CN102790804A - Intelligent mobile agent-based unstructured P2P network load balancing method and system - Google Patents

Abstract

The invention provides a mobile agent-based unstructured peer-to-peer network load balancing method and system, with the purpose of avoiding network node congestion and realizing load balancing. The technical solution for realizing the purpose of the present invention is: a method for load balancing of unstructured peer-to-peer networks based on mobile agents, comprising the following steps: S1, each node in the network regularly initiates a topology adjustment mechanism, connects to nodes with strong processing capabilities, Forming a central topology structure; S2, judging the congestion status of nodes by calculating the congestion level of nodes at a certain moment, and further optimizing the network topology.

Description

Method and system for unstructured peer-to-peer network load balancing based on mobile agent

technical field

The invention relates to the technical field of peer-to-peer networks, in particular to a mobile agent-based unstructured peer-to-peer network load balancing method and system.

Background technique

In recent years, peer-to-peer (Peer-to-Peer, P2P) network has gradually developed into an important network model. This is because compared with the traditional client/server (Client/Server, C/S) model, the peer-to-peer network has the advantages of good fault tolerance, strong scalability, low cost, and full use of distributed resources. Peer-to-peer networks have been widely used in file sharing, distributed computing, and streaming media services.

There are two types of peer-to-peer networks: structured and unstructured. The structured peer-to-peer network has strict restrictions on the connection of nodes, while the unstructured peer-to-peer network has a large degree of freedom in the connection of nodes. Due to the distributed characteristics of unstructured peer-to-peer network nodes, nodes only know the resources of their neighbor nodes, but do not know the distribution of resources on other nodes in the network. Therefore, a great problem in unstructured peer-to-peer networks is that resource location problem. For this reason, many search methods have been proposed to effectively locate resources on unstructured peer-to-peer networks, such as flooding, random walks, APS (Adaptive ProbabilitySearch), etc. However, due to the forwarding of query messages in the process of resource search in the above methods, some nodes in the network need to process too many messages during the search process, resulting in message congestion, which seriously affects the continuation of the search, resulting in a decline in search performance. It is particularly important to realize load balancing in a peer-to-peer network.

The so-called load balancing refers to distributing the network load to each node according to the processing capability of the node. There are two main types of existing load balancing methods. One is to study the load balancing strategy of the network by dynamically changing the network topology. The method of changing the network topology can make nodes with strong processing capabilities have more neighbors and share more burden. Large network load, make full use of its own processing capacity, to achieve the purpose of load balancing. For example, Gia adjusts the network structure according to the satisfaction level of nodes; DANTE proposed by Luis et al., nodes initiate reconnection periodically, and use the attractiveness of nodes to determine which nodes can be reconnected. In these methods, the rate at which nodes initiate queries is fixed. After the topology structure is basically stable, if the query load in the network suddenly increases, it is easy to cause congestion at the central node. The other is to study load balancing methods without changing the topology of the network. This type of method achieves load balancing by transferring redundant query messages from congested nodes to nodes with redundant processing capabilities. The most critical issue in the research is how to discover congestion. Due to the characteristics of high flexibility, high efficiency, low load, low communication delay, and high asynchrony of mobile agents, it has been gradually applied to congestion sensing methods. Shekhar et al. proposed a routing protocol for mobile agents based on congestion awareness. The mobile agent randomly selects the next access node from neighbor nodes; Li et al. proposed a peer-to-peer network load balancing method based on mobile agents. The agent finds congested nodes by visiting all nodes in the node in turn. But the existing load balancing methods based on mobile agents are carried out without changing the topology of the network.

In view of this, it is necessary to propose a new method and system for unstructured peer-to-peer network load balancing based on mobile agents.

Contents of the invention

The purpose of the present invention is to provide a mobile agent-based unstructured peer-to-peer network load balancing method and system, which avoids network node congestion and realizes load balancing.

A kind of unstructured peer-to-peer network load balancing method based on mobile agent of the present invention, described method comprises the following steps:

S1. Each node in the network regularly initiates a topology adjustment mechanism to connect to nodes with strong processing capabilities to form a central topology structure;

S2. Judging the congestion state of the node by calculating the congestion level of the node at a certain moment, and further optimizing the network topology.

As a further improvement of the present invention, the step S1 is specifically:

S11. The node P _i sends a collection message to collect a set of candidate nodes participating in the topology adjustment of the node;

S12. Forward the node collection message to a randomly selected neighbor node, and add the node to the set S _i ; at the same time, the TTL of the survival time is reduced by 1; if the TTL of the survival time is not equal to 0, repeat step S12, and continue to collect nodes;

S13. Return the candidate node set Si to the message originating node P _i ;

S14, select the node P _min(Ni) with the weakest processing capability and a degree of connectivity greater than 1 from the current neighbor node set of node P _i ;

S15. Take out the node with the strongest processing capability from the candidate node set S _i , record it as P _max(Si) , and delete it from S _i , if the processing capability of node P _max(Si) is not greater than P _{min(Ni )} , execute step S17; if the processing capacity of the node P _max(Si) is greater than P _min(Ni) , send a connection request to the node P _max(Si) , if the node P _max(Si) rejects the connection request, repeat the steps S15, continue to select the next reconnection node, if the node P _max(Si) accepts the reconnection request, then execute step S16;

S16. The node P _i disconnects from the selected neighbor node P _min(Ni) and establishes a connection with the node P _max(Si) . If the number of reconnected nodes reaches the preset upper limit n of the number of reconnected nodes, step S17 is performed; if not, step S14 is performed to select the node that will participate in the reconnection next time;

S17. This topology adjustment ends.

As a further improvement of the present invention, the step S2 is specifically:

并利用

的值来判断节点的拥塞状态，包括对每个节点设定过载、正常负载和过载三种负载状态，并设定负载状态阈值，将计算出的

值与设定的负载状态阈值相比较从而确定节点的当前负载状态；若节点P_j负载状态是过载，则执行步骤S22；否则执行步骤S23，进行拓扑的优化，选择下一个访问节点；S21. The mobile agent collects information and calculates the congestion level of node P _j at the current moment t

and use

value to judge the congestion state of the node, including setting three load states of overload, normal load and overload for each node, and setting the load state threshold, the calculated

Value is compared with the load state threshold value of setting so as to determine the current load state of node; If the load state of node _Pj is overload, then execute step S22; Otherwise execute step S23, carry out the optimization of topology, select next visit node;

S22. The node P _j randomly selects some nodes to disconnect, and derives a child mobile agent; the child mobile agent takes its generation node as the parent node, and searches the network in a breadth-first traversal manner. For nodes with excess processing capacity, each time a node arrives, the load status of the node is judged first. If the node is lightly loaded, a node to be reconnected is randomly selected to connect to it, and then the next node is searched until all nodes to be reconnected After the reconnection is completed, the sub-mobile agent actively exits the network;

S23. The mobile agent selects the next visited node according to the attraction table of the neighbor node P _o of the node P _j to the mobile agent and the list of visited nodes, and migrates to this node; the visited node list It is used to record the nodes visited by the agent; the neighbor node attractiveness table is a table for any node in the network to record the attractiveness of its neighbor nodes to the mobile agent.

的计算公式为：As a further improvement of the present invention, in the step S21, the congestion level of the node P _j at the current moment t

The calculation formula is:

${\mathrm{<span\; class="notranslate">\; CL</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

表示节点P_j处理一条查询消息所需的时间；如果结点正在处理消息那么此时到来的消息都被放在该节点的消息缓存队列中，

表示时刻t节点P_j的缓存队列中待处理的消息数。in,

Indicates the time required for node P _j to process a query message; if the node is processing the message, the incoming messages at this time are placed in the message cache queue of the node,

Indicates the number of messages to be processed in the cache queue of node P _j at time t.

As a further improvement of the present invention, the step S21 further includes:

In step S210, the node P _j decides whether to accept connection requests from other nodes according to its own load status: if the node P _j is currently overloaded, the node P _j does not accept other node connection requests; if the node P _j is currently under normal load, and At this time, if the last state of node P _j is overloaded, then node P _j will not accept connection requests from other nodes; if the previous state of node P _j is not overloaded, and node P _j is currently in a normal state, then node P _j will accept other nodes connection request; if the current state of node P _j is light load, then node P _j accepts connection requests from other nodes.

的计算公式为：As a further improvement of the present invention, in the step S21, the congestion level of the node P _j at the current moment t

The calculation formula is:

${\mathrm{<span\; class="notranslate">\; CL</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

表示时刻t节点P_j的缓存队列中待处理的消息数。in, Indicates the time required for node P _j to process a query message; if the node is processing the message, the incoming messages at this time are placed in the message cache queue of the node,

Indicates the number of messages to be processed in the cache queue of node P _j at time t.

节点P_j的状态是过载的，记变量flag＝1，若节点P_j是正常负载的，记变量flag＝0。若

节点P_j的状态是轻载的，记变量flag＝-1。As a further improvement of the present invention, in the step S21 "the calculated The value is compared with the set load state threshold to determine the current load state of the node" specifically: the two set thresholds U _thred and B _thred , if

The state of node P _j is overloaded, record the variable flag=1, if The node P _j is under normal load, and the variable flag=0. like

The state of node P _j is light load, and the variable flag=-1.

As a further improvement of the present invention, the step S210 specifically includes: the current variable state of the node P _j is flag=1, and the node P _j does not accept other node connection requests; if the current state of the node P _j flag=0, and at this time if If the last variable status flag of node P _j is = 1, then node P _j will not accept other node connection requests; if the previous variable status flag of node P _j is = 0 or -1, and the current variable status of node P _j is flag = 0, then the node P _j can accept connection requests from other nodes; if the current variable status of node P _j is flag=-1, then node P _j can accept connection requests from other nodes.

及连通度χ(P_o,k);邻居节点P_o对移动智能体的吸引力计算如下：As a further improvement of the present invention, in the step S23, the node P _j is provided with a neighbor node P _o attraction table of the node, which is used to record the processing capabilities of all neighbor nodes on the node P _j

and connectivity χ(P _o ,k); the attractiveness of the neighbor node P _o to the mobile agent is calculated as follows:

${\mathrm{<span\; class="notranslate">\; A</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

越大，则移动智能体从P_j迁移到节点P_o的可能性越高； is the processing capability of the neighbor node P _o , which means the time required for the neighbor node P _o to process a query message; χ(P _o , k) is the connectivity of the node P _o ; the attractiveness of the node P _o

The larger is, the higher the possibility of the mobile agent migrating from P _j to node P _o ;

The connectivity calculation formula of the neighbor node P _o is as follows:

$\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Among them, N(P _o , h) represents the number of nodes that are h hops away from the neighbor node P _o , k represents the radius for calculating connectivity, δ is the control coefficient, and 1/h ^δ is used to control the connectivity of node pairs at different distances Influenced weight, the larger χ(P _o ,k) is, the more query messages the node P _o may receive.

Correspondingly, a kind of unstructured peer-to-peer network load balancing system based on mobile agent, said system comprises:

The topology adjustment unit is used for each node in the network to periodically initiate a topology adjustment mechanism, connect to nodes with strong processing capabilities, and form a central topology unit;

The topology optimization unit is used to judge the congestion state of the node by calculating the congestion level of the node at a certain moment, and further perform network topology optimization;

Wherein, the topology adjustment unit includes:

The node P _i sends a collection message to collect the units of the candidate node set participating in the topology adjustment of the node;

Forward the node collection message to a randomly selected neighbor node, and add this node to the set S _i ;

Return the candidate node set S _i to the unit of the message originating node P _i ;

Select the unit of the node P _min(Ni) with the weakest processing capability and a connection degree greater than 1 from the current neighbor node set of node P _i ;

Take out the node P _max(Si) with the strongest processing capability from the candidate node set S _i , compare the processing capabilities of the node P _max(Si) and P _min(Ni) , and send a connection request to the node P _max(Si) ;

The node P _i disconnects the connection with the selected neighbor node P _min(Ni) and establishes a connection with the node P _max(Si) ;

The topology optimization unit includes:

对每个节点设定过载、正常负载和过载三种负载状态，并设定负载状态阈值，将计算出的

值与设定的负载状态阈值相比较从而确定节点的当前负载状态的单元；The mobile agent collects information and calculates the congestion level of node P _j at the current moment t

Set three load states of overload, normal load and overload for each node, and set the load state threshold, the calculated

A unit that compares the value with a set load state threshold to determine the current load state of the node;

The node P _j randomly selects a part of the nodes to disconnect, and derives a sub-mobile agent unit;

The child mobile agent takes its generation node as the parent node, and searches for nodes with redundant processing capacity in the network in the way of breadth-first traversal, and judges the load status of the node every time it arrives at a node, and randomly selects the node if it is lightly loaded A node to be reconnected is connected to it, and then continues to search for the next node until all nodes to be reconnected are reconnected, and the sub-mobile agent actively exits the unit of the network;

The mobile agent selects the next visited node according to the attractiveness of the neighbor node P _o of the node P _j to the mobile agent and the list of visited nodes, and migrates to the unit of this node.

As a further improvement of the system of the present invention, the congestion level of the node P _j at the current moment t The calculation formula is:

${\mathrm{<span\; class="notranslate">\; CL</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

表示时刻t节点P_j的缓存队列中待处理的消息数；in, Indicates the time required for node P _j to process a query message; if the node is processing the message, the incoming messages at this time are placed in the message cache queue of the node,

Indicates the number of messages to be processed in the cache queue of node P _j at time t;

The calculation of the attractiveness of the neighbor node P _o to the mobile agent is as follows:

${\mathrm{<span\; class="notranslate">\; A</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

越大，则移动智能体从P_j迁移到节点P_o的可能性越高； is the processing capability of the neighbor node P _o , which means the time required for the neighbor node P _o to process a query message; χ(P _o , k) is the connectivity of the node P _o ; the attractiveness of the node P _o

The larger is, the higher the possibility of the mobile agent migrating from P _j to node P _o ;

The connectivity calculation formula of the neighbor node P _o is as follows:

$\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Among them, N(P _o , h) represents the number of nodes that are h hops away from the neighbor node P _o , k represents the radius for calculating connectivity, δ is the control coefficient, and 1/h ^δ is used to control the connectivity of node pairs at different distances Influenced weight, the larger χ(P _o ,k) is, the more query messages the node P _o may receive.

The beneficial effects of the present invention are: each node in the network of the present invention periodically starts a topology reconnection mechanism to form a centralized topology structure. At the same time, nodes continuously collect information such as processing capabilities and connectivity of neighboring nodes to guide mobile agents to migrate purposefully in the network and constantly monitor the load of important nodes in the network, optimize and adjust the topology of the network to avoid congestion. , so as to realize the load balance on the unstructured peer-to-peer network, and adjust the network topology in time according to the load status of the network, which can effectively avoid network node congestion and improve resource search efficiency.

Description of drawings

FIG. 1 is a flow chart of a mobile agent-based unstructured peer-to-peer network load balancing method in the present invention.

FIG. 2 is a specific flow chart of step S1 in FIG. 1 for periodic network node topology adjustment.

FIG. 3 is a specific flowchart of the topology optimization of the mobile agent in step S2 in FIG. 1 .

Fig. 4 is a network topology simulation diagram in an embodiment of the present invention.

FIG. 5 is a diagram of the degree of aggregation of network nodes after topology reconnection in an embodiment of the present invention.

FIG. 6 is a graph showing changes in the average connectivity of nodes with different capabilities in the network after topology reconnection in an embodiment of the present invention.

Fig. 7 is a change diagram of the average number of nodes passed by the query after topology reconnection in an embodiment of the present invention.

FIG. 8 is a schematic structural diagram of a mobile agent-based unstructured peer-to-peer network load balancing system according to the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with various embodiments shown in the drawings. However, these embodiments do not limit the present invention, and any structural, method, or functional changes made by those skilled in the art according to these embodiments are included in the protection scope of the present invention.

Referring to Fig. 1 shown is a kind of flow chart of the unstructured peer-to-peer network load balancing method based on mobile agent of the present invention, comprises the following steps:

S1. Each node in the network regularly initiates a topology adjustment mechanism to connect to nodes with strong processing capabilities to form a central topology structure;

S2. Judging the congestion state of the node by calculating the congestion level of the node at a certain moment, and further optimizing the network topology.

Referring to Fig. 2, it is a specific flowchart of step S1 network node topology regular adjustment in this embodiment, including the following steps:

S11. The node P _i sends a collection message to collect a set of candidate nodes participating in the topology adjustment of the node. Assuming that the set of candidate nodes is S _i , and the initial S _i is empty, the TTL (Time to Live, TTL) of the message is K;

S12. Forward the node collection message to a randomly selected neighbor node, and add this node to the set S _i . The time-to-live TTL of the collected messages is reduced by 1, and it is judged whether the time-to-live TTL is equal to 0. If so, the node collection process ends, and step S13 is executed. If not, step S12 is repeated to continue collecting the node;

S13. Return the candidate node set S _i to the message originating node P _i ;

S14. Select the node P _min(Ni) with the weakest processing capability and a connection degree greater than 1 from the current neighbor node set of node P _i , where N _i is the set of neighbor nodes of node P _i ;

S15. Take out the node P _max(Si) with the strongest processing capability from the candidate node set S _i , and delete the node P _max(Si) from S _i . Compare the processing capabilities of nodes P _max(Si) and P _min(Ni) , if the processing capability of node P _max(Si) is not greater than P _min(Ni) , execute step S17; if the processing capability of node P _max(Si) If it is greater than P _min(Ni) , a connection request is sent to node P _max(Si) . If the node P _{max (Si)} rejects the connection request, then repeat step S15, continue to select the next reconnection node, if the node P _{max (Si)} accepts the connection request, then perform step S16;

S16. The node P _i disconnects from the selected neighbor node P _min(Ni) and establishes a connection with the node P _max(Si) . If the number of reconnected nodes reaches the preset upper limit n of the number of reconnected nodes, step S17 is performed; if not, step S14 is performed to select the node that will participate in the reconnection next time;

S17. This topology adjustment ends.

Simply relying on the regular topological reconnection of the nodes themselves will eventually form a centralized network centered on the nodes with the strongest processing capabilities. In the absence of congestion, the network resource search efficiency of this structure is the highest, but almost The resource search requests of all nodes will be sent to the central node, and the central node may easily be congested, which will seriously affect the efficiency of resource search. In order to avoid the occurrence of congestion, the mobile agent-based topology optimization method.

Referring to Fig. 3, it is a specific flowchart of the topology optimization of the step S2 mobile agent in the present embodiment, including the following steps:

并利用

的值来判断节点的拥塞状态，确定节点的当前负载状态，若节点P_j负载状态是过载的，则执行步骤S22；否则执行步骤S23，进行拓扑的优化，选择下一个访问节点；S21. The mobile agent collects information and calculates the congestion level of node P _j at the current moment t

and use

value to judge the congestion state of the node, determine the current load state of the node, if the load state of the node P _j is overloaded, then perform step S22; otherwise perform step S23, perform topology optimization, and select the next access node;

S22. Node P _j randomly selects some nodes to disconnect, and derives a child mobile agent; the child mobile agent takes its generation node as the parent node, and searches for redundant processing capacity in the network in the way of breadth-first traversal. Nodes, each time a node is reached, first judge the load status of the node. If the node is lightly loaded, randomly select a node to be reconnected to connect to it, and then continue to search for the next node until all the nodes to be reconnected are reconnected. , the child mobile agent actively exits the network;

S23. The mobile agent selects the next visited node according to the attractiveness list of the neighbors of the node P _j and the list of nodes visited by the mobile agent, and migrates to this node.

的计算公式为：In step S21, the congestion level of node P _j at the current moment t

The calculation formula is:

${\mathrm{<span\; class="notranslate">\; CL</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

表示节点P_j处理一条查询消息所需的时间；如果结点正在处理消息那么此时到来的消息都被放在该节点的消息缓存队列中，

表示时刻t节点P_j的缓存队列中待处理的消息数。该公式表示在时刻t一个查询消息被转发到节点P_j到处理完毕需要花费的时间。

Indicates the time required for node P _j to process a query message; if the node is processing the message, the incoming messages at this time are placed in the message cache queue of the node,

Indicates the number of messages to be processed in the cache queue of node P _j at time t. This formula expresses the time it takes for a query message to be forwarded to node P _j to be processed at time t.

"Determine the current load state of the node" in step S21 is specifically:

的值与设定的两个阈值U_thred、B_thred相比较从而确定节点的当前负载状态：Each node is set to have three load states, according to

The value of is compared with the set two thresholds U _thred and B _thred to determine the current load status of the node:

节点P_j的状态是过载的，记变量flag＝1，此时认为该节点拥塞；like

The state of the node P _j is overloaded, and the variable flag=1 is considered to be congested at this time;

节点P_j是正常负载的，记变量flag＝0；like

Node P _j is under normal load, record variable flag=0;

节点P_j的状态是轻载的，记变量flag＝-1。like

The state of node P _j is light load, and the variable flag=-1.

Further, node _Pj decides whether to accept connection requests from other nodes according to its own load status. That is _, the current variable state _of node P _j is flag=1, and node P _j does not accept connection requests from other nodes; _j does not accept connection requests from other nodes; if a variable status flag on node P _j =0 or -1, and node P _j ’s current variable status flag=0, then node P _j can accept connection requests from other nodes; if node P _j is currently If the variable status flag=-1, the node P _j can accept connection requests from other nodes.

In step S23, in order to determine the attractiveness of all neighbor nodes on node P _j , node P _j maintains a neighbor node attractiveness table of the node, which is used to record the processing capacity and connectivity of all neighbor nodes on node P _j Spend. Assuming that a neighbor node of node P _j is P _o , the calculation formula of the connectivity of P _o is as follows:

$\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Among them, N(P _o , h) represents the number of nodes that are h hops away from node P _o , k represents the radius for calculating connectivity, and δ is the control coefficient. 1/h ^δ is used to control the weight of the impact of nodes with different distances on connectivity. The larger χ(P _o ,k) is, the more query messages the node P _o may receive.

The attractiveness of node P _o to the mobile agent is calculated as follows:

${\mathrm{<span\; class="notranslate">\; A</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

是节点P_o的处理能力，χ(P_o,k)是节点P_o的连通度。节点P_o吸引力

越大，则移动智能体从P_j迁移到节点P_o的可能性越高。

is the processing capability of node P _o , χ(P _o ,k) is the connectivity of node P _o . Node P _o attraction

The larger , the higher the possibility of the mobile agent migrating from P _j to node P _o .

An example of a specific implementation process of the present invention is given below. The example includes 10,000 network nodes, each node has 50 different data, and each data has 100 copies in the network. The mobile agent is generated by the first node joining the network. The processing power of the node is divided into five levels: 0.1, 1, 10, 100, 1000, and the percentage of network nodes is corresponding to: 20, 45, 30, 4.9 , 0.1, use random walks to search for resources, and use the following two methods to reconnect the topology to achieve load balancing. One is the periodic topology reconnection of nodes themselves. The purpose is to enable nodes with strong processing capabilities to have more connections, so that they can process more query messages and make full use of their processing capabilities; the other is based on mobile agents. The purpose of topology optimization is to prevent nodes with strong processing capabilities from having too many neighbors to cause congestion, which will affect the performance of the network instead.

Each node initiates a topology reconnection every 30 seconds to connect itself to a node with strong processing capabilities. Figure 4 is a simulation diagram of the network topology. The following is a topology reconnection process of a node P ₁ in the simulation diagram:

S11. The node P ₁ sends a collection message to collect a set of candidate nodes participating in the topology adjustment of the node. Assume that the set of candidate nodes is S ₁ , and the initial S ₁ is empty, and the TTL is 5.

S12. Forward the node collection message to a randomly selected neighbor node, and add the node to the set S ₁ , and decrease the TTL by 1 at the same time. If the survival time TTL is not equal to 0, then repeat step S12 to continue collecting nodes; otherwise, the node collecting process ends and step S13 is executed.

S13. Return the candidate node set S ₁ to the message originating node P ₁ , assuming that S ₁ is {P, P ₄ , P ₉ , P ₁₂ , P ₁₃ }, and then execute step S14.

S14, from the current neighbor node set of node _P1 , according to Figure 4, the current neighbor node set of _P1 node is {P, _P3 }, select the node with the weakest processing ability and connection degree greater than 1, and record it as _{Pmin(N1 )} , N ₁ is the neighbor node set of node P ₁ ; then execute step S15. For example, among the neighbor nodes of P ₁ , the node with the weakest processing capability is P ₃ , and P _min(N1) is the node P ₃ , and then step S15 is executed.

S15. Take out the node with the strongest processing capability from the candidate node set S ₁ , record it as P _max(S1) , delete P _max(S1) from S ₁ , if the processing capability of node P _max(S1) is not greater than P _min(N1) , it means that the processing capabilities of all the remaining candidate nodes are weaker than the existing neighbor nodes of node P ₁ , go to step S17; if the processing capability of node P _max(S1) is greater than P _min(N1) , send P _max(S1) sends a connection request, if the node P _max(S1) rejects the connection request, then repeat step S15, continue to select the next reconnection node, if the node P _max(S1) accepts the reconnection request, then execute step S16 . For example, the node with the strongest processing capability selected from the candidate node set _S1 is _P4 , and the processing capability of _P4 is greater than the node _P3 selected in step S14, then node _P1 sends a connection request to _P4 , and _P4 accepts If the connection request is received, step S16 is executed.

S16. Node P ₁ disconnects the connection with the selected neighbor node P _{min (N1)} and establishes a connection with node P _{max (S1)} . If there are already n candidate nodes participating in the reconnection, in this example n selects 3, then execute Step S17; otherwise, execute step S14, and start to select a node to participate in the reconnection next time. For example, node _P1 disconnects from node _P3 selected by step S14 and establishes a connection with node _P4 . Since there are no three candidate nodes participating in reconnection, step S14 is executed to start selecting the node to participate in reconnection next time.

S17. This topology adjustment ends.

As shown in Figure 4, assuming that the node where the mobile agent is currently located is P, the topology optimization process of the mobile agent at node P is as follows:

S21. The mobile agent collects information to calculate the congestion level CL _P (T) of node P at the current moment T, and uses the value of CL _P (T) to judge the congestion state of the node. At the same time, each node has three load states. , we compare the calculated _CLP (T) value with the two set thresholds U _thred and B _thred to determine the current load status of the node. The two thresholds U _thred and B _thred are set to be 0.1 and 0.05 respectively. If CL _P (T)>U _thred , the state of the node is overloaded, record the current variable state flag=1 of the node P at time T, and at this time, the node is considered to be congested. If B _thred <=CL _P (T)<=U _thred , node P is under normal load, and flag=0. If CL _P (T)<B _thred , the state of node P is lightly loaded, record flag=-1. The node P decides the next execution step according to its load state, that is, if the state of the node P is overloaded, execute step S22 to optimize the topology; otherwise execute step S23 to select the next access node.

In addition, node P decides whether to accept connection requests from other nodes according to its own load status. That is, the current variable state flag of node P=1, node P does not accept other node connection requests; if the current state flag of node P=0, and if the previous variable state flag of node P=1 at this time, node P does not accept other nodes Connection request; if a variable state flag on node P=0 or -1, and node P’s current variable state flag=0, then node P can accept connection requests from other nodes; if node P’s current variable state flag=-1, then node P P can accept connection requests from other nodes;

S22. The node P randomly selects some nodes to disconnect from it, and in this embodiment, disconnects from two nodes at a time. Assume node P disconnects from two nodes and spawns a child mobile agent.

The child mobile agent is mainly responsible for reconnecting the disconnected connection. The child mobile agent takes its generation node as the parent node, and searches for nodes with redundant processing capacity in the network in the way of breadth-first traversal, and judges each time it reaches a node. The load status of the node. If the node is lightly loaded, randomly select a node to be reconnected to connect with it, and then continue to search for the next node until all the nodes to be reconnected are reconnected. At this time, the sub-mobile agent actively log out of the network;

S23. The mobile agent selects the next visited node according to the attractiveness of the neighbors of node P to it and the list of visited nodes, and migrates to this node, where k and δ are 2 and 1 respectively. Assume that the neighbor nodes of node P in Figure 4 are sorted from large to small according to their attractiveness to the mobile agent: P ₄ , P ₅ , P ₃ , P ₁ , P ₂ , and the list of nodes recently visited by the mobile agent is : P ₁₂ , P ₉ , P ₄ , so the next visiting node of the mobile agent is P ₅ .

Figure 5 is a change diagram of the clustering coefficient (CC) of network nodes during the topology optimization process. The calculation formula of CC is as follows:

$\mathrm{<span\; class="notranslate">\; CC</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; |</span>}\underset{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; V</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; CC</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Where V is the set of all nodes in the network, CC _p is the aggregation coefficient of the neighbor nodes of node p, and the calculation formula is as follows:

${\mathrm{<span\; class="notranslate">\; CC</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&cap;</span>\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; |</span>}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Where N _p is the set of neighbor nodes of node p, k _p = |N _p | is the number of neighbor nodes of node p, and E is the set of connections between nodes in the network. 0<=CC<=1, the smaller the CC value, the more random distribution of network nodes, and the larger the CC value, the more concentrated the nodes.

It can be seen from Figure 5 that as the topology changes, the degree of node aggregation becomes higher and higher, and reaches a basically stable state. It can be seen that the topology of the network becomes more and more concentrated from completely random.

Figure 6 shows the changes in the average node degree (the number of neighbor nodes) of nodes with the same processing capacity during the topology reconnection process. In the experiment, the change of node degree was observed in units of 100 seconds. It can be seen that a large number of nodes began to reconnect to the node with the strongest processing capacity, but it quickly reaches a stable state, and the higher the connection degree of the node with the stronger processing capacity, it means that the node with the stronger processing capacity can share more query load and can make full use of the network nodes processing capacity.

Figure 7 is a change diagram of the average number of hops required for a successful query. In the experiment, the average value of the hops required for every 1000 queries was obtained as shown in the figure. With the optimization of the network topology, the average number of hops required for the query gradually decreased. It can be seen from the figure that the number of hops required after topology optimization is much lower than that at the beginning, and the fewer hops required for query, the higher the efficiency of the search algorithm. It can be seen that the topology optimization method based on mobile agents is greatly improved. network search performance.

Correspondingly, as shown in FIG. 8, a mobile agent-based unstructured peer-to-peer network load balancing system 100 of the present invention includes:

The topology adjustment unit 10 is used for each node in the network to periodically initiate a topology adjustment mechanism to connect to nodes with strong processing capabilities to form a central topology unit;

The topology optimization unit 20 is used to judge the congestion state of the node by calculating the congestion level of the node at a certain moment, and further perform network topology optimization;

The topology adjustment unit 10 includes:

The node P _i sends a collection message to collect the units of the candidate node set participating in the topology adjustment of the node;

Forward the node collection message to a randomly selected neighbor node, and add this node to the set S _i ;

Return the candidate node set S _i to the unit of the message originating node P _i ;

Select the unit of the node P _min(Ni) with the weakest processing capability and a connection degree greater than 1 from the current neighbor node set of node P _i ;

Take out the node P _max(Si) with the strongest processing capability from the candidate node set S _i , compare the processing capabilities of the node P _max(Si) and P _min(Ni) , and send a connection request to the node P _max(Si) ;

The node P _i disconnects the connection with the selected neighbor node P _min(Ni) and establishes a connection with the node P _max(Si) ;

The topology optimization unit 20 includes:

值与设定的负载状态阈值相比较从而确定节点的当前负载状态的单元；其中，节点P_j在当前时刻t的拥塞水平

的计算公式为：The mobile agent collects information and calculates the congestion level of node P _j at the current moment t Set three load states of overload, normal load and overload for each node, and set the load state threshold, the calculated

The value is compared with the set load state threshold to determine the unit of the current load state of the node; where, the congestion level of node P _j at the current time t

The calculation formula is:

${\mathrm{<span\; class="notranslate">\; CL</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}}$

表示节点P_j处理一条查询消息所需的时间；如果结点正在处理消息那么此时到来的消息都被放在该节点的消息缓存队列中，表示时刻t节点P_j的缓存队列中待处理的消息数；in,

Indicates the time required for node P _j to process a query message; if the node is processing the message, the incoming messages at this time are placed in the message cache queue of the node, Indicates the number of messages to be processed in the cache queue of node P _j at time t;

The node P _j randomly selects a part of the nodes to disconnect, and derives a sub-mobile agent unit;

The child mobile agent takes its generation node as the parent node, and searches for nodes with redundant processing capacity in the network in the way of breadth-first traversal, and judges the load status of the node every time it arrives at a node, and randomly selects the node if it is light-loaded A node to be reconnected is connected to it, and then continues to search for the next node until all nodes to be reconnected are reconnected, and the sub-mobile agent actively exits the network unit;

The mobile agent selects the next visited node according to the attractiveness of the neighbor node P _o of the node P _j to the mobile agent and the list of visited nodes, and migrates to the unit of the node; among them, the neighbor node P _{o is} to the mobile agent The attraction for is calculated as follows:

${\mathrm{<span\; class="notranslate">\; A</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

是邻居节点P_o的处理能力，表示该邻居节点P_o处理一条查询消息所需的时间；x(P_o,k)是节点P_o的连通度；节点P_o吸引力

越大，则移动智能体从P_j迁移到节点P_o的可能性越高；

is the processing capability of the neighbor node P _o , which means the time required for the neighbor node P _o to process a query message; x(P _o ,k) is the connectivity of the node P _o ; the attractiveness of the node P _o

The larger is, the higher the possibility of the mobile agent migrating from P _j to node P _o ;

The connectivity calculation formula of the neighbor node P _o is as follows:

$\mathrm{<span\; class="notranslate">\; \&chi;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\frac{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; o</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; h</span>}}^{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}}$

Among them, N(P _o , h) represents the number of nodes that are h hops away from the neighbor node P _o , k represents the radius for calculating connectivity, δ is the control coefficient, and 1/h ^δ is used to control the connectivity of node pairs at different distances Influenced weight, the larger χ(P _o ,k) is, the more query messages the node P _o may receive.

Compared with the prior art, each node in the network of the present invention periodically starts a topology reconnection mechanism to form a centralized topology structure. At the same time, nodes continuously collect information such as processing capabilities and connectivity of neighboring nodes to guide mobile agents to migrate purposefully in the network and constantly monitor the load of important nodes in the network, optimize and adjust the topology of the network to avoid congestion. , so as to realize the load balance on the unstructured peer-to-peer network, and adjust the network topology in time according to the load status of the network, which can effectively avoid network node congestion and improve resource search efficiency.

For the convenience of description, when describing the above devices, functions are divided into various units and described separately. Of course, when implementing the present application, the functions of each unit can be implemented in one or more pieces of software and/or hardware.

It can be known from the above description of the implementation manners that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general-purpose hardware platform. Based on this understanding, the essence of the technical solution of this application or the part that contributes to the prior art can be embodied in the form of software products, and the computer software products can be stored in storage media, such as ROM/RAM, disk , optical disc, etc., including several instructions to enable a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

The device implementation described above is only illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network elements. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. It can be understood and implemented by those skilled in the art without creative effort.

The application can be used in numerous general purpose or special purpose computing system environments or configurations. Examples: personal computers, server computers, handheld or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, etc.

This application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

It should be understood that although this description is described according to implementation modes, not each implementation mode only contains an independent technical solution, and this description in the description is only for clarity, and those skilled in the art should take the description as a whole, and each The technical solutions in the embodiments can also be properly combined to form other embodiments that can be understood by those skilled in the art.

The series of detailed descriptions listed above are only specific descriptions for feasible implementations of the present invention, and they are not intended to limit the protection scope of the present invention. Any equivalent implementation or implementation that does not depart from the technical spirit of the present invention All changes should be included within the protection scope of the present invention.

Claims (10)

1. A method for load balancing of a structureless peer-to-peer network based on a mobile agent is characterized by comprising the following steps:

s1, each node in the network periodically initiates a topology adjustment mechanism, and the topology adjustment mechanism is connected to the nodes with strong processing capacity to form a central topology structure;

and S2, judging the congestion state of the node by calculating the congestion level of the node at a certain moment, and further performing network topology optimization.

2. The method according to claim 1, wherein the step S1 specifically includes:

s11 and node P_iSending a collecting message to collect a candidate node set participating in the node topology adjustment;

s12, forwarding the node collection message to a randomly selected neighbor node, and adding the node to the set S_i(ii) a Simultaneously, the survival time TTL is reduced by 1; if the time to live TTL is not equal to 0, repeating step S12, and continuing to collect nodes;

s13, collecting candidate nodes S_iReturned to the message originating node P_i；

S14, slave node P_iSelecting the node P with the weakest processing capability and the connectivity degree larger than 1 from the current neighbor node set_min(Ni)；

S15, slave candidate node set S_iThe node with the strongest processing capability is taken out and recorded asAnd slave it to S_iIn deletion, if the node

Has a processing capacity of not more than

Step S17 is executed; if node

Has a processing capacity greater thanThen to the node

Sending out a connection request if the node

Rejecting a connection requestThen step S15 is repeated to continue selecting the next reconnected node, if the node is not the next reconnected node

If the reconnect request is accepted, go to step S16;

s16 and node P_iNeighboring node P of disconnection and selection_min(Ni)And is connected with node P_max(Si)Establishing connection; if the number of the reconnected nodes reaches the upper limit number n of the preset number of the reconnected nodes, executing step S17; if not, executing step S14 to select the node to join the reconnection next time;

and S17, finishing the topology adjustment.

3. The method for load balancing of a structureless peer-to-peer network based on mobile agents according to claim 1, wherein the step S2 specifically comprises:

s21 calculation node P for collecting information by mobile agent_jCongestion level at the current time t

And use

The congestion state of the node is judged by the value of (A), the congestion state of the node comprises setting three load states of overload, normal load and overload for each node, setting a load state threshold value, and calculating

Comparing the value with a set load state threshold value to determine the current load state of the node; if node P_jIf the load status is overload, go to step S22; otherwise, executing step S23, optimizing the topology, and selecting the next access node;

s22, the node P_jRandomly selecting a part of nodes to disconnect, and deriving a sub-mobile agent; the child mobile agent takes its generation node as a parent nodeSearching nodes with redundant processing capacity in the network in a breadth-first traversal mode, judging the load state of the nodes when reaching one node, randomly selecting one node to be reconnected to connect with the node if the node is lightly loaded, then continuously searching the next node until all the nodes to be reconnected are reconnected, and actively quitting the network by the sub-mobile intelligent agent;

s23, the mobile agent according to the node P_jOf neighbor node P_oSelecting a next access node according to the attraction table of the mobile agent and the accessed node list, and migrating the next access node to the node; the accessed node list is used for recording the nodes accessed by the agent; the neighbor node attraction table is a table of the attractions of the neighbor nodes to the mobile agents, which is used by any node in the network for recording.

4. The method for load balancing of unstructured peer-to-peer networks based on mobile agents according to claim 3, wherein the step S21 further comprises:

step S210 said node P_jAnd determining whether to accept the connection request of other nodes according to the load state of the node: if node P_jCurrently in an overload state, node P_jNot accepting other node connection requests; if node P_jIs currently under normal load, and at this time if node P_jIs overloaded, node P_jNot accepting other node connection requests; if node P_jThe last state is not overloaded, but node P_jWhen the current state is normal, the node P_jReceiving connection requests of other nodes; if node P_jNode P when the current state is light load_jAnd accepting connection requests of other nodes.

5. The method according to claim 3, characterized in that in step S21, the node P_jCongestion level at the current time t

The calculation formula of (2) is as follows:

wherein,

representing a node P_jThe time required to process a query message; if a node is processing a message then all incoming messages at that time are placed in the node's message buffer queue,

indicating the time t node P_jThe number of messages to be processed in the buffer queue.

6. The method for load balancing unstructured peer-to-peer networks based on mobile agents according to claim 3, characterized in that in step S23, the node P_jNeighbor node P provided with one such node_oAn attraction meter for recording the node P_jProcessing capability of all neighbor nodesAnd connectivity χ (P)_o,k)；

Neighbor node P_oThe attraction to a mobile agent is calculated as follows:

<math> <mrow> <msub> <mi>A</mi> <msub> <mi>P</mi> <mi>o</mi> </msub> </msub> <mo>=</mo> <msub> <mi>C</mi> <msub> <mi>P</mi> <mi>o</mi> </msub> </msub> <mo>&times;</mo> <mi>&chi;</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>)</mo> </mrow> </mrow> </math>

is a neighbor node P_oRepresents the neighboring node P_oThe time required to process a query message;

χ(P_ok) is a node P_oThe connectivity of (c); node P_oAttractive force

The larger the mobile agent slave P_jMigration to node P_oThe higher the probability of (c);

the neighbor node P_oThe connectivity calculation formula is as follows:

<math> <mrow> <mi>&chi;</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>h</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>h</mi> <mo>=</mo> <mi>k</mi> </mrow> </munderover> <mfrac> <mrow> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>h</mi> <mo>)</mo> </mrow> </mrow> <msup> <mi>h</mi> <mi>&delta;</mi> </msup> </mfrac> <mo>,</mo> </mrow> </math>

wherein, N (P)_oH) represents a neighbor node P_oThe number of nodes which are h hops away, k represents the radius of the calculated connectivity, delta is a control coefficient, 1/h^δWeight, χ (P), for controlling influence of nodes of different distances on connectivity_oK) the larger the node P_oThe more query messages that may be received.

7. The method for load balancing unstructured peer-to-peer networks based on mobile agents of claim 4, wherein "will be calculated" in step S21

The value is compared with a set load state threshold value to determine the current load state of the node, which is specifically: two set thresholds U_thred、B_thredIf, ifNode P_jIs overloaded, the variable flag =1, if

Node P_jIs normally loaded, the variable flag = 0. If it is

Node P_jIs lightly loaded, and the variable flag = -1.

8. The method for load balancing of a structureless peer-to-peer network based on mobile agents according to claim 7, wherein the step S210 specifically comprises: the node P_jCurrent variable state flag =1, node P_jNot accepting other node connection requests; if node P_jCurrent state flag =0, and at this time if node P_jThe last variable state of flag =1, then the node P_jNot accepting other node connection requests; if node P_jLast variable state flag =0 or-1, and node P_jNode P with current variable state flag =0_jConnection requests of other nodes can be accepted; if node P_jIf the current variable state flag = -1, the node P_jConnection requests of other nodes may be accepted.

9. A system for load balancing a structureless peer-to-peer network based on mobile agents, the system comprising:

the topology adjusting unit is used for each node in the network to periodically initiate a topology adjusting mechanism and connect the topology adjusting mechanism to the nodes with strong processing capacity to form a unit of a central topology structure;

a topology optimization unit for judging the congestion state of the node by calculating the congestion level of the node at a certain moment, and further performing network topology optimization;

wherein the topology adjusting unit includes:

node P_iA unit for sending a collecting message to collect a candidate node set participating in the node topology adjustment; forwarding the node collecting message to a randomly selected neighbor node, and adding the node to a set Si; set of candidate nodes S_iReturned to the message originating node P_iA unit of (1);

slave node P_iSelecting the node P with the weakest processing capability and the connectivity degree larger than 1 from the current neighbor node set_min(Ni)A unit of (1);

from the set of candidate nodes S_iThe node P with the strongest processing capability is taken out_max(Si)Comparing node P_max(Si)And P_min(Ni)To the node P_max(Si)A unit that issues a connection request;

node P_iNeighboring node P of disconnection and selection_min(Ni)And is connected with node P_max(Si)A unit for establishing a connection;

the topology optimization unit includes:

computing node P for collecting information by mobile agent_jCongestion level at the current time t

Setting three load states of overload, normal load and overload for each node, setting load state threshold value, calculatingThe value is compared with a set load state threshold value to determine the current load state of the nodeA unit of a state;

node P_jRandomly selecting a part of nodes to disconnect, and deriving a unit of a sub-mobile agent;

the child mobile intelligent agent takes the generation node as a father node, searches nodes with redundant processing capacity in the network in a breadth-first traversal mode, judges the load state of the node when reaching one node, randomly selects a node to be reconnected to connect with the node if the node is lightly loaded, and then continuously searches the next node until all the nodes to be reconnected are reconnected, and the child mobile intelligent agent actively quits the network unit;

mobile agent according to node P_jOf neighbor node P_oThe attraction to the mobile agent and the list of visited nodes to select the next visited node and migrate to the unit of that node.

10. The system according to claim 9, wherein the node P is configured to perform load balancing for unstructured peer-to-peer networks based on mobile agents_jCongestion level at the current time tThe calculation formula of (2) is as follows:wherein,representing a node P_jThe time required to process a query message; if a node is processing a message then all incoming messages at that time are placed in the node's message buffer queue,

indicating the time t node P_jThe number of messages to be processed in the buffer queue;

the neighbor node P_oThe attraction to a mobile agent is calculated as follows:

<math> <mrow> <msub> <mi>A</mi> <msub> <mi>P</mi> <mi>o</mi> </msub> </msub> <mo>=</mo> <msub> <mi>C</mi> <msub> <mi>P</mi> <mi>o</mi> </msub> </msub> <mo>&times;</mo> <mi>&chi;</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>,</mo> </mrow> </math>

is a neighbor node P_oRepresents the neighboring node P_oThe time required to process a query message; x (P)_oK) is a node P_oThe connectivity of (c); node P_oAttractive force

The larger the mobile agent slave P_jMigration to node P_oThe higher the probability of (c);

the neighbor node P_oThe connectivity calculation formula is as follows:

<math> <mrow> <mi>&chi;</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>h</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>h</mi> <mo>=</mo> <mi>k</mi> </mrow> </munderover> <mfrac> <mrow> <mi>N</mi> <mrow> <mo>(</mo> <msub> <mi>P</mi> <mi>o</mi> </msub> <mo>,</mo> <mi>h</mi> <mo>)</mo> </mrow> </mrow> <msup> <mi>h</mi> <mi>&delta;</mi> </msup> </mfrac> <mo>,</mo> </mrow> </math>

wherein, N (P)_oH) represents a neighbor node P_oThe number of nodes which are h hops away, k represents the radius of the calculated connectivity, delta is a control coefficient, 1/h^δWeight, χ (P), for controlling influence of nodes of different distances on connectivity_oK) the larger the node P_oThe more query messages that may be received.

Images