CN1710885A - A method of QoS multicast routing based on Ant-Like mobile agent - Google Patents

Abstract

Implementing mobile agent under network simulator NS2, the invention includes steps: (1) building network model; multicasting tree T(VT, ET) satisfies following conditions: time delay constraint delay(P(s, m)) less than and equal to D, Pm is a member of the set M; bandwidth constraint bandwidth(P(s, m)) larger than and equal to B, e is a member of the P(s, t), where P as route, s as source node, m as target node, D as maximal time delay, B as minimal bandwidth; (2) expanding NS2 in following four sides: 1, expansion of class i.e. mobile agent MA classifier is expanded in address classifier; 2, expansion of packet i.e. MA head is expanded in packet head of NS2; 3, expansion of node, i.e. special data packet including mobile agent is added in node structure of original NS2; 4, expansion of agent, i.e. implementation environment MAE of mobile agent is built and loaded to node.

Description

A QoS Multicast Routing Method Based on Ant-Like Mobile Agent

technical field

The invention relates to a QoS multicast routing method (MAMRQoS) based on Ant-Like mobile agent, which belongs to the field of network technology, communication technology and multimedia application technology (the meaning of QoS is quality of service).

Background technique

The ant algorithm is a population-based simulated evolutionary algorithm. After a lot of research, it has been found that ants have the ability to find the shortest (optimal) path between the ant nest and the food source. A volatile secretion pheromone (Pheromone) is realized, and the probability that the ant chooses the path when advancing on a path is proportional to the intensity of the pheromone on this path. For the shortest path, more and more ants choose it, resulting in more pheromone accumulation, attracting more ants, thus forming a positive feedback, so that the shortest path will be found by most ants soon, the ants The exchange of information and mutual cooperation between individuals achieve the purpose of searching for food.

AntNet is a distributed adaptive shortest path calculation method based on mobile agents, which originated from the famous ant colony algorithm in bionics. The basic idea of AntNet is: any node in the network is the source node, and a randomly selected node is the destination node at a certain time interval, and a mobile agent is sent as the forward agent to search for the shortest path between the two nodes, and the records are collected at the same time. Path state information, the agent selects the next hop node at each intermediate node according to the ratio of the orientation probability of the adjacent node in the routing table or randomly, and records the identifier of the node and the travel experience from the source node to the node Time, repeating this process until reaching the destination node, the forward agent that reaches the destination node generates a backward agent, and passes all the collected information to the agent and then dies, the backward agent moves along the opposite direction of the same path, and uses its The information carried updates the parameters such as the orientation probability value and travel time in the routing table of the passing node, and the simulation experiment proves that its performance is better than the traditional link state algorithm and distance vector algorithm, but this algorithm is only for best-effort However, for the datagram delivery, it does not consider the routing calculation problem of real-time data flow with service quality requirements, which has certain limitations.

Agent-based routing algorithm is an application of artificial intelligence in the network. Appleby and Steward initially proposed that an algorithm similar to ant agent can be used for network control such as routing and load balancing, but they did not give a prototype. Distributed QoS (Quality of Service) routing calculation method based on mobile agent. This method uses mobile agent to search the network to find the path that meets the QoS request. After introducing mobile agent, the system is more flexible and adaptable.

Ant routing is a novel routing algorithm, which is derived from the model of ant colony searching for food in biology. It is called Agents in ant routing. A group of ants is used to explore the shortest path between a specified source node and destination node. For a destination node, instead of having a fixed next-hop node in the routing table, it has A plurality of optional next-hop nodes, each candidate next-hop node has a probability value, indicating the possibility of the candidate value being selected in the case of calculating the shortest path. The probability values of these candidate nodes are equal at the beginning, and are updated by the passing ant packets at any time during the operation of the algorithm. For a pair of given source node and destination node, the source node will generate a certain number of ants, and send these ant packets to the network according to its own routing table entries, and they will complete the task of exploring the network to find the path. Ants can remember the path nodes they have passed, so when an ant reaches the destination node, it will be able to return to the source node along the source path. In the process of returning, the ant is used to changing its path according to the following rules according to the information state it carries The routing table of each node. The update rule of the routing table is to increase the probability value of the candidate node that the ant comes from in the routing table of the current node, and reduce the probability value of other candidate nodes. The path with a higher probability value is always preferred, which increases the path With the selection probability of , more and more ants will follow this path, further increasing the probability value, and eventually more ants will follow this path. It is because of this positive feedback loop that a best path quickly Appears, and when the network load changes and a new best path appears, ants will quickly recognize and strengthen it, so ant routing is dynamic, adaptive, and has good scalability and flexibility.

None of the above is used to solve the constraint problem of QoS multicast routing.

Contents of the invention

The purpose of the present invention is to solve the constraint problem of QoS multicast routing, propose a kind of method based on the QoS multicast routing of Ant-Like mobile agent, use it to solve time delay, time delay vibration, broadband and packet loss in real-time applications rate constraints.

Technical scheme of the present invention: the present invention provides a kind of method based on the QoS multicast routing of Ant-Like mobile agent, realizes mobile agent under network simulator NS2, is characterized in that:

1. Establish a network model: the multicast tree T (VT, ET) satisfies:

Delay constraint, delay(P(s,m))≤D, Pm∈M. (1)

Bandwidth constraint bandwidth(P(s, m))≥B, e∈P(s, t) (2)

In the formula: P represents the path, s represents the source node, m represents the target node, D represents the maximum delay, and B represents the minimum bandwidth;

2. Expand NS2 in the following four aspects:

1) class extension: extend the mobile agent MA classifier in the address classifier;

2) Packet extension: extend the MA header in the packet header of NS2;

3) Extension of node node: In the node node structure of the original NS2, a special datagram containing mobile agent is added;

4) Agent extension: establish the mobile agent execution environment MAE, and load it on the node; the extension includes two parts: the first is to enhance the function of switcn_, so that it can judge whether the received Packet is a mobile agent; the second is to add a MA classifier that handles mobile agent packets.

The invention has the advantages of improving the success rate of the mobile agent in obtaining the shortest path and solving the problem of network balance.

Description of drawings

Figure 1 is a schematic diagram of a network model;

Fig. 2 establishes the flow chart of network model;

Figure 3 Schematic Diagram of Structural Expansion;

Figure 4 Schematic diagram of the extension of Node.

Detailed ways

1. Theoretical basis:

The present invention first proposes a network model of multicast routing with multiple QoS constraints. The purpose of QoS multicast routing is to find the optimal path in a distributed network, which requires starting from the source node, passing through all destination nodes, and satisfying all constraints. Conditions, to achieve the minimum cost or to achieve a specific service level, the QoS multicast routing problem is NP-complete.

d(u，v)：

其中d(u，v)包含了分组在节点及链路上的排队、发送和传播三部分时延，假设P(u，v)代表从节点u到节点v的一条路径，而多播树为T，那么应存在下列关系：A network can be represented as a weighted graph G=(V, E), where V represents the set of nodes, E represents the set of communication links connecting nodes, |V| and |E| represent the number of nodes and links in the network, respectively number. For the sake of generality, we only consider such a type of graph, that is, there is at most one link between a pair of nodes in this type of network, and the parameters beside the link can be used to describe the current state of the link. Let T(s, M) represent a multicast tree, where s∈V is the source node of a multicast tree, M{V-{s}} is the end node or leaf node set of the multicast tree, let R ⁺ is the set of positive real numbers. Define two functions here: c(u, v) and d(u, v), respectively represent the cost and delay on the link l, let l=(u, v)∈E, then c(u, v ):

d(u,v):

Among them, d(u, v) includes the queuing, sending and propagation delays of packets on nodes and links. Assume that P(u, v) represents a path from node u to node v, and the multicast tree is T, then the following relationship should exist:

Latency: refers to the sum of the delays of the link, which is:

$\mathrm{<span\; class="notranslate">\; delay</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Broadband: refers to the minimum bandwidth value of the link, which is:

bandwidth(P(s,m))=min{bandwidth(e), e∈P(s,t)} (1-2)

Cost: refers to the sum of the cost of the link:

$\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span>$

Cost minimum multicast routing problem with time delay and bandwidth constraints, given a network G=(V, E), multicast source node s∈V, multicast group M{V-{s}}, delay function d (u, v)∈R ⁺ , the cost function c(u, v)∈R ⁺ requires finding a multicast tree T(V _T , E _T ) starting from s and connecting all target nodes to satisfy:

Delay constraint, delay(P(s,m))≤D, Pm∈M. (1)

Bandwidth constraint bandwidth(P(s, m))≥B, e∈P(s, t) (2)

Cost Constraints In all multicast trees satisfying the conditions (1) and (2), the cost cost(T) of T(V _T , E _T ) is the smallest.

Secondly, a QoS multicast routing method (MAMRQoS) based on Ant-Like mobile agent is proposed. The main idea is to let the mobile agent move only along those links that meet the QoS requirements at each node according to the QoS service request. , search for a feasible path that satisfies the QoS request, and record the path information hop by hop at the node. In order to better describe MAMRQoS, a descriptive definition is made, including routing table, link available bandwidth table, active path and path vector.

2. Specific implementation

Implemented under NS2, Mobile Agents is a rapidly developing research topic in recent years. It has unique autonomy, intelligence and mobility, which introduces a new model for distributed computing and network control. In order to make the simulation software adapt to a variety of network environments, it should be an efficient and reliable method to develop through the expansion of mature general network simulation tools. The preferred general simulation tool is NS2 simulation software.

1. The establishment of the Mobile Agents model, as shown in Figure 1: Establish the model and realize the mobile agent. This model can simulate the behavior of the real mobile agent, such as the execution of scheduled tasks, access to the current node, and active migration.

Establish a network model: the multicast tree T(V _T , E _T ) satisfies:

Delay constraint, delay(P(s,m))≤D, Pm∈M. (1)

Bandwidth constraint bandwidth(P(s, m))≥B, e∈P(s, t) (2)

In the formula: P represents the path, s represents the source node, m represents the target node, D represents the maximum delay, and B represents the minimum bandwidth.

Specifically, the flow chart of establishing a network model is shown in Figure 2: after initialization, use the matrix formula p _t'rs to calculate all nodes that meet the two conditions of (1) and (2); adjust the parameters, and then calculate, and take all nodes that reach the minimum cost; use All nodes satisfying the two conditions of (1) and (2) with the minimum cost generate a multicast tree.

Finally, the implementation of MAMRQoS is described, and the main pseudocode of the implementation is as follows:

In QoSMRMA, it is implemented by a simple mobile agent. Each node in the network is assumed to save the required automatic regression parameters, the current address (address) and the current minimum cost (cost). When the network topology After creation, the network id is determined, and the update of the minimum cost (cost) is realized by visiting the node through the agent and adjusting the value of parameter b, so the main pseudocode of the algorithm can be expressed as:

switch (control of the main loop)

case init(tabulist, t)

tabulist={};

t = current time

L_min=0;

  home_node = current_node.address;

b = initial value;

break

case forward search(tabulist, current_node.address, s)

L_min=Min(L_min, current_node.L_min);

r = current_node.address;

p _(s) = calculate p _t,rs ;

X = (0.0, 1.0);

if(X<p _(s) ) //s is a neighbor node

Move to node s;

else p _(s) = 0

L_min＝Calculation cost (cost) value

end if

break

case adjustment parameters (b)

    b = reassignment; //reassignment>initial value

The parameter values of r _t+1 , r _t+1 and cost

break

case returns

while (if tabulist is not empty)

s=current_node;

  Move to the last node of the tabulist

r = current_node;

Upgrade P _t'rs

Where r _t+1 path vector, P _t'rs pheromone change matrix.

The above code can be expressed as a logic diagram as follows:

2. According to the functional requirements of the mobile agent operating environment, analyze the existing functions and structural characteristics of NS2 and the characteristics of the multicast routing algorithm with multiple QoS constraints. NS2 needs to be expanded in the following four aspects.

1) Expansion of structure:

NS2 connects objects and variables in C++ and Otcl languages through Tclcl. The component library of NS2 is a hierarchical structure, and the components in it are usually realized by two interrelated classes, one in C++ and the other in C++. One is in Otcl. Therefore, NS2 contains a hierarchy of C++ classes and a hierarchy of Otcl classes. The main functions of the components are realized by C++ programs, and the classes in Otcl mainly provide user-oriented configuration interfaces for C++ objects.

As shown in Figure 3: In NS2, a node Node contains an address classifier classifier_, type (multicast, unicast and mobile agent MA) classifier or a port classifier dmux_, their role is to send incoming data packets Distribute to the correct link, agent or mobile agent, extend the class address AddressClassifier, derive the class MAAddressClassifier, the following is the extension of the address classifier: class SwitchClassifier: public Classifier{public:...

SwitchClassifier(): mask_(:0), shift_(0){}...

} class_switch_classifier;

2) Extension of Packet:

Figure 2: In the network simulation of NS2, the basic unit of interaction between Packet objects consists of a series of packet headers and an optional data space. The structure of the packet header is initialized when the Simulator object is created. At the same time, the offset of each packet header relative to the starting address of the packet is also recorded. Here, the packet header of NS2 includes commom header, IP header, TCP header, RTP header, MA header and trace header, which means By default a packet header is initialized regardless of whether it is used or not. Users can define their own packet headers for new protocols, or extend existing packet headers by adding domains. The method of adding a new packet header is to first define the required domain structure in C++, then define a static class to provide a connection to Otcl, and then modify some simulated initialization codes to specify the new packet header in the packet The byte offset of .

Edit the relevant files in the NS2 source code (such as ns-packet.tcl, etc.), so that this new header structure is automatically loaded and initialized when NS2 starts, so the main part of the packet header can be expressed as:

enum packet_t{

...

PT_MA,

... };

...

class p_info{

public:

p_info(){

...

name_[PT_MA] = "ma";

... }};

3) Node extension, as shown in Figure 3:

In the node structure of the original NS2, after a datagram arrives at the node, it is first judged whether it is a unicast datagram or a multicast datagram, and accordingly the datagram is passed to the corresponding unicast module or multicast module. Since a special datagram containing a mobile agent is now added, the original node structure needs to be modified. When it is judged that a datagram of a mobile agent type is received, it should be passed to the mobile agent execution environment (MAE) of the node run in.

4) Agent extension

Build MAE and load it on the node. After the mobile agent is migrated to the node loaded with MAE, it can run on it. In NS2, the functional components that process various protocol data packets in nodes are called "Agents". Therefore, the Agent that can support the operation of mobile agents is derived from the most basic Agent as MAE.

As shown in Figure 3, the extension includes two parts: firstly, the function of switch_ is enhanced, so that it can judge whether the received Packet is a mobile agent; secondly, an MA classifier for processing mobile agent data packets is added. When switch_ determines that a Packet is a mobile agent, it will be sent to the MA classifier, and then the latter will send the mobile agent to the MAE attached to the node. After that, the MAE can call the Activate function of the mobile agent to activate the It activates. In addition, it is also possible to specify whether a node can load MAE. The mobile agent can only run on the node that has loaded MAE, but in the node that has not loaded MAE, the mobile agent will be processed as a normal datagram.

2. Implementation of MAMRQoS in NS2:

The architecture of NS2 is open, so it allows researchers to add new functions to it, realize new network topology and new traffic model. NS uses two languages, C++ and Otcl, C++ program runs faster, is a mandatory type language, requires strict data type checking, easy to implement complex data types, easy to implement accurate and complex algorithms, but modification, debug It takes longer to recompile and recompile, so C++ is adapted to the implementation of specific protocols. Otcl runs slowly, but it is very convenient for interactive modification, no compilation is required, and it is not mandatory, and it is not easy to make mistakes. It is suitable for simulation configuration. In the simulation configuration, different languages are selected according to the task, and the new code is added to the NS2 system to realize the simulation of the QoS multicast routing algorithm of the Ant-like mobile agent. Since the QoS mechanism of Ant-like mobile agent is to be realized, software with Mobile Agent mechanism is written, and then added to NS2 for simulation and simulation.

3. Simulation results:

The average performance of the algorithm is evaluated through the simulation comparison of Ant, OQMRA and MAMRQoS. In the simulation experiment, it is generated by the improved NS2 software package. The number of nodes is 50. The end-to-end delay is determined by the transmission delay and line propagation delay. Compared the average connection rate and corresponding communication overhead of the three algorithms under different network load conditions, each point on the curve is the average value of 20 simulation results, each simulation generates 100 connection requests, and the generation of requests obeys the following rules, The source and destination pairs of the call are randomly selected from the node set with a uniform probability, and the bandwidth, duration, and end-to-end maximum allowable delay of the call request are all subject to a uniform distribution. It is proved by experiments that OQMRA can not only effectively improve the transmission quality of network data packets, but also can greatly save routing time and find the optimal solution (or near-optimal solution) quickly. OQMRA can also achieve global optimization. According to the characteristics of QoS multicast routing of ant mobile agent, the method of MAMRQoS is proposed. The simulation results show that MAMRQoS solves the problem of network balance well, and provides a new thinking method for solving the multi-constraint QoS multicast routing problem.

Claims (1)

1. a method based on the QoS multicast routing of Ant-Like mobile agent, realize mobile agent under network simulator NS2, it is characterized in that: 1. Establish a network model: the multicast tree T(V _T , E _T ) satisfies: Delay constraint, delay(P(s,m))≤D, Pm∈M. (1) Bandwidth constraint bandwidth(P(s, m))≥B, e∈P(s, t) (2) In the formula: P represents the path, s represents the source node, m represents the target node, D represents the maximum delay, and B represents the minimum bandwidth; 2. Expand NS2 in the following four aspects: 1) class extension: extend the mobile agent MA classifier in the address classifier; 2) Packet extension: extend the MA header in the packet header of NS2; 3) Extension of node node: In the node node structure of the original NS2, a special datagram containing mobile agent is added; 4) Agent extension: establish mobile agent execution environment MAE, and load it on the node; the extension includes two parts: the first is to enhance the function of switch_, so that it can judge whether the received Packet is a mobile agent; secondly, add a MA classifier that handles mobile agent packets.

Images