JP5621565B2 - Multicast system and node of multicast system - Google Patents

Description

  The present invention relates to a multicast system. In particular, the present invention relates to a multicast system having a multicast tree structure in which a plurality of nodes are connected in a tree shape from a root node. Further, the present invention relates to a multicast system node, a multicast tree update method, and a computer program for reconfiguring a tree by adding a node in the multicast system.

  There is a multicast system that distributes update information of a WEB server or data generated by a sensor or the like to a large number of devices using a multicast tree having a data distribution source as a root.

  This multicast tree is a logical network composed of nodes that use the data, and may be created for each distribution source or may share one multicast tree. Data is transferred between nodes according to the structure of the multicast tree.

  Regarding the multicast tree construction method, several means for constructing a multicast tree for distributing data in real time to a large number of users have been proposed.

  For example, in Patent Document 1, by exchanging information such as CPU usage rate, memory usage rate, CPU specification, physical proximity, and survival time of a node with surrounding nodes, a plurality of pieces of information received from the surroundings are compared and evaluated. A means for constructing a multicast tree is disclosed. The “physical proximity” can be obtained by measuring the response time of the peer-to-peer connection operation between the nodes. In addition, “survival time” can be defined as the average connection time of the terminal connected to the network in one connection.

  Further, Patent Document 2 discloses a means for building a multicast tree by evaluating the estimated number of hops or the sequence number of an arrival packet at a certain time.

  Further, Patent Document 3 is a multicast transfer route calculation method for obtaining a multicast transfer route from a given start node to a plurality of end nodes in a network composed of a plurality of nodes, and includes network topology information and Using the delay information, a minimum delay path from the start node to the end node is obtained for each end node, and a node on one delay minimum path among a plurality of minimum delay paths to the start node and each end node is determined. , Select as a candidate node of the aggregation point node in multicast forwarding, and for each candidate node, calculate a minimum delay path from the candidate node to the end node for each end node, and a plurality of minimum delay paths for each end node Find the difference between the maximum and minimum values of the delay, and select the candidate node with the smallest error as the aggregation node To select it, and minimum delay paths from the start node to the aggregation point node, and outputs the respective minimum delay path from the aggregation point node to each end node as a multicast forwarding path. According to the above-mentioned patent document 3, it is described that the calculation speed of the multicast transfer route can be improved and the delay dispersion between users can be reduced.

JP 2005-252596 A JP 2004-246790 A JP 2004-208289 A

  Note that the entire disclosure of Patent Documents 1 to 3 is incorporated herein by reference.

  The following analysis is given by the present invention.

  The first problem in the prior art is that in data distribution in a multicast tree, the distribution delay cannot be quickly kept below a specified value.

  The reason is that Patent Documents 1 and 2 do not consider the worst value of the delivery delay. Further, according to Patent Document 3, the worst value of the delivery delay is considered for the time being. However, when a node located in the middle of the root node is dropped during delivery, a rapid multicast tree reconstruction is possible. There is a problem that you can't.

  A second problem in the prior art is that there may be cases where the distribution delay variation in the multicast system cannot be within a certain range.

  In general, nodes participating in a multicast tree have different timings for receiving data depending on positions on the multicast tree. As a result, the dispersion of the distribution delay in the multicast tree becomes large. Patent Document 3 describes that a route is selected so as to suppress a variation in distribution delay, but there is a limit to suppressing a variation in distribution delay by selecting a route.

  The present invention has been made in view of such circumstances, and the first object of the present invention is to participate by quickly constructing a multicast tree that takes into account the intra-node transfer processing delay and the network delay of the transfer link. It is an object of the present invention to provide a multicast system capable of keeping the delivery delay of a node to be within a specified value.

  A second object of the present invention is to provide a multicast system with small variation in distribution delay between nodes participating in the multicast tree.

  According to the first aspect of the present invention, in a multicast tree having a plurality of nodes, the predicted delay amount when assuming that an additional node is newly connected is calculated for each node that can be a connection destination before the additional node is provided. Means for calculating in advance, means for storing the calculated predicted delay amount, and means for determining a node to which the additional node is connected using the predicted delay amount when the additional node is provided. A multicast system is provided.

  According to a second aspect of the present invention, there is provided a multicast system having a multicast tree structure in which a plurality of nodes from a root node are connected in a tree shape, and means for calculating a delay amount from the root node to each node; Means for calculating a delay amount intentionally inserted when data arrives from a difference between a reference delay amount and the calculated delay amount of each node; and the calculated intentional time when data arrives at each node. A multicast system comprising: means for inserting a delay amount.

  According to the third aspect of the present invention, it is assumed that a node is used in a multicast system having a multicast tree structure in which a plurality of nodes are connected in a tree shape, and that a new node is added under its own node. A node of a multicast system is provided that includes an estimated delay amount calculation function unit that calculates a predicted delay amount in advance in this case and a node information storage unit that stores the calculated predicted delay amount.

  According to a fourth aspect of the present invention, there is provided a root node used in a multicast system having a multicast tree structure in which a plurality of nodes are connected in a tree form from a root node, and each of the nodes connected in the tree form An estimated delay amount collecting function unit that collects a predicted delay amount calculated in advance assuming that a new node is added under the node of the own node, and the multicast tree structure using the collected predicted delay amount. There is provided a root node of a multicast system including a topology determination function unit for making a change.

  According to the fifth aspect of the present invention, in a multicast system having a multicast tree in which a plurality of nodes are connected in a tree shape, the estimated delay amount when an additional node is newly connected is provided for the additional node. Before connecting the additional node using the step of calculating in advance for each node that can be a connection destination and the predicted delay amount calculated in advance when the additional node needs to be provided There is provided a multicast tree update method for a multicast system, comprising: determining a node; and, based on the determination, adding an additional node to the multicast tree to perform data distribution.

  According to a sixth aspect of the present invention, there is provided a computer program used in a multicast system having a multicast tree in which a plurality of nodes are connected in a tree shape, and an additional node is newly connected to any of the plurality of nodes. Then, it is assumed that the predicted delay amount when it is assumed is calculated in advance for each node that can be a connection destination before providing the additional node, and stored in the storage unit, and when the additional node needs to be provided, There is provided a computer program of a multicast system for causing at least one computer included in the multicast system to execute a process of determining a node to which the additional node is connected using the predicted delay amount stored in a storage unit. .

  According to the first, third to sixth viewpoints of the present invention, in data distribution in a multicast system, it becomes possible to quickly keep a distribution delay below a specified value.

  Further, according to the second aspect of the present invention, it is possible to keep the distribution delay variation in the multicast system within a certain range.

It is a conceptual diagram which shows (a) the structure of the multicast tree before reconfiguration | reconstruction in the multicast system of the 1st Embodiment of this invention, and (b) the structure of the multicast tree after reconfiguration | reconstruction. It is a conceptual diagram which shows the structural example of the multicast tree in a multicast system. It is a block diagram which shows the internal structure of the root node used for the multicast system by embodiment of this invention. It is a block diagram which shows the internal structure of nodes other than the root node used for the multicast system by embodiment of this invention. It is a conceptual diagram which shows an example of a data structure of the node information database in 1st Embodiment. It is a process flowchart of the root node in 1st Embodiment. It is a process flowchart of nodes other than the root node in 1st Embodiment. It is a block diagram which shows the internal structure of nodes other than the root node used for the multicast system by 2nd Embodiment.

  An outline of an embodiment of the present invention will be described. As shown in FIG. 1, the multicast system according to the embodiment of the present invention has a predicted delay amount (assuming each node in FIG. 1) when it is assumed that an additional node is newly connected in a multi-chast tree having a plurality of nodes. Means for pre-calculating the three items of ID) for each node that can be a connection destination before providing an additional node (for example, 212 in FIG. 4) and means for storing the calculated predicted delay amount (in FIG. 4) 221 and 121) of FIG. Furthermore, a unit (113 in FIG. 3) for determining a node to which the additional node is connected using the predicted delay amount when providing the additional node is provided.

  According to the above-described embodiment, the estimated delay amount when an additional node is newly connected under each node is calculated in advance before making a change to the multicast tree. When necessary, the optimum multicast tree can be changed quickly. Therefore, even if a situation occurs in which a part of an intermediate node is dropped during data distribution, a multicast tree can be quickly reconstructed by keeping the delay amount below a predetermined delay value.

  In addition, in the multicast system according to another embodiment of the present invention, as shown in FIG. 8, a plurality of nodes (representing 200 in FIG. 2 as 200A) are connected in a tree form from the root node (100 in FIG. 2). A multicast system having a multicast tree structure, comprising: means for calculating a delay amount from the root node 100 to each node 200A (212 in FIG. 8); a reference delay amount and the calculated delay amount of each node; A unit 231 that calculates a delay amount that is intentionally inserted when data arrives from the difference, and a unit 232 that inserts an intentional delay amount calculated when data arrives at each node.

  According to the multicast system according to the another embodiment, the delay amount of each node can be accurately aligned.

  It should be noted that the reference numerals attached to the outline of the above embodiment are merely examples for helping understanding, and are not intended to be limited to the illustrated modes. The description of the outline has been completed above, and specific embodiments will be described below in more detail with reference to the drawings.

[First Embodiment]
FIG. 2 is a conceptual diagram illustrating a configuration example of a multicast tree in the multicast system. In FIG. 2, the multicast tree 1 forms a logical network by a root node 100 and a plurality of nodes 200. The root node 100 is connected to at least one node 200, and the node 200 is connected to at least one other node 200 except for the node 200 that is the end of the tree. By branching and connecting from the root node 100 and the node 200 to a plurality of nodes 200, a net having the root node 100 as a vertex spreads in a tree shape, and information transmitted from the root node 100 includes a large number of nodes 200 in a tree shape. Will be communicated to.

  The node 200 is connected to the root node 100 or one or more nodes 200. When each node 200 receives data from the connected root node 100 or node 200, the node 200 transfers the data to the downstream node 200 in the multicast tree 1.

  FIG. 1 is a diagram illustrating the concept of the first embodiment, FIG. 1A is a conceptual diagram illustrating a configuration of a multicast tree before reconfiguration, and FIG. 1B is a configuration of a multicast tree after reconfiguration. FIG. In the multicast tree 1 of FIG. 1, a configuration example is shown in which the configuration of the multicast tree is changed by moving a node in order to keep the delivery delay below a specified value. FIG. 1 (a) is before node movement, and FIG. 1 (b) is after node movement.

  In FIG. 1, the three symbols attached to the root node 100 and the node 200, respectively, separated by a colon, the first symbol at the left end indicates the node ID of each node. The second code at the center is a delivery delay amount from the root node 100 (node ID: n2) to its own node. This delay amount may be an estimated delay amount, or may be an actual measurement value if an actual measurement value is obtained. Furthermore, the third code at the right end is a predicted value of the delivery delay when it is assumed that a new node is directly connected to the own node and added. This predicted value represents the predicted delay amount from the root node to the additional node. This predicted delay amount is a numerical value that serves as a guide when a node newly connected to the multicast tree selects a connection destination.

  The delay amount of a certain node i is D_i, the processing delay amount for distributing data to the jth node under the first generation is DP_ij, and the network delay between the node i and the jth node under the one generation Is DN_ij, the estimated delay amount D_ij of the j-th node one generation below the node i can be expressed by Equation 1.

D_ij = D_i + DP_ij + DN_ij (Formula 1)

Here, P_i0 is the intra-node transfer processing delay from when node i receives the packet from the parent node of node i until it transmits data to the first subordinate node, and P_ij is subordinate j at node i. Deliver data to the jth node under the 1st generation when the intra-node transfer processing delay from sending data to the node under the 1st generation to sending data to the subordinate j + 1st generation node The processing delay amount DP_ij for
DP_ij = DP_i (j-1) + P_i (j-1)
DP_i1 = P_i0
Can be written.

  At this time, if there is no difference in data transmission processing between subordinate nodes in node i and subordinate nodes, Equation 1 can be simplified. Specifically, when it can be assumed that P_i0 = P_i1 = P_i2 = P_i3... And DN_i1 = DN_i2 = DN_i3 = DN_i4.

D_ij = D_i + DP_i * j + DN_i (Formula 2)

  Here, DP_i is a processing delay required for one data transfer in the node i. DN_i is a network delay to a node one generation lower in node i. This value is calculated statically or dynamically from, for example, the actual processing delay, distribution delay, node processing capacity (CPU speed, CPU utilization), the value of a representative subordinate node or one generation higher node. To do. The “processing delay” is a delay inside the node, and the “distribution delay” is a delay of the network connecting the nodes. That is, the calculation is performed using the actual processing delay of each node, the actual delivery delay between the nodes, and the actual processing capability of each node. The “node processing capability” is fixed because it is a machine performance, but the “processing delay” varies depending on the processing status in the machine, and the “delivery delay” varies depending on the network status. The changing parameter may be statically calculated using, for example, an average value, or may be dynamically calculated using a moving average over a fixed time.

  Further, in any node participating in the multicast tree, when DP_1 = DP_2 =... = DP and DN_1 = DN_2 =... = DN, it is possible to further simplify Equation 2 to Equation 3.

D_ij = D_i + DP × j + DN (Formula 3)

  Here, DP is a processing delay required for one data transfer. DN is a network delay to a node one generation below. This value is calculated statically or dynamically from, for example, actual processing delay, delivery delay, node processing capability (CPU speed, CPU utilization), and arbitrary node value.

  In the example of FIG. 1, the estimated delay amount of the root node n2 is set to 0, Expression 3 is applied, and DP = DN = 1. For example, the delay amount of the root node n2 is 0, the delay amount of the subordinate first node n8 is 2, and the delay amount of the node n11 to which the root node n2 transfers data following the node n8 is 3.

  Specifically, the predicted delay amount of a newly connected node is defined as the estimated delay amount of the newly connected node when a new node is connected to that node. For a newly connected node or a node whose connection destination is changed, the delivery delay is smaller when the node is connected to a node with a small (estimated) delay amount of the newly connected node. When the predicted delay amount of a newly connected node is expressed by an expression, the (estimated) delay of a certain node i is assumed that the newly connected node is the n-th child of the node i (a node directly connected under the node). The amount of D_i, the predicted delay amount of the newly connected node of node i, D_in, the predicted value of the processing delay until data is delivered to that node when the new node is connected, DP_in, and the new node connected When the predicted value of the network delay is DN_in, Expression 4 is established.

D_in = D_i + DP_in + DN_in (Formula 4)

  Here, as in the derivation of Equation 1 to Equation 2, Equation 4 can be simplified if there is no difference in data transmission processing between subordinate nodes in node i and subordinate nodes. Specifically, when P_i0 = P_i1 = P_i2 = P_i3... And DN_i1 = DN_i2 = DN_i3 = DN_i4.

D_in = D_i + DP_i * n + DN_i (Formula 5)

  Further, as in the derivation of Equation 2 to Equation 3, Equation 5 is further simplified when DP_1 = DP_2 =... = DP and DN_1 = DN_2 =... = DN can be assumed at any node participating in the multicast tree. Equation 6 can be obtained.

D_in = D_i + DP × (c + 1) + DN (Formula 6)

  Further, in Expressions 1 to 6, D_i is the estimated delay amount of node i. However, when the actually measured delay amount is known, the actually measured delay amount may be used as D_i.

  In the example of FIG. 1, when the estimated delay amount of the root node n2 is set to 0, Equations 3 and 6 are applied, and DP = DN = 1, the estimated delay amount is within the predetermined delay amount 8. The example of a structure is shown. Specifically, in FIG. 1A, the delay amount of the node n9 is 9, and the delay amount of the node n15 is 10, which is larger than the predetermined delay amount 8. Therefore, node recombination is required. Further, in FIG. 1A, there is a node having a predicted delay amount smaller than a predetermined delay amount 8 when an additional node is newly provided under its own node (prediction of nodes n2, n8, n11, n22). The delay amounts are 0, 6, 6, and 7), respectively. Therefore, by changing the connection of the nodes n9 and n15 to any one of the nodes n2, n8, n11, and n22 whose predicted delay amount is less than 8, the delay amount of all the nodes can be kept within the predetermined delay amount 8. It is thought that there is sex. In FIG. 1B, the node n9 is moved under the node n11, and the node n15 is moved under the root node n2.

  As a result, the predicted delay amount of the newly connected additional node of the root node n2 is 4 to 5, the predicted delay amount of the newly connected additional node of the node n11 is 6 to 7, and the estimated delay amount of the node n15 and the new The predicted delay amount of the additional node connected to the node is updated from 10 to 4 and 12 to 6, respectively, the estimated delay amount of the node n9 and the predicted delay amount of the newly connected node are updated from 9 to 6 and 11 to 8, respectively. The (estimated) delay amount of all the nodes does not exceed the specified predetermined delay amount 8. Note that since there is no node connected under the node n23, the predicted delay amount of the additional node of the node n23 is also updated from 11 to 9. In this way, a node whose distribution delay does not fall below the specified value is changed to a node whose distribution delay falls below the specified value based on the predicted delay amount of the additional node, and the multicast tree is reconfigured Thus, it is possible to keep the delivery delay below the specified value in the multicast tree.

  Next, the configuration of the node 200 and the root node of the multicast system that changes the multicast tree according to the first embodiment described with reference to FIG. 1 will be described.

  FIG. 4 is a block diagram showing an internal configuration of a node other than the root node 100 in FIGS. 1 and 2. In FIG. 4, the node 200 includes a main control unit 210, a network interface 220, and a node information database 221. The network interface 220 is connected to the root node 100 or another node 200 via the net.

  The main control unit 210 is a main part of a computer configured by a CPU, a RAM, an OS, and the like. The main control unit 210 includes a transmission / reception unit 211 configured as an operating program and an estimated delay amount calculation function. Part 212.

  FIG. 5 is a conceptual diagram showing an example of the internal data configuration of the node information database 221 in FIG. As an example of the data, the node information database 221 of the node n3 in FIG. 2 is illustrated. The node information database 221 stores node information in the multicast tree 1. The node information is information necessary for transmitting / receiving a message to / from another node 200 such as a node ID, an IP address, and a port number that uniquely identifies the node 200. In the tree structure of the multicast tree 1, This is information of the node 200 below. The node information database 221 also stores (estimated) delay amounts of each node and predicted delay amounts of newly added additional nodes.

  The estimated delay amount calculation function unit 212 calculates the (estimated) delay amount of the node 200 connected under the multicast tree and the predicted delay amount of the additional node newly connected to the own node in the multicast tree. For calculating the (estimated) delay amount of the node 200 connected to the subordinate, Expression 1 or Expression 2 is used. Further, Formula 4 or Formula 5 is used to calculate the predicted delay amount of the newly connected additional node. However, since the additional node that is assumed to be newly connected has not yet been identified, the estimated value DN_in of the network delay is a virtual value. The calculation result is stored in the node information database 221, and the estimated delay amount of the node 200 connected to the subordinate is notified to the corresponding subordinate node 200.

  The estimated delay amount calculation function unit 212 calculates the (estimated) delay amount and the predicted delay amount of the newly connected additional node by calculating the (estimated) delay amount from the root node 100 and the predicted estimated delay amount of the newly connected node. This is performed when a collection request is received, when there is a change such as increase or decrease in the subordinate node 200, or when there is an instruction from an operator of the node 200.

  FIG. 3 is a block diagram showing an internal configuration of the root node 100 used in the multicast system according to the first embodiment. In FIG. 3, the root node 100 includes a main control unit 110, a network interface 120, and a node information database 121. The main control unit 110 is a main part of a computer configured with a CPU, a RAM, an OS, and the like. The root node 100 includes a transmission / reception unit 111 configured as an operating program, an estimated delay amount collection function unit 112, and a topology determination function unit 113 in the main control unit 110.

  The node information database 121 is a database corresponding to the node information database 221 of the node 200 already described with reference to FIG. However, in the case of the node information database 121 of the root node 100, the node information of each node collected from each node by the estimated delay amount collection function unit 112, the (estimated) delay amount, and an additional node is newly connected to each node The predicted delay amount is also stored.

  The transmission / reception unit 111 receives a message from the node 200 via the network interface 120, checks the message type, and sends the message to the estimated delay amount collection function unit 112 or the topology determination function unit 113.

  Further, the transmission / reception unit 111 receives a message from the estimated delay amount collection function unit 112 or the topology determination function unit 113, and transmits the message to the node 200 that is the destination described in the message via the network interface 120.

  The estimated delay amount collection function unit 112 collects each (estimated) delay amount and the predicted delay amount of the newly added additional node from the nodes participating in the multicast tree 1 and updates the node information database 121. (Estimated) As a method of collecting the delay amount and the predicted delay amount of the newly connected node, for example, the root node 100 requests the subordinate node 200 for the estimated delay amount and the predicted delay amount of the newly connected additional node. The node 200 that has received the message requesting the estimated delay amount and the predicted delay amount of the newly connected additional node further forwards the message to the subordinate node 200, and determines the position of the leaf of the multicast tree 1. There is a method in which the (estimated) delay amount from the node 200 and the predicted delay amount of a newly connected node are returned in reverse order. Alternatively, the root node 100 may directly communicate with the node 200 connected to the multicast tree 1 to collect (estimated) delay amount and the predicted delay amount of a newly connected node. When the actually measured delay amount is known, the actually measured delay amount may be used instead of the estimated delay amount.

  The topology determination function unit 113 refers to the node information database 121 and refers to the (estimated) delay amount of the node 200 participating in the multicast tree 1 and the predicted delay amount of the newly connected node. Next, the node 200 having a delay amount larger than the specified delay amount (estimated) is extracted. The node group extracted here is referred to as a node group A. Further, the node 200 having the predicted delay amount of the newly connected node smaller than the designated delay amount is extracted. The node group extracted here is referred to as a node group B. That is, the node of the node group A is a node whose current (estimated) delay amount is larger than the specified reference delay amount, and the node of the node group B is newly added under the node The predicted delay amount is a node having a delay amount smaller than a designated reference delay amount.

  At this time, if one or more nodes 200 are included in the node group A, the node group A is connected under the node group B.

  Which node 200 of the node group A is connected to which node 200 of the node group B is connected to, for example, a node with a small estimated delay amount of a newly connected node in order of the node having the largest predicted delay amount. it can. Further, it may be connected under the node whose position on the multicast tree 1 is as small as possible. Although other methods may be used, the maximum value of the delay amount can be reduced by connecting a node having a large delay amount to a node having a predicted delay amount smaller than the current delay amount of the node. When the measured delay amount is known by using a time stamp or the like, the actually measured delay amount may be used instead of the estimated delay amount when extracting the node group A.

  The topology determination function unit 113 determines the destination of the node and then transmits a movement message to the node 200 to be moved.

  Here, the delivery delay time is specified statically or dynamically by the multicast tree operator. There is a lower limit in the designation of the delivery delay time depending on the number of nodes 200 participating in the multicast tree 1. When a value smaller than the lower limit is specified, the number of nodes in the node group A is larger than the number of nodes in the node group B, and there are nodes 200 whose distribution delay is not less than the specified value. At this time, a method of notifying the node of an alarm or forcibly leaving the multicast tree 1 is employed.

  In addition, the topology determination function unit 113 provides the information of the predicted delay amount of the additional node newly connected to the node 200 already participating in the multicast tree to the node 200 newly participating in the multicast tree. The node 200 newly participating in the multicast tree selects the node 200 as a connection destination that does not exceed the specified delivery delay time from the obtained information of the predicted delay amount of the newly connected node, and connects to the multicast tree by connecting. Join the tree. At this time, if there is no connection destination that does not exceed the specified delivery delay time, a connection is made after raising the alarm, or the connection is abandoned.

  The estimated delay amount collection function unit 112 collects the (estimated) delay amount and the predicted delay amount of the newly connected node, and the topology determination function unit 113 determines the topology and moves the node. This is performed when there is a change such as increase or decrease in 200 or when there is an instruction from the operator of the root node 100.

  Next, a particularly characteristic part of the operation of the first embodiment will be described.

  FIG. 6 is a flowchart showing an operation in the case where the topology is changed so that the delivery delay in the multicast tree 1 falls within a certain value in the root node 100 in the present embodiment. When the operation is started, first, the transmission / reception unit 111 enters a state of waiting for communication from another node 200 or the main control unit 110 in the root node 100 (hereinafter referred to as an idle state) (step S300).

  When the root node 100 is in an idle state (step S300), when the node 200 connected to the multicast tree 1 changes such as increase or decrease, or when an instruction from the operator of the root node 100 is given, the multicast tree 1. Collect the (estimated) delay amount and the predicted delay amount of the newly connected node from the node 200 participating in 1 (step S301), and update the node information database 121 (step S302).

  Subsequently, the node information database 121 is referred to, and a node 200 having a larger (estimated) delay amount than the designated delay amount is extracted (step S303). Further, the node information database 121 is referred to, and a node 200 having a predicted delay amount of the additional node smaller than the designated delay amount is extracted (step S304). Further, the node extracted in step S303 is connected to the node extracted in step S304 (step S305). Next, the number of nodes extracted in step S303 is compared with the number of nodes extracted in step S304 (step S306). If the number of nodes extracted in step S304 is larger, the process returns to the idle state in step S300. . If the number of nodes extracted in step S303 is larger (transition down to step S306), an alarm is transmitted to the node 200 that could not move (step S307), and the state returns to the idle state.

  The operation content of each step according to the flowchart described in FIG. 6 can be configured to be executed as a program that operates on a computer (main control unit) provided in advance in the root node 100.

  In the above description, the reconfiguration of the existing multicast tree 1 has been described. However, when a new node newly joins the multicast tree 1, it connects to which node 200 that already configures the multicast tree 1. When determining whether or not, the predicted delay amount of the newly connected node is collected from each node 200 participating in the multicast tree 1, and the predicted delay amount of the additional node is small with reference to the predicted delay amount of the additional node. By connecting to, the delivery delay of the multicast tree 1 can be kept below a specified value. At this time, if there is no connection destination that can keep the delivery delay of the multicast tree 1 below the specified value, it is possible to refuse new participation or allow participation after notifying an alarm.

  FIG. 7 is a process flowchart of the nodes 200 other than the root node in the first embodiment. When the operation is started, first, the transmission / reception unit 211 enters a state of waiting for communication from another node 200 or the main control unit 110 in the root node 100 (hereinafter referred to as an idle state) (step S400). When the connection of the node changes according to an instruction from the root node (Yes in step S401), the topology information in the node information database 221 is updated (step S402). In step S402, when the actual measured value of the delay value is obtained using a time stamp or the like, the estimated delay value may be replaced with the actual measured value.

  In step S403, when there is a query of the node information database 221 from the root node 100 or a node (parent node) interposed between the node and the root node (Yes in step S403), a response is made to the content of the query. Then, the contents of the latest node information database 221 are answered to the inquiry (step S404), and then the process returns to the idle state of step S400. On the other hand, if there is no inquiry in the node information database 221 (No in step S403), nothing is done and the process returns to the idle state in step S400.

  In the above description, the root node 100 collects the (estimated) delay amount and the predicted delay amount of the newly connected node from the node 200 that participates in the multicast tree 1, and determines the topology of the multicast tree 1. The root node 100 may not have this function. That is, it may be a certain node 200 or a third computer. In this case, the root node 100 also operates as a node similar to the node 200.

[Second Embodiment]
Next, a second embodiment of the present invention will be described with reference to the drawings.

  In the second embodiment of the present invention, by introducing a delay when data arrives on the basis of a specified estimated delay amount, variation in distribution delay among nodes participating in the multicast tree can be reduced.

  The concept of the second embodiment will be described with reference to FIG. In FIG. 1A, the (estimated) delay amount of the node n8 is 2, and the (estimated) delay amount of the node n22 is 5. At this time, if the node n8 delays inside the node by the amount of (estimated) delay amount 3 at the time of data reception, it corresponds to the estimated delay amount 5, and the difference in distribution delay between the node n8 and the node n22 becomes small.

  Further, when the delay amount 10 is used as a reference, the node n8 has a delay amount 8 delay, the node n22 has a delay amount 5 delay, and the node n11 has a delay amount 7 delay. If the difference is delayed within the node at the time of data reception, the distribution delay of data reception becomes substantially equal at all nodes. Note that a node to which another node is connected is immediately transferred to the subordinate node when the data arrives without inserting a delay for the data to be transferred to the subordinate node. For example, the node n8 transfers data to the node n3 and the node n44, and the node n11 transfers data to the node n22, but no delay is inserted into the data transferred to the other nodes. By performing such processing, it is possible to reduce the variation in the delivery delay at the end of the delivery destination. Expressed as an expression, if the delay amount of the node i is D_i, the reference delay amount is DB, and the intentionally inserted delay amount is ΔD_i, Expression 7 is established.

ΔD_i = DB−D_i (Formula 7)

  At this time, the reference delay amount needs to be the same as or larger than the maximum delay amount among the nodes 200 participating in the multicast tree 1. In order to determine the delay amount as a reference, it is sufficient to know the maximum delay amount among the nodes 200 participating in the multicast tree 1. The maximum delay amount among the nodes 200 participating in the multicast tree 1 is obtained in the process of determining the topology. However, when it is difficult to obtain information on delay amounts (measured or estimated) of all nodes, for example, when the number of nodes 200 participating in the multicast tree is enormous, among the nodes 200 participating in the multicast tree 1 Estimate the maximum amount of delay. The method includes, for example, estimating from the number of nodes 200 participating in the multicast tree, and estimating from the estimated delay amount of a node at the lowest layer of the multicast tree.

  In addition, since it is preferable that the amount of delay to be inserted when data arrives is small, the multicast tree 1 is preliminarily joined using the means of the first embodiment of the present invention before obtaining the amount of delay to be intentionally inserted. It is preferable to reduce the maximum estimated delay amount in the node 200 to be operated. Specifically, the maximum estimated delay amount among the nodes 200 participating in the multicast tree 1 can be reduced by setting the estimated delay amount as a lower limit and changing the topology of the multicast tree 1.

  Also, the reference delay amount can be made known by, for example, sending it as the first data. By making the delay amount used as a reference known to the nodes participating in the multicast tree, a maximum delay amount or a value larger than the maximum delay amount is given in advance as a reference value, and intentionally inserted based on that reference value. The delay amount ΔD_i can be calculated.

  FIG. 8 is a block diagram showing an internal configuration of a node other than the root node used in the multicast system according to the second embodiment. The node 200A in FIG. 8 is used in place of the node 200 of the first embodiment. A configuration of the node 200A in FIG. 8 that is different from the node 200 of the first embodiment described with reference to FIG. 4 will be described. Parts having the same configuration as that of the node 200 of the first embodiment are denoted by the same reference numerals, and redundant description is omitted.

  The insertion delay amount calculation unit 231 receives the designation of the reference delay amount from the root node 100, and calculates the delay amount to be inserted at the time of data reception based on the (estimated) delay amount of the own node included in the node information database. To do.

  When receiving data distribution from the root node 100, the delay insertion unit 232 delays the data received by the transmission / reception unit based on the insertion delay amount calculated by the insertion delay amount calculation unit 231.

  Further, the root node 100 can use the root node 100 of the first embodiment as it is, except for transmitting the reference delay amount to each node 200A in the initial data distribution.

  In the above description, the distribution delay variation can be reduced by notifying the estimated delay amount as one reference in the entire multicast tree 1, but the nodes participating in the multicast tree 1 can be divided into a plurality of groups. Dividing, setting an estimated delay amount as a reference in each group, calculating a delay amount ΔD_i intentionally inserted from the estimated delay amount as a reference in the group, and inserting a delay when data arrives, Variation in distribution delay among the nodes can be reduced.

  Further, in the above description, it is assumed that the dispersion of the distribution delay can be reduced by making the estimated delay amount as a reference in the entire multicast tree 1 known. However, nodes 200 having the same or similar estimated delay amount are grouped. By doing so, it is possible to reduce the variation in the delivery delay within the group.

  Taking FIG. 1A as an example, the node n44 and the node n22 both have an estimated delay amount 5. When these nodes are grouped, the estimated delay amount in the group is the same value, so that the variation in delivery delay within the group can be reduced.

  As described above, the estimated delay amount is calculated from the sum of the intra-node transfer processing delay and the network delay of the transfer link, and the value specified by the delivery delay based on the estimated delay amount and the estimated delay amount of the newly connected node The delay of nodes participating in the multicast tree is moved by moving the nodes that are estimated to be less than the following before the start of the multicast service so that the delivery delay falls below the specified value, and reconfiguring the multicast tree. Can be kept below the specified value.

  In addition, as described above, the estimated delay amount is calculated from the sum of the intra-node transfer processing delay and the network delay of the transfer link, and the delay when data arrives based on the estimated delay amount and the estimated delay amount of the newly connected node. By inserting, distribution delay variation among nodes participating in the multicast tree can be reduced.

In the present invention, the following modes are possible.
[Form 1]
As in the multicast system according to the first aspect.
[Form 2]
The means for determining a node to which the additional node is connected connects the additional node to a node whose predicted delay amount is a predetermined delay value or less,
When there is no node whose predicted delay amount is equal to or smaller than the predetermined delay amount, the multicast tree is set so that the delay amounts of the plurality of nodes including the additional node are all equal to or smaller than the predetermined delay amount. It is preferable to rearrange.
[Form 3]
Each of the nodes is
Means for calculating a predicted delay amount assuming that an additional node is connected downstream of the own node;
Means for storing the calculated predicted delay amount,
The multicast system includes means for collecting and storing the predicted delay amount of each node calculated by each node in a batch,
The means for determining the node preferably determines a node to which the additional node is connected using the predicted delay amount of each node stored in a batch.
[Form 4]
The means for calculating the expected delay amount assumes that the processing capability of the node that can be the connection destination, the delay amount from the root node to the node, and the network delay from the node to the additional node are average. It is preferable to calculate the expected delay amount based on the virtual network delay amount.
[Form 5]
As in the multicast system according to the second aspect.
[Form 6]
As a node of the multicast system according to the third aspect.
[Form 7]
The multicast tree structure is a multicast system in which the plurality of nodes are connected in a tree form from a root node;
It is preferable that the node including the estimated delay amount calculation function unit and the node information storage unit is the node other than the root node.
[Form 8]
As the root node of the multicast system according to the fourth aspect.
[Form 9]
The multicast tree update method of the multicast system according to the fifth aspect.
[Mode 10]
As described in the computer program of the multicast system according to the sixth aspect.
[Form 11]
A multicast system comprising a multicast tree, and means for calculating, for each connected node, an estimated delay amount of the node after joining before joining the multicast tree.
[Form 12]
A multicast system constituted by a multicast tree, which is estimated from a node constituting a multicast tree, a means for calculating an estimated delay amount of a node after joining before joining the multicast tree, and a node constituting the multicast tree. From the estimated or measured delay amount, the means to collect the measured delay amount and the estimated delay amount when it is assumed that a new node is connected under the node, and the node whose distribution delay does not fall below the specified value And a means for relocating a multicast tree by moving the node under a node whose distribution delay falls below a specified value.
[Form 13]
A multicast system composed of a multicast tree, for calculating an estimated delay amount of a node after joining for each connected node before joining the multicast tree, and a node newly joining the multicast tree A means for notifying an estimated delay amount when it is assumed that a node is newly connected under the nodes constituting the multicast tree; and a node that is a connection destination within which the delivery delay falls within a specified value from the estimated delay amount A multicast system comprising: means for selecting
[Form 14]
14. A multicast system comprising a multicast tree according to any one of forms 11 to 13, wherein when calculating an estimated delay amount of a certain node, an actual processing delay, distribution delay, and node of an arbitrary node Using the processing capacity (CPU speed, CPU utilization), calculating a processing delay amount for distributing data to a node under one generation, and a network delay between the node and the node under the one generation, It is preferable that the multicast system calculates an estimated delay amount of a certain node from the accumulated processing delay amount and the network delay amount.
[Form 15]
The multicast system configured by the multicast tree according to the fourteenth aspect, wherein when calculating an estimated delay amount of a certain node, an average actual processing delay, distribution delay of all the nodes configuring the multicast tree, Using the processing capacity of the node (CPU speed, CPU utilization), calculate a processing delay amount for distributing data to a node under one generation and a network delay between the node and the node under the one generation, It is preferable that the multicast system calculates an estimated delay amount of a certain node from the accumulation of the processing delay amount and the network delay amount.
[Form 16]
15. A multicast system comprising a multicast tree according to the fourteenth aspect, wherein when an estimated delay amount of a certain node is calculated, an actual processing delay, a delivery delay, and a node processing of an adjacent node in the multicast tree Using the capacity (CPU speed, CPU utilization), a processing delay amount for distributing data to a node under one generation and a network delay between the node and the node under the one generation are calculated, and the processing delay It is preferable that the multicast system calculates the estimated delay amount of a certain node from the amount and the accumulated network delay amount.
[Form 17]
15. The multicast system according to the fourteenth aspect, including a multicast tree, and when calculating the estimated delay amount of a certain node, assuming that the processing delay and the delivery delay are equal, data is transmitted to a node one generation lower A multicast system for calculating a processing delay amount for distributing a network, a network delay between the node and a node one generation lower than the node, and calculating an estimated delay amount of a node from the accumulation of the processing delay amount and the network delay amount It is preferable that
[Form 18]
The multicast system configured by the multicast tree according to the fourteenth aspect described above, in which when the estimated delay amount of a certain node is calculated, the actual processing delay, distribution delay, and node Using an average value weighted by processing capacity (CPU speed, CPU utilization), the amount of processing delay for distributing data to a node under one generation, and the network delay between the node and the node under the one generation Preferably, the multicast system calculates the estimated delay amount of a certain node from the accumulated processing delay amount and the network delay amount.
[Form 19]
14. A multicast system comprising a multicast tree according to any one of forms 11 to 13, wherein an estimated delay amount of a node after participation is calculated for each connection destination node before joining the multicast tree. In addition, using the actual processing delay, distribution delay, and node processing capability (CPU speed, CPU utilization) of any node, processing delay until data is distributed to that node when a new node is connected The estimated value and the estimated value of the network delay when the new node is connected are calculated, and the newly connected node is newly connected from the accumulation of the estimated value of the processing delay and the estimated value of the network delay. A multicast system that calculates an estimated delay amount of a node is preferable.
[Mode 20]
The multicast system according to the nineteenth aspect, which includes a multicast tree, and calculates the estimated delay amount of a node after joining for each connected node before joining the multicast tree. The processing delay amount for distributing data to a node one generation lower, using the average, actual processing delay, distribution delay, and node processing capacity (CPU speed, CPU utilization rate) of all the nodes constituting It is preferable that the multicast system calculates a network delay between a node and a node that is one generation lower than the node, and calculates an estimated delay amount of a node from the accumulated processing delay amount and the network delay amount.
[Form 21]
The multicast system configured by a multicast tree according to the nineteenth aspect, wherein the estimated delay amount of the node after participation is calculated for each connection destination node before joining the multicast tree. Estimated processing delay until data is distributed to a new node connected using the actual processing delay, distribution delay, and node processing capability (CPU speed, CPU utilization) of the adjacent node And an estimated value of the network delay when the new node is connected, and from the accumulation of the estimated value of the processing delay and the estimated value of the network delay, the newly connected node is connected before the newly connected node is connected. A multicast system that calculates an estimated delay amount is preferable.
[Form 22]
The multicast system configured by a multicast tree according to the nineteenth aspect, wherein the estimated delay amount of the node after participation is calculated for each connection destination node before joining the multicast tree. Estimated processing delay until data is distributed to a new node connected using the actual processing delay, distribution delay, and node processing capability (CPU speed, CPU utilization) of the adjacent node And an estimated value of the network delay when the new node is connected, and from the accumulation of the estimated value of the processing delay and the estimated value of the network delay, the newly connected node is connected before the newly connected node is connected. A multicast system that calculates an estimated delay amount is preferable.
[Form 23]
The multicast system comprising the multicast tree according to the nineteenth aspect, wherein when calculating the estimated delay amount of the node after joining before joining the multicast tree for each connection destination node, processing delay and Assuming that delivery delays are equal, an estimated value of processing delay until data is delivered to that node when a new node is connected and an estimated value of network delay when a new node is connected are calculated, It is preferable that the multicast system calculates an estimated delay amount of a newly connected node before the newly connected node is connected from the accumulated delay estimated value and the estimated network delay value.
[Form 24]
The multicast system according to the nineteenth aspect, which includes a multicast tree, and calculates the estimated delay amount of a node after joining for each connected node before joining the multicast tree. Using the average value weighted by the actual processing delay, distribution delay, and node processing capacity (CPU speed, CPU utilization) of all the nodes to configure, until data is distributed to that node when a new node is connected The estimated value of the processing delay and the estimated value of the network delay when the new node is connected are calculated, and the accumulated value of the estimated value of the processing delay and the estimated value of the network delay is calculated before the newly connected node is connected. A multicast system that calculates an estimated delay amount of a newly connected node is preferable.
[Form 25]
Means for collecting the estimated or measured delay amount and the estimated delay amount of the newly connected node from the nodes constituting the multicast system and participating in the multicast tree, and the collected estimated or actually measured delay amount And a node having an estimated or actually measured delay amount larger than the specified estimated delay amount from the estimated delay amount of the newly connected node and a newly connected node having an estimated delay amount smaller than the specified estimated delay amount And means for holding node information including an estimated or measured delay amount of a node participating in the multicast tree and an estimated delay amount of a newly connected node.
[Form 26]
A node constituting a multicast system, a means for calculating an estimated or measured delay amount of a node connected to the subordinate system and an estimated delay amount of a newly connected node of the own node; an estimation or actual measurement of the own node; A node that holds node information including a delay amount and an estimated delay amount of a newly connected node.
[Form 27]
A multicast system composed of a multicast tree, which is designed to calculate the estimated delay amount of the nodes constituting the multicast tree and the difference between the reference delay amount and the estimated or actually measured delay amount when data arrives. A multicast system comprising means for calculating a delay amount to be inserted automatically and means for inserting the intentional delay amount upon arrival of data.
[Form 28]
28. A multicast system comprising a multicast tree according to the embodiment 27, wherein the reference value is the worst value of the actually measured delay amount when calculating the delay amount intentionally inserted when data arrives. A system is preferred.
[Form 29]
28. A multicast system comprising a multicast tree according to the embodiment 27, wherein the reference value is a maximum value of the estimated delay amount when calculating the delay amount intentionally inserted when data arrives. A system is preferred.
[Form 30]
28. The multicast system configured by a multicast tree according to the above-described aspect 27, wherein a position at which an estimated delay amount is likely to be large when calculating a delay amount to be intentionally inserted when data arrives. It is preferable that the multicast system be an actual measurement value of a node in
[Form 31]
28. A multicast system comprising a multicast tree as set forth in the embodiment 27, wherein means for dividing a node participating in the multicast tree into a plurality of groups, and an estimated or actually measured delay amount for each group are set. A multicast system, means for calculating a delay amount intentionally inserted from an estimated or actually measured delay amount as a reference in the group, and a means for inserting a delay when data arrives by the amount of the delay amount It is preferable that
[Form 32]
It is preferable that the multicast system includes a multicast tree described in the form 27, and has a means for grouping nodes having the same or close delay amounts estimated or actually measured.
[Form 33]
A multicast system comprising a multicast tree, the step of collecting an estimated or actually measured delay amount and an estimated delay amount when a node is newly connected under the node from a node constituting the multicast tree Extracting a node having a delay amount larger than the designated estimated or actually measured delay amount, and extracting a node having an estimated delay amount of a newly connected node smaller than the designated estimated or actually measured delay amount; Moving a node whose distribution delay does not fall below a specified value to a subordinate of a node whose distribution delay falls below a specified value from an estimated or measured delay amount, and reconfiguring a multicast tree. Casstree construction method.
[Form 34]
Estimated when it is assumed that a computer that controls a node constituting a multicast system composed of a multicast tree is assumed to have newly connected a node under the estimated or actually measured delay amount from the node constituting the multicast tree. A function that collects the delay amount, a function that extracts a node whose delay amount is larger than the specified estimated or actually measured delay amount, and an estimated delay amount of a newly connected node that is smaller than the specified estimated or actually measured delay amount. The function to extract nodes and the nodes whose distribution delay does not fall below the specified value are moved from the estimated or measured delay amount to the nodes under the distribution delay below the specified value, and the multicast tree is reconfigured A function that executes a function.
[Form 35]
A means for specifying a reference delay amount;
At least some of the plurality of nodes have means for obtaining a difference between the reference delay amount and its own delay amount, and means for inserting a delay amount that fills the difference when data arrives at the node It is preferable that it is a multicast system in any one of form 1 thru | or 3 provided with these.
[Form 36]
The estimated delay amount calculation function unit includes a function of collecting predicted delay amounts calculated by other nodes included in the multicast tree;
The node of the multicast system according to aspect 6 or 7, further comprising a topology determination function unit that reconstructs the multicast tree using the collected predicted delay amount.
[Form 37]
The estimated value delay amount collection function unit also has an estimated delay amount calculation function unit that calculates in advance a predicted delay amount when it is assumed that a node is newly added under its own node,
A node information storage unit that stores the current topology of the multicast tree structure including the predicted delay amount accumulated by the estimated delay amount collection function unit;
It is preferable that the topology determination function unit is a root node of the multicast system according to the eighth aspect in which the multicast tree structure is changed using node information of the entire multicast tree stored in the node information storage unit.
[Form 38]
In each of the nodes, which is a data distribution destination and also relays data transfer to another node, the means for inserting the intentional delay amount includes data to be relayed to the other node. It is preferable that the multicast system of the fifth aspect performs data transfer promptly after receiving data without inserting the delay amount.

  Within the scope of the entire disclosure (including claims and drawings) of the present invention, the examples and the examples can be changed and adjusted based on the basic technical concept. In addition, various combinations or selections of various disclosed elements are possible within the scope of the claims of the present invention. That is, the present invention naturally includes various modifications and changes that could be made by those skilled in the art according to the entire disclosure including the claims and the drawings, and the technical idea.

1: Multicast tree 100: Root node 200, 200A: Node 110, 210: Main control unit 120, 220: Network interface 121, 221: Node information database 111, 211 Transmission / reception unit 112: Estimated delay amount collection function unit 113: Topology determination Function unit 212: Estimated delay amount calculation function unit 231: Insertion delay amount calculation unit 232: Delay insertion unit

Claims (9)

In a multicast tree having a plurality of nodes, a means for calculating in advance a prediction delay amount when assuming that an additional node is newly connected for each node that can be a connection destination before providing the additional node;
Means for storing the calculated predicted delay amount;
Means for determining a node connecting the additional node using the predicted delay amount when providing the additional node;
A multicast system comprising: The means for determining a node to which the additional node is connected connects the additional node to a node whose predicted delay amount is a predetermined delay value or less,
When there is no node whose predicted delay amount is equal to or smaller than the predetermined delay amount, the multicast tree is set so that the delay amounts of the plurality of nodes including the additional node are all equal to or smaller than the predetermined delay amount. The multicast system according to claim 1, wherein the multicast system is rearranged. Each of the nodes is
Means for calculating a predicted delay amount assuming that an additional node is connected downstream of the own node;
Means for storing the calculated predicted delay amount,
The multicast system includes means for collecting and storing the predicted delay amount of each node calculated by each node in a batch,
The multicast system according to claim 1 or 2, wherein the means for determining the node determines a node to which the additional node is connected using the predicted delay amount of each node stored in a batch.   The means for calculating the expected delay amount assumes that the processing capability of the node that can be the connection destination, the delay amount from the root node to the node, and the network delay from the node to the additional node are average. 4. The multicast system according to claim 1, wherein an expected delay amount is calculated based on the virtual network delay amount. A node used in a multicast system having a multicast tree structure in which a plurality of nodes are connected in a tree shape,
An estimated delay amount calculation function unit for calculating in advance a predicted delay amount assuming that a new node is added under the own node;
A node information storage unit for storing the calculated predicted delay amount;
A node of a multicast system comprising: The multicast tree structure is a multicast system in which the plurality of nodes are connected in a tree form from a root node;
The node of the multicast system according to claim 5 , wherein the node including the estimated delay amount calculation function unit and the node information storage unit is the node other than the root node. A root node used in a multicast system having a multicast tree structure in which a plurality of nodes are connected in a tree form from the root node;
An estimated delay amount collection function unit that collects a predicted delay amount calculated in advance on the assumption that a node is newly added under the own node from each of the nodes connected in the tree shape;
A topology determination function unit for changing the multicast tree structure using the collected predicted delay amount;
A root node of a multicast system, comprising: In a multicast system having a multicast tree in which a plurality of nodes are connected in a tree shape, the estimated delay amount when assuming that an additional node is newly connected is set in advance for each node that can be a connection destination before providing the additional node. A step to calculate,
When it is necessary to provide the additional node, using the predicted delay amount calculated in advance, determining a node to connect the additional node;
Based on the determination, incorporating additional nodes into the multicast tree for data distribution;
A multicast tree update method for a multicast system, comprising: A computer program used in a multicast system having a multicast tree in which a plurality of nodes are connected in a tree shape,
A process of calculating a predicted delay amount when it is assumed that an additional node is newly connected to any of the plurality of nodes, for each node that can be a connection destination before providing the additional node, and storing the estimated delay amount in a storage unit; ,
When it becomes necessary to provide the additional node, a process for determining a node to which the additional node is connected using the predicted delay amount stored in the storage unit;
Is executed by at least one computer included in the multicast system.

Images