CN1845524A - Peer-to-peer network system realization process using cell dynamic division technology - Google Patents

Abstract

The provided P2P method for information interaction and system share comprises: applying the unique register server in system to manage all node ID for uniqueness as well as some node information; according to performance and system state dividing all nodes into the common ones and superior ones both managed by the super ones; taking dynamic subarea management to all users. This invention improves system performance and expandability.

Description

Adopt the peer-to-peer network system implementation method of cell dynamic division technology

Technical field the present invention relates to Peer-to-Peer (P2P) peer-to-peer network technology, refers to that especially a kind of P2P of utilization network technology carries out information interaction, shared systems approach on Internet.

Background technology is made a general survey of the P2P development of internet technology, because its dependence to server service quality little, that provide for client is better than the client-server pattern to a certain extent, makes it at the recent universal great development that obtained along with broadband access.Search for and safeguard but generally adopt in the existing systems, consumed massive band width and caused resource utilization low, in fact brought great burden, become a barrier that hinders the P2P development to telecom operators based on " inundation method ".

Summary of the invention main purpose of the present invention is to provide a kind of information interaction based on the P2P network technology, shared system method.System carries out dynamic partition management with all users, has simultaneously to elect some nodes the information of sub-district is managed.When the node in the system is initiated another node of search, will search by the node that elects, changed the pattern that floods on a large scale.

Embodiment adopts the peer-to-peer network system implementation method of cell dynamic division technology to have following characteristics:

1. whole system does not need other servers except that a unique registrar of the overall situation, and adopts directed resource locate mode can reduce the propagation of search information in the network greatly.Under the situation of optimum, one query only need be done in the resource location on a node; In the worst case, also only need on three nodes, do one query respectively.

2. each node in the system all has the unique sign of system (ID), the Performance Evaluation value in network.The ID of each node by the user first when the system registry by user record, registrar is guaranteed its uniqueness of the overall situation; According to the performance of node, one co-exists in three types node in the system: ordinary node (Common Node, CN), higher-level node (Higher Node, HN), (Super Node, SN), and performance increases progressively super node successively.Quantitative proportion between each category node is determined by system design parameters, and is supported dynamically to adjust.

3. each node is preserved an adjacent IP-ID routing table according to its type, is unique instrument of other nodes of search.All kinds of node IP-ID routing table structure is all different: CN preserves the IP-ID information of its HN, SN of management, so HN preserves the IP-ID information of CN of its management and the IP-ID information of managing its SN, SN preserves the IP-ID information of all HN of the IP-ID information of all SN in the system and its management.

4. because the sub-district bound of node unique ID and SN, HN is all English alphabet, numeric string, so the reference character dictionary order can be determined the partial ordering relation of any two signs or sub-district bound: only need to press bit comparison from left to right, in the time can distinguishing partial ordering relation, can stop.

5. establishing RH, RL is respectively the upper and lower border of a HN, and ID is the unique identification of CN, and a HN manages N CN at most, and then according to lexicographic order, the CN that this HN managed is such set:

{ID _i，ID _j|RL＜ID _i≤RH(i＝0，1，2L，N-1)，ID _j＜ID _j+1(j＝0，1，2L，N-2)，ID _i≠ID _j(i≠j)}。

6. the union of the sub-district of all HN of the some SN management cell range of SN administration therewith equates, and the cell range that each HN administered do not have overlapping, that is:

$\underset{i}{\∪}{\mathrm{HN}}_i.\mathrm{Area}=\mathrm{SN}.\mathrm{Area},\∀{\mathrm{HN}}_i\∈\mathrm{SN}$

And {  i ≠ j|HN _iAreaI HN _jArea=; HN _i, HN _j∈ SN}.

7. the CN of a HN management some when the quantity of management CN reaches the threshold value (N) of HN, will carry out the ordinary node cell splitting.The principle of division is that CN that HN managed of script is divided equally, and half CN continues management by original HN; And from second half CN, select the highest node of performance number and upgrade to HN, the CN that management is remaining.In the process of division, need safeguard that also need upgrade the routing table of the SN of management HN to the routing table of new HN.The number of node is time dependent in the network, and CN can withdraw from a large number in the sub-district in a period of time, produces the CN number situation seldom of a HN management.If adjacent two CN number sums that HN managed of a SN subordinate are less than k ₁, then these two sub-districts that HN managed will be merged.

8. a plurality of HN of the same management of SN when the quantity of HN reaches the threshold value (N ') of SN, adopt and the similar method of ordinary node cell splitting, and the sub-district is divided equally.New SN node is a node of selecting best performance from the subordinate CN of first HN of this SN management, upgrades to SN.Relevant HN, SN routing table all needs to upgrade.

When node after server registration finishes, server will return in the node system IP-ID routing iinformation of a current online SN, node is kept at it in IP-ID routing table.When a node adds fashionablely, its default type is CN, and this node is the record of inquiry HN and send packet and confirm that it is online in the IP-ID routing table at first:

A. if this HN is online, then CN sends to HN with its ID.HN carries out the comparison of ID and this sub-district bound, if CN belongs to this sub-district, then IP, the id information with CN is kept in the IP-ID routing table, HN judges whether the CN quantity of being managed surpasses threshold value (N) then, if do not surpass, then CN logins successfully, otherwise after HN carried out the ordinary node cell splitting, CN finished login.If CN does not belong to this sub-district, HN searches another one HN in the IP-ID routing table, if existence and CN belong to this HN, then its IP, ID is returned to CN, and CN restarts login process; If do not exist, then the regional extent that ID and SN managed with CN compares, if CN belongs to this SN, the IP, the ID that then return SN give CN, and CN restarts login process again, does not belong to this SN as if CN, and then SN will return the correct SN of CN CN is logined again.

B. if this add ingress do not find or write down in HN not online, will get in touch with SN.SN at first will add the comparison of ID and this sub-district bound of ingress after receiving CN information.If CN belongs to this sub-district, then SN searches the HN of management CN in its IP-ID routing table, if find, then IP, the ID with HN sends to CN, and CN continues HN login from then on; If do not find HN, illustrating does not have HN to exist among the SN, then this CN is upgraded to HN.If CN does not belong to this sub-district, because the information synchronization of other SN maintenances IP, ID in SN and the system, then SN searches the SN of management CN in its IP-ID routing table, and the result is returned to CN, restarts login process.

When CN in HN, SN login when all failing, the SN that exists in will inquiry registrar system, server return to IP, the id information of the current online SN of CN, CN restarts login process.

10. when just setting up in system, without any node, only there is a registrar in the system.When first node adds system, server will be set to SN to this node this moment.

11. after each node login system, regularly send online message to upper layer node, upper layer node confirms that with this this node is in system.Upper layer node will return to lower level node online information equally after receiving the online information of lower level node.Node failure is in two kinds of situation: normally withdraw from and go offline.

A. when CN normally withdraws from, need to notify it to withdraw to its HN of management, HN is the IP-ID item of this node of deletion in the IP-ID routing table then; When CN went offline, HN behind overtime certain hour, judged that this CN logs off owing to do not receive the online information that CN sends on time, and corresponding record deletion in the IP-ID routing table.

B. when HN normally withdraws from, to in the CN of its management, select a highest node of performance number, and it will be made as HN, after the IP-ID of the HN that will withdraw from information reproduction is given new HN, the IP-ID routing table update of SN, CN that new HN is correlated with, this moment, old HN logged off.When HN goes offline, the CN of its management will can not get the online information returned from HN, and this moment, CN was with the duplicate node login process, picked up HN.In this process, because the SN of management HN also will can not receive the online information of HN, SN will at first select less one of management CN number from two HN that HN is adjacent therewith, revise the going up or lower bound of scope of its administration, make the range of management that covers the HN that goes offline.The IP-ID routing table update of corresponding C N, HN, SN also is necessary.

During the c.SN node failure, the HN of its management also can not receive the online feedback information of SN.This moment, a plurality of HN respectively selected the node of a best performance from the own CN that is managed, and through relatively, selected a highest CN of performance number as new SN again, and corresponding IP-ID routing table also will obtain upgrading.

Claims (9)

1, a kind of peer-to-peer network system implementation method that adopts cell dynamic division technology, it is characterized in that: this method is carried out the system node district dynamic according to the unique ID (English alphabet, numeric string) of node in the network and is divided, and node contains relevant for the routing table of other nodes in the network (reflecting the ID of other nodes and the corresponding relation of IP address), and system configuration belongs to half distributed structural model.

2, method according to claim 1 is characterized in that: each node in the system all has the unique sign of system (ID), the Performance Evaluation value in network.

3, method according to claim 1 is characterized in that: the scale of sub-district, promptly the number of node in sub-district is set when system initialization, can adjust according to network condition, and have the bound threshold value.

4, method according to claim 3 is characterized in that: when number of users surpasses upper bound threshold value in the sub-district, carry out the division of sub-district; , carry out the sub-district and merge during at the community user number less than the lower bound threshold value.

5, method according to claim 1 is characterized in that: node type has three kinds in the system: ordinary node (CN), higher-level node (HN), super node (SN).

6, method according to claim 5 is characterized in that: each node types determine with its Performance Evaluation refer to system in the state of sub-district relevant, the ratio of three types of number of nodes is set when system initialization, and dynamically adjusts.

7, method according to claim 5 is characterized in that: the node type in the sub-district is an ordinary node all, and is managed by higher-level node, and the higher-level node of some is managed by super node again.

8, method according to claim 1 is characterized in that: when a node is initiated other nodal informations of inquiry, at first search in the sub-district by the higher-level node of managing this sub-district, if find destination node, then finish to search.

9, method according to claim 1, it is characterized in that: when a node is initiated other nodal informations of inquiry, at first in the sub-district, search, if do not find by the higher-level node of managing this sub-district, then assist to search, finally return the positional information of destination node by super node.