be considered for multicast path determination in a VP-based ATM network, a method of generating a reduced network, and a multicast routing algorithm will be described with reference to FIGS. 1 and 2.

The ATM network may be modeled by a graph G (V, E) of V representing a switching node and E representing a physical link. Then, the routing decision problem of multicast considering the delay time is defined as follows. In other words, the ATM network graph G (V, E), the transmitting node s, the set of receiving nodes D = {d ₁ ,... , d _n } and allowable delay time between nodes H _{sd i} Given is a problem of finding a routing tree with minimum cost for a multicast connection that includes all nodes belonging to s and D while satisfying the delay condition.

If the multicast routing tree is R = {R (v), R (e)}, where R (v) and R (e) represent a set of nodes and a set of links in the multicast tree, respectively. The cost C _T (R) of the multicast routing tree consists of three types of cost functions: switching cost C _S (R), connection establishment cost C _E (R), and band cost C _B (R) at the switching node. . First, the switching cost depends on the multicast cell processing time at that node and assumes that the multicast cell processing time at the node is proportional to the required bandwidth capacity. In addition, the switching cost is divided into the switching cost at the VPX node and the switching cost at the VCX node. If t _sd is a bandwidth capacity required for multicast communication, the switching cost in the tree R can be expressed as Equation 1 below.

C _S (R) = t _sd (n _vpx (R) + _{δn vcx} (R))

Here, n _vpx (R) and n _vcx (R) refer to the number of VPX nodes in the tree R and the number of VCX nodes including the transmitting node, respectively, δ is the VC switching cost versus the VP switching cost per unit band. The relative value is usually δ≥1.

Connection establishment cost C _E (R) is a measure of the complexity of the routing establishment process between all transmitting and receiving nodes. To provide multicast communication, assuming that the transmitting node is a VCX node capable of duplicating a multicast cell. It is formulated to be proportional to the number of VCX nodes used. Therefore, the connection establishment cost C _E (R) is expressed as in Equation 2 below.

C _E (R) = n _vcx (R) -1

또한, 노드 i와 j사이의 경로를 연결하는 VP 중에서 이용 가능한 대역용량이 가장 큰 VP를 ω_ij라고 하면, VP ω_ij의 대역 비용 C_VPB (ω_ij) 는 [수학식 3]과 같이 나타낼 수 있다Bandwidth C _B (R) is the sum of the bandwidth used for multicast communication from the transmitting node to all receiving nodes, and is generally assumed to be the sum of the required bandwidth capacity allocated to all links in the tree. However, network resources can be efficiently used by first selecting a VP having a large available bandwidth among all VPs connecting two VCXs. By doing this, a small available band of VP may be used for other services such as unicast. For this purpose the band cost C _B (R) can be formulated taking into account the available bands on each path. That is, the bandwidth cost C _B (R) may be expressed as a function of the sum of the bandwidth capacities required for multicast communication and the available bandwidth of the VP in the tree. the number of VPs connecting τ _ij between two VCX nodes i and j, m = 1,... , Assumed to be ω ^m _ij from the m- th VP VP connecting two nodes i and j when the VCX τ _ij. In addition, it is assumed that bandwidth capacity that can be used as a, and a _ij available bandwidth capacity of a a ^m m-th VP _{ij ij} ω ^m at the maximum. In other words,

In addition, suppose that VP having the largest usable bandwidth capacity among the VPs connecting the paths between the nodes i and j is ω _ij , the bandwidth cost of VP ω _ij C _{VP B} (ω _ij ) Can be expressed as [Equation 3]

In this case, k _ij means the number of physical links between the node i and j.

Accordingly, the total bandwidth cost is obtained by adding the bandwidth costs of all VPs in the tree R as shown in Equation 4 below.

Therefore, the total cost C _T (R) of the multicast routing tree is the sum of C _S (R), C _E (R), and C _B (R), and is formulated as in Equation 5 below.

C _T (R) = αC _S (R) + βC _E (R) + γC _B (R)

Here, α, β and γ denote weights of switching cost, connection establishment cost and bandwidth cost with respect to the total cost of the multicast routing tree.

In a VP-based ATM network, the end node of each VP or VC may be only a VCX node, and the intermediate nodes via the VP or VC may be VPX or VCX. One VP may be via a plurality of VPX nodes, and one VC is set on one VP or on a connection of a plurality of VPs. All VPX nodes on the routing tree can be considered as relay nodes in a multicast connection. Then, all receiving nodes of the multicast connection are located in the VCX node, and only the VCX node has a function of replicating the multicast cell and transmitting it to the next node connected to it. Therefore, it is necessary to obtain a spanning tree for multicast routing on a reduced network G 'consisting of only VCX nodes. To do this, first create a reduced graph G 'for a given graph G. This network is easily obtained by removing all VPX nodes in graph G and then introducing a logical link to connect two adjacent VCX nodes directly, which is called a virtual link. In this way, we get a reduced network G 'that is much simpler than the original given network graph G.

와 같이 표현된다고 하면, 가상링크 l^V _ij상의 비용함수 C(l^V _ij)는 가상링크 상의 스위칭 비용, 접속설정 비용 및 대역 비용의 합으로 [수학식 6]과 같이 표현될 수 있다. ^V _ij is called a virtual link introduced after removing all VPX nodes between two adjacent VCX nodes i and j , and the routing tree in this reduced network G 'is called R' = {R '(v), R' ( e)}. Here, R '(v) and R' (e) represent a set of VCX nodes and a set of virtual links in the multicast routing tree, respectively. The cost function in the reduced graph G 'can be formulated using the above equations. For example, the connection setup cost has a unit cost for one VC setup, and the bandwidth cost is

That when expressed as, the cost function C (l ^V _ij) on the virtual link l ^V _ij can be expressed by [Equation 6] by the sum of the cost of switching, connection setup cost, and bandwidth cost on the virtual link.

Here, a ^v _ij means available bandwidth on the virtual link l ^v _ij , and k ^v _ij means the number of physical links in the virtual link l ^v _ij . Therefore, the total cost for multicast routing in the graph G 'is expressed by Equation 7 by adding the costs of each virtual link in the tree R'.

과 2가 된다.In the present invention, the number of hops between all transmit / receive node pairs, that is, the number of physical links is considered as a delay constraint. The delay time on each virtual link in the reduced network G 'is equal to 1 in the graph G, which is the number of VPX nodes located between the two VCX nodes that are the start and end nodes of the virtual link. That is, the delay time on the virtual link l _ij ^v is equal to the value of k _ij ^v . FIG. 1 (b) shows an example in which a reduced network G 'is obtained by removing all VPX nodes and introducing virtual links in the circle graph given in FIG. 1 (a). For example, a virtual link cost C l ^v (l ^v _s8), and the delay time of l ^v _{s8 s8} are each

And 2

The multicast routing algorithm using the cost function of the routing tree and the generation method of the reduced network is as follows.

First, C _k (s, d) is defined as the cost of path p (s, d) while the delay time from node s to node d is k (that is, h _sd = k). This can be easily calculated using dynamic programming and is represented by Equation 8 below.

Where h _ud represents the delay between node pair u and d. If H _sd is the allowable value of the delay time between the transmitting node s and the receiving node d, the minimum cost that satisfies the allowable value of the delay time H _sd is obtained by [Equation 9].

C (s, d) = min _{k≤H SD} C _k (s, d)

Currently used multicast routing algorithms use the dynamic programming schemes of Equations 8 and 9 to allow latency values for all transmit and receive node pairs. H _{sd i} (d _i ∈ D) Find the path with the lowest cost while satisfying. This least cost path is called p (s, d _i ) when d _i ∈ D for all i. The least cost path is selected among the obtained least cost paths, and this is defined as an initial tree R '= {R' (v), R '(e)}. For every receiving node whose path is not determined, the path of least cost from the tree R '= {R' (v), R '(e)} is obtained, and the path between the transmission and reception having the smallest value is selected. When this selected path is n _i ^* ∈R '(v), it is defined as p (n _i ^* , d _i ^* ). Then, the path p (n _i ^* , d _i ^* ) is added to the tree R 'with R' ← R '+ p (n _i ^* , d _i ^* ). The above process is repeated until all the multicast receiving nodes are included in one tree R '.

However, since the static multicast routing algorithm aims at optimal routing at connection request, when the number of receiving nodes of the multicast group changes within the service time of the multicast connection, network resources cannot be efficiently used.

Fig. 2 (a) shows that when the transmitting node is s, the set of receiving nodes is D = {2, 3, 4, 5}, and the allowable value of delay time between all transmitting and receiving nodes is i = 2, 3, 4, 5; When H _1i = 4, it represents the routing tree obtained using the static multicast algorithm. The total cost of the initial multicast tree in this case is 41. Assuming that node 3 stops receiving service during the talk time of the multicast connection, the cost of the tree is equal to the initial cost 41 even after node 3 is deleted from the set of receiving nodes.

Therefore, when the membership of a set of multicast receiving nodes changes frequently within the service time of a connection, a dynamic multicast routing algorithm that can efficiently use network resources is required. To do this, we first introduce the frequency components f _a and f _d representing the probability that a non-receiving node will newly join the multicast connection within the multicast connection time, and the probability that the currently receiving node will leave the multicast connection, where 0≤f _a ≤1, 0≤f _b ≤1. It is assumed that the values of the frequency components f _a and f _b depend on the application. If the value of f _a is 0, it means that the non-receiving node never joins the multicast connection. Although it is not possible to obtain an accurate value for the frequency factor, it can be given an approximate value depending on the application. For the receiving node d _i to be connected to the current tree R ', the weight W (d _i ) is calculated for each receiving node d _i using [Equation 10] in consideration of the frequency components f _a and f _b .

라 한다. 만일, 최소의 W(d_i)값을 가진 노드 d_i ^*가 복수개 존재하면, 중계 노드의 수가 제일 많은 노드 d_i ^*를 선택하고, 현재 트리상의 노드를 다음의 [수학식 11]에 의해 선택하여 경로 p(n_i ^*, d_i ^*)를 현재의 트리에 추가한다.Then, among the receiving nodes included in set D, but the path has not been determined so far. h _{su i} + h _{u i d i} ≤H _{sd i} And h _{sv i} + h _{v i d i} ≤H _{sd i} The node d _i with the minimum W (d _i ) value is selected while satisfying the delay constraint of. This selected node is called d _i ^* . That is, d _i ^* is

It is called. If, when the node with the minimum of W (d _i) value d _i ^* a plurality is present, the number of relay nodes, select the best number of nodes d _i ^* and the current selection, the node in the tree by the following formula 11] Add the paths p (n _i ^* , d _i ^* ) to the current tree.

Here, u _i (#R '(v)) represents the node on tree R' that provides the least cost tree when connecting node d _i to the current tree R '. On the other hand, v _i (#R '(v)) means a transmitting node or a relay node on the tree R' when the node d _i is connected to the current tree R '. Therefore, the first term in Equation 10 represents the weight of the cost of connecting the receiving node d _i with the least cost path in consideration of the frequency factors f _a and f _b , and the second term is the frequency factor f _a. Considering f and f _b , it means the relative weight of the cost required to connect the receiving node d _i from the transmitting node or the relay node on the current tree. Then, the selected receiving node d _i ^* is connected to the tree R 'via node n _i ^* on the current tree determined by Equation (11). This process is performed repeatedly until all receiving nodes are included in a single tree.

In this way, after the multicast routing tree R '= {R' (v), R '(e)} is determined in consideration of the delay time, even if the number of receiving nodes of the multicast group changes within the service time, FIG. As shown in Fig. 2), the network resources can be used efficiently. Since the multicast routing algorithm proposed in the present invention sets up the routing tree in consideration of the frequency components f _a and f _b , the probability that each receiving node is located in the leaf node of the tree is higher than that of the static algorithm. do. As in the example of the static algorithm, assuming that node 3 stops receiving service within the service time of the multicast connection, the tree cost goes from 42 to 35 by eliminating the link transmitting the multicast cell to receiving node 3 and receiving node 3. Will decrease.

3 is a flowchart illustrating a multicast path determination method in an ATM network according to the present invention.

First, a given ATM network G, composed of VPX and VCX nodes, has a transmitting node s, a set of receiving nodes D = (d ₁ , d ₂ ,…, d _n }, frequency components f _a , f _b, and allowable values of delay time. H _{sd i} And a reduced network consisting of only VCX nodes is obtained from this network (301). Then, the allowable value of the delay time for all the transmission and reception node pairs using Equation 8 and Equation 9 for the obtained reduced network. H _{sd i} The path p (s, d _i ) having the minimum cost while satisfying (d _i ∈ D) is obtained. The path P ₁ (s, d _i ) having the largest number of relay nodes among the obtained least cost paths is selected and defined as an initial tree R '= {R' (v), R '(e)} (302). Then, for every receiving node whose path is not determined, the cost required by using Equation 10, that is, the weight of resource W (d _i ) is calculated and the receiving node d having the smallest W (d _i ) value. Select _i The selected receiving node d _i is defined as d _i ^* (303). Also, using Equation 11, node n _i ^* on the tree determined up to now is determined. p (n _i ^* , d _i ^* ) is called a path connecting nodes d _i ^* and n _i ^* (304). Then, the path p (n _i ^* , d _i ^* ) is added to the tree R 'up to the present (305). Next, verify that the tree R 'includes all receive modes (306) and if it does not include all receive modes, proceed to step 303 until all receive nodes of the multicast connection are included in one tree R'. Repeat. When all the transmitting and receiving nodes have determined the spanning tree while satisfying the delay time on the reduced graph G ', the virtual link of the tree R' is converted into the physical link on the given circle graph G (307).

As described above, according to the present invention, when providing a multicast service in an asynchronous transmission mode network based on a virtual path, a probability that a node joins a service, a probability that a receiving node in the service quits a service, and By determining the service path in consideration of the delay time between the transmitting and receiving nodes, there is an excellent effect of efficiently using network resources.

Claims (1)

Obtaining a reduced network consisting of only virtual channel switching nodes from a given asynchronous transmission mode network consisting of a virtual path switching node and a virtual channel switching node; Obtaining a path having a minimum cost while satisfying a tolerance value of delay time for all transmission / reception node pairs for the reduced network; Selecting a path having the largest number of relay nodes among the least cost paths and defining the path as an initial tree; Selecting a receiving node having a minimum weight by calculating weights of costs for all receiving nodes whose paths are not determined; Determining a receiving node having the minimum weight and a node on a tree determined so far, and using the same, selecting a transmission / reception path having a minimum cost weight; Adding a transmission / reception path of the minimum cost weight to an initial tree; Verifying that the tree obtained contains all of the receive modes, As a result of confirming that the obtained tree includes all the reception modes, if the obtained tree does not include all the reception modes, the weight having the minimum weight is calculated by calculating the weight of the cost for all the reception nodes whose path is not determined. Proceeding to selecting a node, Determining that the obtained tree includes all reception modes, and if the obtained tree includes all reception modes, converting a virtual link of the obtained tree into a physical link on a given circle graph. Multicast path determination method in asynchronous transmission mode network.

Images