CN101378325A - Superposition network and building method thereof - Google Patents
Superposition network and building method thereof Download PDFInfo
- Publication number
- CN101378325A CN101378325A CNA2007101458430A CN200710145843A CN101378325A CN 101378325 A CN101378325 A CN 101378325A CN A2007101458430 A CNA2007101458430 A CN A2007101458430A CN 200710145843 A CN200710145843 A CN 200710145843A CN 101378325 A CN101378325 A CN 101378325A
- Authority
- CN
- China
- Prior art keywords
- node
- peer
- father
- information
- peer node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Computer And Data Communications (AREA)
Abstract
The invention discloses an overlay network and a construction method thereof, and a peer node discovery method and user nodes, which pertain to the network communication field, wherein, the construction method comprises the steps that: a rebroadcast node list is created for each live broadcast channel, a distributed hash table ring is created for each live broadcast channel. The discovery method comprises the steps that: peer node information corresponding to the channel is searched in the rebroadcast node list; performance detection is carried out to optimal nodes and the optimal peer nodes are taken as the candidate father nodes; the minimum amount of optimal peer nodes meeting the system performance requirements are taken as the father nodes, transmission requests for different block data are respectively initiated to the father nodes. The overlay network comprises a list server, a streaming server and user nodes. The user node comprises a peer node searching module, a candidate father node storage module and a father node selection module. By adopting the technical proposal, the invention has better transmission performance in the start-up initial stage, is convenient for being expanded and can raise the peer node discovery speed.
Description
Technical field
The present invention relates to network communication field, particularly a kind of overlapping network and construction method thereof that is applicable to multi-source peer-to-peer network flow medium live system, and the user node in the discover method of peer node and this overlapping network in this overlapping network.
Background technology
Along with extensively popularizing and the continuous maturation of multimedia technology of the Internet, how the key that provides high-quality live streaming media service to become to promote internet development to the large-scale concurrent customer group on Internet is used.Recent years, the researcher has proposed based on peer-to-peer network technology (P2P, flow medium live system Peer-to-Peer).Its basic thought is, the data route relevant with the IP multicast, duplicate, application layer (rather than router) that function such as forwarding is transferred to member's main frame (user node) realizes, makes up an overlapping network of realizing multicast functionality between user node.As the most popular current Internet technology, various fields have been widely used in, for example video conference (ESM application software), Web TV (PPLive application software, PPstream application software) etc. based on the flow medium live system of peer-to-peer network.
Provide the mode of data according to its source node, can be divided into two kinds based on the flow medium live system of peer-to-peer network: single source flow medium live system and multi-source flow media live broadcast system.
Referring to Fig. 1, single source flow medium live system is meant that all relay user node and multicast tree of streaming media server formation is carried out transfer of data.Streaming media server and reception/provide the user node of same channel content to be called peer node.Each peer node is the recipient of data, also is the supplier of data.Streaming media server is a special node, only as the supplier of data, is called source node.Each peer node in the tree structure that single source flow medium live system constitutes has and has only data source supplier, i.e. a streaming media server.Among Fig. 1, heavy line is represented the flow direction of data.
Referring to Fig. 2, the multi-source flow media live broadcast system is meant that all relay user node and netted (Mesh) structure of streaming media server formation is carried out transfer of data.Each peer node on the network structure can have a plurality of data source suppliers simultaneously, and the recipient is respectively to the different data block of different pieces of information source request.Multi-source system is united by a plurality of data sources provides data to single recipient, not only can effectively improve the quality of reception of each user node, and can alleviate in the peer-to-peer network because user node frequently adds/withdraw from the dynamic problem that causes.
Typical case's representative of multi-source flow media live broadcast system is the DONet (Data DrivenOverlay Network, the overlapping network of data-driven) of people's propositions such as the Zhang Xin of Hong Kong University of Science and Thchnology grinds.In DONet, the maintenance and management of peer node information is as follows:
Each node all has an identifier (for example IP address) that the overall situation is unique in DONet, safeguards a buffer memory mCache simultaneously, in order to the information of other peer node in the storage system.Some other peer node of safeguarding among each node meeting and the mCache is set up cooperation relation, and these peer node of setting up cooperation relation are called partner node.
In dynamical system, in order to create and safeguard mCache, each node needs periodically to send the message of declaration oneself survival, and message format is (seq_num, id, num_parnter, time_to_live), wherein seq_num is the sequence number of message, and id is the sign of node, num_parnter is partner's number of node, and time_to_live is the remaining life span of this message.
When node was received a new information about firms, its can find the information about firms record of corresponding id in mCache, if seq_num then upgrades this record greater than the seq_num in the record.If do not find the record of corresponding id, then in mCache, add this information about firms of record storage.In mCache, every information about firms record comprises 5: seq_num, id, num_parnter, time_to_live, last_update_time.
The key issue that nodal information is safeguarded among the DONet is how to create and upgrade mCache.Although it is an extendible agreement of lightweight that DONet compares based on the scheme of gossip with tradition, be the information of part peer node in each node maintenance system, but can see along with the increase that adds system time, the number of peer node can increase thereupon among the mCache of each node maintenance, thereby causes memory cost to increase.
In addition, it is also bigger that the membership message that each node periodically sends the own survival of declaration among the DONet spreads the control expense of bringing, and the implementation complexity of SCAM agreement is also higher.
In DONet, add fashionable as new user node, it at first connect (Contact) with source node, source node selects a node as agent node (deputy) from the mCache of local maintenance at random, new joining request of user node is transmitted to this agent node, new user node obtains partner node information from agent node, and with these partner node swap data validity information, carry out transfer of data.Owing to adopt the transmission performance select agent node and partner node at random and not consider node, may make that the transmission performance of the partner node selected when new user node is initial is relatively poor, thereby cause total reception bandwidth can not satisfy the transmission requirement of Streaming Media.
In order to improve the quality of reception of node, each node can periodically select a group node to rebulid partnership as new partner node from its mCache at random among the DONet, has the node of better transmission performance with detection.Though this optimized choice mechanism can be improved the quality of reception to a certain extent, but can periodically introduce the control expense, and be subjected to the restriction of the characteristics of SCAM agreement own, because nodal information transmits and safeguards that it is slower to seek the process with better transmission performance node by probability.
Summary of the invention
In order to make overlapping network have better transmission performance in initial start stage, to be convenient to expansion and accelerate to find peer node speed, the embodiment of the invention provides the discover method and the user node of a kind of overlapping network and construction method thereof, peer node.Described technical scheme is as follows:
A kind of construction method of overlapping network said method comprising the steps of:
Create one for each direct broadcast band and relay node listing, in described relay node listing, safeguard the part peer node information of described direct broadcast band correspondence;
Create a distributed hashtable ring for each direct broadcast band, on described distributed hashtable ring, safeguard all peer node information of described direct broadcast band correspondence by the distributed pointer gauge of each peer node.
The discover method of peer node in a kind of overlapping network said method comprising the steps of:
Wish to receive the new user node of certain channel data to relaying the peer node information that node listing is searched this channel correspondence;
Described new user node receives the preferred peer node information in the relay node listing of the described channel correspondence of selecting according to metric parameter, described preferred node is carried out after performance surveys, described preferred peer node as alternative father node;
Described new user node is according to multi-parameter father node selection strategy, and the preferred peer node of selecting to satisfy the minimum number that systematic function requires from described alternative father node is as father node, and initiates transmission request to the different pieces of information piece respectively to described father node.
A kind of overlapping network, described overlapping network comprises LIST SERVER, streaming media server and user node;
Described LIST SERVER is used to each direct broadcast band to create one and relays node listing, safeguards the part peer node information of described direct broadcast band correspondence in the described relay node listing;
Described streaming media server and user node constitute the distributed hashtable ring, the corresponding distributed hashtable ring of each direct broadcast band is safeguarded all peer node information of described direct broadcast band correspondence by the distributed pointer gauge of each peer node on described distributed hashtable ring.
User node in a kind of overlapping network, described user node comprises: peer node is searched module, alternative father node memory module and father node and is selected module;
Described peer node is searched the preferred peer node information that module is used for searching to LIST SERVER the relay node listing of the channel data correspondence of wishing reception, and described preferred peer node information is sent to described alternative father node memory module;
Described alternative father node memory module is used for receiving described preferred peer node information, and described preferred node is carried out storing described preferred peer node information after performance surveys;
Described father node selects module to be used for according to multi-parameter father node selection strategy, the preferred peer node of selecting to satisfy the minimum number that systematic function requires from described alternative father node memory module is as father node, and initiates transmission request to the different pieces of information piece respectively to described father node.
The embodiment of the invention adopts the peer node maintenance mechanism that mixes, and safeguards whole peer node information by the distributed hashtable ring by certain regular distribution formula ground, and when the user node number increased, memory cost can sharply not increase, and had improved the autgmentability of system.Because the distributed hashtable ring is the distributed maintenance of structuring, with respect to the non-structured maintenance mode of DONet, maintenance costs is controlled, makes maintenance costs remain on lower grade.
By the information of the preferred node of centralized ground service portion on LIST SERVER, accelerated the speed that new user node is found other peer node information, shortened the startup time-delay of system, and user node just can obtain the quality of reception preferably in initial start stage.
Because user node only just obtains peer node information from LIST SERVER when adding system, therefore can effectively avoid LIST SERVER to become the single-point bottleneck of system.
Description of drawings
Fig. 1 is single source peer-to-peer network flow medium live system schematic diagram that prior art provides;
Fig. 2 is the multi-source peer-to-peer network flow medium live system schematic diagram that prior art provides;
Fig. 3 is the flow chart of the overlapping network construction method that provides of the embodiment of the invention 1;
Fig. 4 is the structural representation that the peer node information that provides of the embodiment of the invention 1 is mixed service mode;
Fig. 5 is that node number that the embodiment of the invention 1 provides is the schematic diagram of 6 DHT ring;
Fig. 6 is that the new user node that the embodiment of the invention 1 provides adds the sequential schematic diagram that DHT encircles;
Fig. 7 is the structural representation after the new user node (N=20) that provides of the embodiment of the invention 1 adds the DHT ring;
Fig. 8 is the flow chart of the discover method of the peer node that provides of the embodiment of the invention 2;
Fig. 9 is the flow chart of the multi-parameter father node selection strategy method that provides of the embodiment of the invention 2;
Figure 10 is the structural representation of the overlapping network that provides of the embodiment of the invention 3;
Figure 11 is the user node structural representation in the overlapping network that provides of the embodiment of the invention 4.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, embodiment of the present invention is described further in detail below in conjunction with accompanying drawing.
The embodiment of the invention provides a kind of construction method that is used for the overlapping network of multi-source peer-to-peer network flow medium live system, comprising:
Create one for each direct broadcast band and relay node listing, in relaying node listing, safeguard the part peer node information of direct broadcast band correspondence; Create a distributed hashtable ring for each direct broadcast band, on the distributed hashtable ring, safeguard all peer node information of direct broadcast band correspondence by the distributed pointer gauge of each peer node.
In this overlapping network construction method, the maintenance of peer node information and discovery mechanism can adopt the network architecture that combines LIST SERVER and DHT (Distributed Hash Table, distributed hashtable) ring based on.Under this framework, peer node information adopts the service mode that mixes, it is included in the information of service portion preferred peer node in centralized ground on the LIST SERVER, and create a DHT for each direct broadcast band and encircle, user node adds different DHT rings respectively according to the channel of desired reception, by the distributed pointer gauge of each peer node, distributed earth is safeguarded all peer node information on the DHT ring.Referring to Fig. 3, the said method concrete steps are as follows:
Step 101: LIST SERVER is set, on LIST SERVER, is each channel of being issued on it, create and safeguard a tabulation of relaying the preferred peer node information of part of this channel data, be called the relay node listing.
When a streaming media server is wished when the Internet user node provides streaming media service, it need be registered the relevant information of issue channel to LIST SERVER, just channel information (channel designation, content introduction, creation-time, code rate) and this nodal information (IP address, port numbers, uplink bandwidth, son node number, time-to-live) are sent to LIST SERVER.Accept when LIST SERVER after the registration of this streaming media server, can on Web, show the relevant information of issue channel, check and click to offer user node.
Referring to Fig. 4, be used for the structural representation that peer node information is mixed service mode in the construction method of the overlapping network of the multi-source peer-to-peer network flow medium live system that the embodiment of the invention provides.The IP address of LIST SERVER is well-known fixed address (for example 202.38.72.133) among Fig. 4.
LIST SERVER is relayed node listing for one of each channel creation registering on this machine, relaying the preferred part peer node information of this channel in order to storage.Referring to table 1, in this table content, the first half comprises: the nodal information of the user node of streaming media server and this channel correspondence, the latter half comprise the channel information of streaming media server to the LIST SERVER registration.
Table 1
In the constructive process of relaying node listing, the streaming media server of the streaming media service of at first knowing that provides each channel, initiate then user node need be to the LIST SERVER request, the peer node of knowing when LIST SERVER outnumbers when relaying the maximum node number that node listing can safeguard, just no longer receives new user node.
Step 102: for each direct broadcast band is created a distributed hashtable ring, safeguard all peer node information of direct broadcast band correspondence on the distributed hashtable ring, each peer node on the distributed hashtable ring realizes the maintenance of peer node information by distributed pointer gauge.
Organize themselves into a circulus by all peer node (comprising streaming media server) that receive same channel, be called the DHT ring, the form of each peer node on the ring by distributed pointer gauge realizes the maintenance to part peer node information, all nodes information of all nodes of joining together just to have safeguarded on the DHT ring.
Referring to the ring of the DHT among Fig. 41, be that peer node (server that comprises channel 1) self-organizing by all receiving channelss 1 forms, each peer node on the ring is all safeguarded a pointer gauge, passes through the distributed maintenance of the form realization of pointer gauge to peer node information just.Wherein, the storage rule of pointer gauge is as follows:
Node identifier is N, then in the pointer gauge the 1st maintenance be on the annulus node identifier more than or equal to N+2
0First peer node (with 2
mBe mould) information, by that analogy, the i item is safeguarded in the pointer gauge be on the annulus node identifier more than or equal to N+2
I-1The information of first peer node.
Referring to table 2, each peer node is safeguarded following parameter in the pointer gauge: the node identifier of this node correspondence, IP address, port numbers, uplink bandwidth, time-to-live, child node number.
Table 2
| Node identifier | The IP address | Port numbers | Uplink bandwidth (Kbps) | Son node number | Time-to-live (s) |
| N+2 0 | 210.45.121.107 | 1024 | 400 | 2 | 1500 |
| N+2 1 | 210.45.121.107 | 1024 | 400 | 2 | 1500 |
| N+2 2 | 210.45.121.107 | 1024 | 400 | 2 | 1500 |
| N+2 3 | 210.45.121.107 | 1024 | 400 | 2 | 1500 |
| N+2 4 | 210.45.121.107 | 1024 | 400 | 2 | 1500 |
| N+2 5 | 211.86.149.94 | 1024 | 500 | 3 | 600 |
Referring to Fig. 5, the example of the DHT ring of a m=6 is provided, 6 nodes that distributed in the ring, and provided the pointer gauge of the peer node of node identifier N=20, for example the peer node of N=41 is that ring tightens and is connected on N=(20+2
0) mod2
6First node after=21 is so the 1st of the pointer gauge that the peer node of N=20 is safeguarded is the information of the peer node of N=41; In like manner, because being ring, the peer node of N=63 goes up node identifier more than or equal to N=(20+2
5) mod2
6First node of=52 is so the 6th of the pointer gauge that the peer node of N=20 is safeguarded is the information of the peer node of N=63.Wherein, N=41, the peer node of N=63 all is called the descendant node of the peer node of N=20, and the peer node of N=41 is called the 1st descendant node again.
Referring to Fig. 6, the sequential chart that provides concrete new user node (202.38.75.11) to add the DHT ring is supposed existing 5 peer node on the DHT ring.
(1) new user node (202.38.75.11) signs in to LIST SERVER (202.38.72.133) and searches, and the request relevant information;
(2) LIST SERVER is peer node information＜202.38.64.62,202.38.64.10,210.45.121.107,211.86.149.94,148.233.229.235 in the relay node listing of this channel correspondence〉return to new user node;
(3) new user node uses the IP address (202.38.75.11) of hash function Hash oneself to obtain the node identifier N=20 of this node, select a node (N=43 then at random, IP=202.38.64.10) as the search start node, initiate searching request to it, the search node identifier is keyword (key=21,22,24,28,36,52) descendant node information;
(4) N43 is after the request of receiving, at first judge search key key=21,22,24,28,36,52 are positioned at outside the interval that this node identification symbol N43 and the 1st descendant node identifier N46 identified, so search pointer gauge, find key=21,22,24,28,36 are positioned within the interval of N63 and N41 sign (crossing the N0 point of DHT ring), so to key=21,22,24,28,36 searching request is handed to N63, and key=52 is positioned within the interval of N46 and N63 sign, so the searching request of key=52 is handed to N46;
(5) N63 receives key=21,22,24,28,36 searching request is searched pointer gauge, finds key=21,22,24,28,36 are positioned within the interval of N5 and N41 sign, so searching request is handed to N5, N5 finds key=21,22,24,28,36 are positioned at the interval of the 1st descendant node N41 sign of N5 and it, then N41 as key=21,22,24,28,36 descendant node information returns to new user node; Equally, N46 finds that key=52 is positioned at the interval of N46 and its 1st descendant node N63 sign, so N63 is returned to new user node as the descendant node information of key=52;
(6) new user node is after receiving the descendant node information of returning, the local pointer gauge key=21 of initialization, 22,24,28,36 → N41, key=52 → N63;
(7) new user node sends response message respectively to N41 and N63, asks these descendant nodes to add the backward pointer table that they are safeguarded through the information with oneself;
(8) N41 and N63 join the information of N20 in the backward pointer table of oneself safeguarding after receiving response message;
(9) N41 and N63 send reverse affirmation to N20;
(10) N20 searches the information N41 of the 1st descendant node in the pointer gauge after receiving for the first time the reverse acknowledge message that descendant node is sent, and sends to it and searches message;
(11) N41 receive search message after, judge that N20 is the new forerunner's node of oneself, so search all the node＜N5 in the own backward pointer table, N41, N43, N46, N63 〉, send update request to these nodes, tell these nodes the information of N20;
(12) after N5 receives updating message, check pointer gauge, discovery need be key=6, and the descendant node information updating of 7,9,13 correspondences is N20; Same N41 finds to be N20 to the descendant node information updating of key=9 correspondence; It need be the descendant node information updating of key=11 correspondence N20 that N43 finds; It need be the descendant node information updating of key=14 correspondence N20 that N46 finds; It need be the descendant node information updating of key=15 correspondence N20 that N63 finds;
(13) N5, N41, N43, N46, N63 find that N20 is an emerging node in the pointer gauge, so notice N20 adds the backward pointer table that N20 safeguards to oneself;
(14) N41, N43, N46 find that respectively N41 disappears from the pointer gauge of oneself, so notice N41 deletes N41, N43, N46 from the backward pointer table that N41 safeguards.
Referring to Fig. 7, provide the structural representation after new user node (N=20) adds the DHT ring.
Above-mentioned two steps have realized the mixing service mode of peer node information, except information, also the information of all peer node is carried out distributed maintenance by a circulus (DHT ring) by the preferred peer node of centralized ground of LIST SERVER service portion.Wherein the corresponding DHT of each channel encircles, and each ring all keeps being connected with LIST SERVER.
Step 103: survey the peer node information of safeguarding on the distributed hashtable ring, upgrade according to metric parameter and relay node listing.
Because on LIST SERVER, for each channel safeguards that is all relayed a node listing, whenever new user node adds the fashionable peer node information of safeguarding in the node listing that all at first obtains to relay.If it is constant to remain the information of these peer node, may make these nodes become potential performance bottleneck.Adopt selection strategy to upgrade the relay node listing of each channel according to metric parameter.Its concrete steps are as follows:
(1) LIST SERVER just selects a peer node as destination node from relay node listing every a timing cycle at random, sends update request to it;
(2) after destination node was received request, all peer node information that this node is safeguarded, promptly all peer node information of safeguarding in the local pointer gauge of this node, backward pointer table, the tabulation of alternative father node returned to LIST SERVER;
(3) for LIST SERVER, if arriving, the peer node number of newly knowing requires (for example 10), then stop to survey, and according to selection strategy, promptly select metric parameter that all nodes of knowing are sorted, the information of the peer node of poor-performing is deleted from relay node listing, upgrade relaying node listing; If do not reach requirement, then repeat above-mentioned steps (1) (2) up to the peer node number of newly knowing meet the demands or number of repetition reach set require till.
Wherein, use the metric parameter of the combination of following parameter or these parameters as selection strategy:
(1) time-to-live of relay user node, promptly this user node receives the time span of this channel, and suggestion keeps long relay user node of time-to-live;
(2) uplink bandwidth of relay user node, promptly this user node can provide the bandwidth of transfer of data for other user nodes, and suggestion keeps the bigger relay user node of uplink bandwidth;
(3) uplink bandwidth and the son node number purpose ratio of relay user node, be the fan-out bandwidth, the fan-out bandwidth is big more, illustrates to relay user node to offer the output bandwidth of its each child node big more, transmission performance is good more, and therefore suggestion keeps the bigger relay user node of fan-out bandwidth.
The embodiment of the invention provides the discover method of peer node in a kind of overlapping network, comprising:
Wish to receive the new user node of certain channel data to relaying the peer node information that node listing is searched this channel correspondence;
New user node receives the preferred peer node information in the relay node listing of the channel correspondence of selecting according to metric parameter, preferred node is carried out after performance surveys, preferred peer node as alternative father node;
New user node is according to multi-parameter father node selection strategy, and the preferred peer node of selecting to satisfy the minimum number that systematic function requires from alternative father node is as father node, and initiates transmission request to the different pieces of information piece respectively to father node.
Referring to Fig. 8, the concrete implementation step of the discover method of peer node in a kind of overlapping network that present embodiment provides:
Step 201: when new user node wishes to receive the data of certain channel, sign in to the peer node information that LIST SERVER is searched this channel correspondence, to obtain suitable relay node.
Step 202: after the preferred peer node information of new user node in the relay node listing that receives the channel correspondence that LIST SERVER selects according to metric parameter, M the peer node of knowing carried out the performance detection respectively.
Wherein, in actual conditions, for the phenomenon that prevents that a plurality of nodes from breaking away, it is more generally to relay the interstitial content that node listing safeguards, and for example 40 or more, and it is generally less to return to the interstitial content of new user node, for example 10.
Wherein, performance parameter comprises: peer node time-to-live, internodal end-to-end delay, the uplink bandwidth of peer node, child node number etc.
Step 203: new user node deposits these information in the alternative father node tabulation of oneself after detecting above-mentioned performance parameter.
New user node can be known the information of other active on current network peer node.
Wherein, because the node in the peer-to-peer network may add and withdraw from continually, the also possibility dynamic change of the performance of each node, therefore user node need upgrade the alternative father node tabulation cycle of carrying out of local maintenance, in embodiments of the present invention, can adopt a kind of correlation inquiry strategy, make user node search the information of more enlivening peer node from the DHT ring.Concrete steps are as follows:
(1) user node selects a node as destination node from alternative father node tabulation at random, sends query requests to it;
(2) after destination node is received request, all peer node information of safeguarding in the own pointer gauge are returned to user node;
(3),, then stop to survey if the peer node number of newly knowing arrives requirement for user node; If do not reach requirement, then repeating step (1) (2) meets the demands or till the iterative search number of times reaches limits value up to the peer node number of newly knowing.
(4) the maximum node number that can safeguard less than alternative father node tabulation of the peer node number of knowing when user node, then the peer node that will newly know joins alternative father node and tabulates; The peer node of knowing when user node outnumbers the maximum node number that alternative father node tabulation can be safeguarded, the metric parameter of the selection strategy correspondence that need provide according to embodiment 1 is (as the uplink bandwidth and the son node number purpose ratio of peer node, be the fan-out bandwidth) performance of these nodes is assessed, according to assessment result, with node deletion from alternative father node tabulation of poor-performing.
User node only just obtains peer node information from LIST SERVER when adding system, other the time can by in this step to the update method of alternative father node tabulation, on the DHT ring, survey peer node information, therefore can effectively avoid LIST SERVER to become the single-point bottleneck of system.Owing to adopted the correlation inquiry mode, inquiry mode is simple, can keep lower inquiry expense.
Step 204: new user node is from the alternative father node tabulation of local maintenance, according to multi-parameter father node selection strategy, selection is satisfied the preferred peer node of the minimum number that systematic function requires as the father node number, promptly select the father node tabulation that MIN peer node deposits a local maintenance in, as the data source supplier of oneself.
Step 205: new user node as father node, is initiated the transmission request to the different pieces of information piece to this MIN peer node respectively.
User node can also be initiated the optimized choice that father node is carried out with after MIN peer node set up transfer of data respectively and be connected.Simultaneously MIN is connected and carry out performance monitoring, when abnormal conditions take place when, for example, the total receiving velocity that monitors MIN connection does not satisfy the performance requirement of Streaming Media, user node can pass through multi-parameter father node selection strategy, selects peer node from alternative father node tabulation, increases the father node number one by one, meet the demands up to transmission rate, perhaps reselect new father node to replace original father node.Its basic step is as follows:
(1) user node can be monitored the message transmission rate of each connection after connecting with a plurality of father nodes;
(2) if when in monitoring period, detecting total receiving velocity and not arriving the transmission rate that this channel requires, and the father node number is counted MAX less than the maximum father node that system allows, then from alternative father node tabulation, select a new peer node to add the father node tabulation, set up a new transfer of data simultaneously and connect according to multi-parameter father node selection strategy;
(3) if the father node number has reached MAX, then user node is selected a new peer node and is connected from alternative father node tabulation, if in next monitoring period, the new transmission rate that connects is higher than transmission rate minimum in the old connection, the father node that then new peer node is replaced the minimum transmission rate correspondence upgrades corresponding father node tabulation then.
When the message transmission rate that monitors certain father node significantly descend (may cause) by multiple reasons such as node collapse or network congestions; and in a period of time, do not recover; then can from alternative father node tabulation, reselect a new peer node and replace this node, this node is deleted from the alternative node list of oneself.
Above-mentioned multi-parameter father node selection strategy is as follows, chooses first metric parameter, and alternative father node is arranged according to the order of performance of first metric parameter; First metric parameter of adjacent two alternative father nodes of Pai Lieing more according to the order of sequence, if the difference of first metric parameter of adjacent two alternative father nodes is greater than the predefined first metric parameter threshold value, then select the bigger alternative father node of first metric parameter as father node, otherwise, from selection strategy, choose second metric parameter, second metric parameter of more adjacent two alternative father nodes; If the difference of second metric parameter of adjacent two alternative father nodes is greater than the predefined second metric parameter threshold value, then select the bigger alternative father node of second metric parameter as father node, otherwise, select the bigger alternative father node of first metric parameter as father node.Referring to Fig. 9, with a specific embodiment above-mentioned multi-parameter father node selection strategy is set forth below.
Each alternative father node P (i) is by one 3 yuan group id＜IP (i), Bc (i), T (i) 〉.Wherein, i is arbitrary positive integer, distinguish alternative father node in order to label, IP (i) is the IP address of alternative father node, and Bc (i) is the fan-out bandwidth of alternative father node, as first metric parameter, T (i) is the time-to-live of alternative father node, as second metric parameter, N is alternative interstitial content, and △ Bc, △ T are predefined threshold value.
Step 301: calculate the fan-out bandwidth B c (i) of all alternative father nodes, and be metric parameter with Bc (i), alternative father node is arranged P (j) according to the order of sequence, wherein the span of j is 1 to N.In this specific embodiment, be example according to metric parameter fan-out bandwidth value by descending with alternative father node.
Step 302: parameter j is set, and making its initial value is 1.
Step 303: the fan-out bandwidth B c (j) and the Bc (j+1) that compare two alternative father node P (j) and P (j+1), judge that whether Bc (j)-Bc (j+1) is greater than △ Bc, if illustrate that then the fan-out bandwidth B c (j) of alternative father node P (j) more satisfies the selection strategy requirement, execution in step 307; Otherwise, Bc (j)-Bc (j+1) is not more than △ Bc, illustrate that fan-out bandwidth B c (j) and fan-out bandwidth B c (j+1) difference do not satisfy predefined fan-out bandwidth threshold value, the fan-out bandwidth difference that is P (j) and P (j+1) is little, then need further to select as metric parameter according to the time-to-live, promptly execution in step 304.
Step 304: relatively time-to-live T (j) and the T (j+1) of two alternative father node P (j) and P (j+1), judgement | T (j)-T (j+1) | whether greater than △ T, if, illustrate T (j) and T (j+1) between difference satisfy predefined threshold value.But need further from P (j) and P (j+1), to select long alternative father node of time-to-live, i.e. execution in step 305 then, otherwise, illustrate that the time-to-live difference of P (j) and P (j+1) is little, then select the bigger P of fan-out bandwidth (j) as father node, promptly execution in step 307.
Step 305: whether judge T (j) greater than T (j+1), if, the time-to-live of the time-to-live of P (j) greater than P (j+1) is described, then execution in step 307, otherwise, illustrate that the time-to-live of P (j) is not more than the time-to-live of P (j+1), then execution in step 306.
Step 306: selection P (j+1) is as father node and exchange P (j+1) and the position of P (j).
Step 307: select P (j) as father node.
Step 308: whether judge j less than N-1, if, then make j=j+1, promptly execution in step 309; When j=N-1, finish selection strategy.
In the foregoing description multi-parameter father node selection strategy be with fan-out bandwidth and time-to-live as the metric parameter of selecting, also can adopt other performance parameter, for example relay user's uplink bandwidth etc.
Referring to Figure 10, present embodiment provides a kind of overlapping network, comprises LIST SERVER 401, streaming media server 402 and user node 403, wherein:
LIST SERVER 401 is used to each direct broadcast band to create one and relays node listing, safeguards the part peer node information of direct broadcast band correspondence in relaying node listing.
Streaming media server 402, be used for constituting the distributed hashtable ring with user node 403, corresponding distributed hashtable ring of each direct broadcast band wherein, safeguard all peer node information of this direct broadcast band correspondence on the distributed hashtable ring, each peer node on the distributed hashtable ring realizes the maintenance of peer node information by distributed pointer gauge.
User node 403, be used for constituting the distributed hashtable ring with streaming media server 402, corresponding distributed hashtable ring of each direct broadcast band wherein, on the distributed hashtable ring, safeguard all peer node information of this direct broadcast band correspondence, each peer node on the distributed hashtable ring realizes the maintenance of peer node information by distributed pointer gauge, and is used for asking and provide data at overlapping network.
In actual applications, the function that above-mentioned LIST SERVER 401 and streaming media server 402 are realized can use a station server to realize.
Referring to Figure 11, the embodiment of the invention provides the user node in a kind of overlapping network, comprising:
Peer node is searched module 501, is used for searching to LIST SERVER 401 the preferred peer node information of the relay node listing of the channel data correspondence of wishing reception, and preferred peer node information is sent to alternative father node memory module 502.
Alternative father node memory module 502 is used for receiving preferred peer node information, and preferred peer node is carried out storing preferred peer node information after performance surveys.
Father node is selected module 503, be used for according to multi-parameter father node selection strategy, the preferred peer node of selecting to satisfy the minimum number that systematic function requires from alternative father node memory module 502 is as father node, and initiates transmission request to the different pieces of information piece respectively to father node.
Further, user node can also comprise alternative father node update module, is used for selecting one or more nodes as destination node at random from alternative father node memory module 502, sends query requests to destination node; After receiving all peer node information of being safeguarded that destination node returns, if arriving, the peer node number that user node is newly known sets requirement, then stop to survey, if the peer node number that user node is known is less than the maximum node number of the alternative father node that can safeguard, then the peer node that will newly know joins in the alternative father node; Otherwise the node of metric parameter poor-performing is deleted from alternative father node.
Further, user node can also comprise the father node optimal module, be used for user node and being connected of father node are carried out performance monitoring, when abnormal conditions take place when, from alternative father node, select peer node according to multi-parameter father node selection strategy, increase the father node number one by one, meet the demands, perhaps reselect new father node to replace original father node up to performance.
Part steps can realize by software in the technical scheme that above embodiment provides, and software is stored on the storage medium that can read, as the floppy disk of computer, and hard disk or CD etc.
The technical scheme that the embodiment of the invention provides adopts the peer node maintenance mechanism that mixes, safeguard whole peer node information by the DHT ring by certain regular distribution formula ground, when the user node number increased, memory cost can sharply not increase, and had improved the autgmentability of system.Because the DHT ring is the distributed maintenance of structuring, with respect to the non-structured maintenance mode of DONet, maintenance costs is controlled, makes maintenance costs remain on lower grade.
By the information of the preferred node of centralized ground service portion on LIST SERVER, accelerated the speed that new user node is found other peer node information, shortened the startup time-delay of system, and user node just can obtain the quality of reception preferably in initial start stage.
Because user node only just obtains peer node information from LIST SERVER when adding system, other the time can be by on DHT ring, surveying peer node information, so can effectively avoid LIST SERVER to become the single-point bottleneck of system.Owing to adopted the correlation inquiry mode, inquiry mode is simple, can keep lower inquiry expense.
By the optimized choice to father node, each user node always attempts to maximize own total reception bandwidth, promptly can maximize the quality of reception of each user node, thereby improve bandwidth availability ratio.And leaving the node influence that exists of arbitrary node is little, has improved the robustness of system.
Because the cycle of the optimized choice mechanism of father node is longer with respect to optimization cycle of the prior art, thereby has saved the control expense.And in the process of the optimized choice of father node, the father node of when the data transmission performance that monitors father node descends, just from alternative father node tabulation, selecting new father node to replace this data transmission performance to descend, so with respect to prior art, no longer produce periodic Control, further saved the control expense.
In addition, the technical scheme that the embodiment of the invention provides can support each receiving terminal to reselect father node adaptively or increase new father node, thereby can effectively adapt to the dynamic of overlapping network environment, strengthens the flatness of each user's quality of reception.
The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (14)
1. the construction method of an overlapping network is characterized in that, said method comprising the steps of:
Create one for each direct broadcast band and relay node listing, in described relay node listing, safeguard the part peer node information of described direct broadcast band correspondence;
Create a distributed hashtable ring for each direct broadcast band, on described distributed hashtable ring, safeguard all peer node information of described direct broadcast band correspondence by the distributed pointer gauge of each peer node.
2. the construction method of overlapping network as claimed in claim 1 is characterized in that, described step for a relay of each direct broadcast band establishment node listing specifically comprises:
The relevant information of the channel of issuing of receiving stream media server registration;
Relay node listing for one of each channel creation on this machine, registering, relaying the peer node information of this channel in order to storage.
3. the construction method of overlapping network as claimed in claim 1 is characterized in that, described method also comprises:
Survey the peer node information of safeguarding on the described distributed hashtable ring, upgrade the step of described relay node listing according to metric parameter.
4. the construction method of overlapping network as claimed in claim 3 is characterized in that, the step of the peer node information of safeguarding on the described distributed hashtable ring of described detection specifically comprises:
From relay node listing, select a peer node every a timing cycle at random, send update request to described destination node as destination node;
Receive all peer node information that this node that described destination node returns is safeguarded, set requirement, then stop to survey if the peer node number of newly knowing arrives.
5. as the construction method of claim 3 or 4 described overlapping networks, it is characterized in that the described step of upgrading described relay node listing according to metric parameter specifically comprises:
Select to relay one of the time-to-live of user node, the uplink bandwidth of relaying user node, fan-out bandwidth or its combination as metric parameter;
According to the metric parameter of selecting all nodes of knowing are sorted, the information of the peer node of poor-performing is deleted from relay node listing.
6. the discover method of peer node in the overlapping network is characterized in that, said method comprising the steps of:
Wish to receive the new user node of certain channel data to relaying the peer node information that node listing is searched this channel correspondence;
Described new user node receives the preferred peer node information in the relay node listing of the described channel correspondence of selecting according to metric parameter, described preferred node is carried out after performance surveys, described preferred peer node as alternative father node;
Described new user node is according to multi-parameter father node selection strategy, and the preferred peer node of selecting to satisfy the minimum number that systematic function requires from described alternative father node is as father node, and initiates transmission request to the different pieces of information piece respectively to described father node.
7. the discover method of peer node is characterized in that in the overlapping network as claimed in claim 6, and described method is further comprising the steps of:
User node selects one or more nodes as destination node from described alternative father node at random, sends query requests to described destination node;
After described user node receives all peer node information of being safeguarded that described destination node returns, if arriving, the peer node number that described user node is newly known sets requirement, then stop to survey, if the peer node number that described user node is known less than the maximum node number of the alternative father node that can safeguard, then joins the described peer node of newly knowing in the alternative father node; Otherwise alternative father node deletion with the metric parameter poor-performing.
8. the discover method of peer node is characterized in that in the overlapping network as claimed in claim 6, and described method is further comprising the steps of:
To user node and being connected of described father node carrying out performance monitoring, when abnormal conditions take place when, then from described alternative father node, select peer node according to multi-parameter father node selection strategy, increase the father node number one by one, meet the demands up to described performance, perhaps reselect new father node to replace original father node.
9. the discover method of peer node is characterized in that in the overlapping network as claimed in claim 8, and described multi-parameter father node selection strategy specifically comprises:
From metric parameter, choose first metric parameter, alternative father node is carried out sequence arrangement according to the performance of described first metric parameter;
First metric parameter of more described adjacent two alternative father nodes of arranging according to the order of sequence, if the performance difference of first metric parameter of described adjacent two alternative father nodes is greater than the predefined first metric parameter threshold value, then select the more excellent alternative father node of the first metric parameter performance as father node, otherwise, from metric parameter, choose second metric parameter, the second metric parameter performance of more described adjacent two alternative father nodes;
If the performance difference of second metric parameter of described adjacent two alternative father nodes is greater than the predefined second metric parameter threshold value, then select the more excellent alternative father node of the second metric parameter performance as father node, otherwise, select the more excellent alternative father node of the first metric parameter performance as father node.
10. as the discover method of peer node in any described overlapping network among the claim 7-9, it is characterized in that described metric parameter comprises one of following condition or combination:
Relay the time-to-live of user node;
Relay the uplink bandwidth of user node;
The fan-out bandwidth.
11. an overlapping network is characterized in that described overlapping network comprises LIST SERVER, streaming media server and user node;
Described LIST SERVER is used to each direct broadcast band to create one and relays node listing, safeguards the part peer node information of described direct broadcast band correspondence in the described relay node listing;
Described streaming media server and user node constitute the distributed hashtable ring, the corresponding distributed hashtable ring of each direct broadcast band is safeguarded all peer node information of described direct broadcast band correspondence by the distributed pointer gauge of each peer node on described distributed hashtable ring.
12. the user node in the overlapping network is characterized in that described user node comprises: peer node is searched module, alternative father node memory module and father node and is selected module;
Described peer node is searched the preferred peer node information that module is used for searching to LIST SERVER the relay node listing of the channel data correspondence of wishing reception, and described preferred peer node information is sent to described alternative father node memory module;
Described alternative father node memory module is used for receiving described preferred peer node information, and described preferred node is carried out storing described preferred peer node information after performance surveys;
Described father node selects module to be used for according to multi-parameter father node selection strategy, the preferred peer node of selecting to satisfy the minimum number that systematic function requires from described alternative father node memory module is as father node, and initiates transmission request to the different pieces of information piece respectively to described father node.
13. the user node in the overlapping network as claimed in claim 12, described user node also comprises:
Alternative father node update module is used for selecting one or more nodes as destination node at random from described alternative father node memory module, sends query requests to described destination node; After receiving all peer node information of being safeguarded that described destination node returns, if arriving, the peer node number that described user node is newly known sets requirement, then stop to survey, if the peer node number that described user node is known less than the maximum node number of the alternative father node that can safeguard, then joins the described peer node of newly knowing in the alternative father node; Otherwise alternative father node deletion with the metric parameter poor-performing.
14. the user node in the overlapping network as claimed in claim 12, described user node also comprises:
The father node optimal module, be used for user node and being connected of described father node are carried out performance monitoring, when abnormal conditions take place when, from described alternative father node, select peer node according to multi-parameter father node selection strategy, increase the father node number one by one, meet the demands up to described performance, perhaps reselect new father node to replace original father node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101458430A CN101378325A (en) | 2007-08-31 | 2007-08-31 | Superposition network and building method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101458430A CN101378325A (en) | 2007-08-31 | 2007-08-31 | Superposition network and building method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101378325A true CN101378325A (en) | 2009-03-04 |
Family
ID=40421685
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101458430A Pending CN101378325A (en) | 2007-08-31 | 2007-08-31 | Superposition network and building method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101378325A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011006375A1 (en) * | 2009-07-14 | 2011-01-20 | 中兴通讯股份有限公司 | System for network deployment and method for mapping and data forwarding thereof |
| CN101521591B (en) * | 2009-04-17 | 2011-05-04 | 杭州华三通信技术有限公司 | Multicast realization method and device of shared network |
| CN102799659A (en) * | 2012-07-05 | 2012-11-28 | 广州鼎鼎信息科技有限公司 | Overall repeating data deleting system and method based on non-centre distribution system |
| WO2013078611A1 (en) * | 2011-11-29 | 2013-06-06 | 华为技术有限公司 | Data processing method, device and client in distributed storage system |
| US8549175B2 (en) | 2009-06-09 | 2013-10-01 | Qualcomm Incorporated | Methods and apparatus for adaptively scheduling a finger stabilization algorithm |
| CN103747297A (en) * | 2014-01-09 | 2014-04-23 | 能力天空科技（北京）有限公司 | Video dynamic dispatching method, device and video playing system |
| US9240927B2 (en) | 2009-02-26 | 2016-01-19 | Qualcomm Incorporated | Methods and apparatus for enhanced overlay state maintenance |
| CN109121436A (en) * | 2015-11-25 | 2019-01-01 | 蒂米菲尔股份有限公司 | Augmentation, exploration and Maintenance Significant Items method layered |
| CN109862376A (en) * | 2019-02-28 | 2019-06-07 | 广州华多网络科技有限公司 | Live content jettison system, method, apparatus, listserv and storage medium |
| CN110932980A (en) * | 2019-11-27 | 2020-03-27 | 达闼科技成都有限公司 | Communication method, terminal, and computer-readable storage medium |
| CN111064641A (en) * | 2019-12-31 | 2020-04-24 | 上海焜耀网络科技有限公司 | Node performance detection system and method for decentralized storage network |
| CN114390338A (en) * | 2022-01-26 | 2022-04-22 | 湖南快乐阳光互动娱乐传媒有限公司 | Method and device for rapidly acquiring p2p video and electronic equipment |
-
2007
- 2007-08-31 CN CNA2007101458430A patent/CN101378325A/en active Pending
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9240927B2 (en) | 2009-02-26 | 2016-01-19 | Qualcomm Incorporated | Methods and apparatus for enhanced overlay state maintenance |
| CN101521591B (en) * | 2009-04-17 | 2011-05-04 | 杭州华三通信技术有限公司 | Multicast realization method and device of shared network |
| US8549175B2 (en) | 2009-06-09 | 2013-10-01 | Qualcomm Incorporated | Methods and apparatus for adaptively scheduling a finger stabilization algorithm |
| US8898334B2 (en) | 2009-07-14 | 2014-11-25 | Zte Corporation | System for network deployment and method for mapping and data forwarding thereof |
| CN101958830B (en) * | 2009-07-14 | 2014-03-12 | 中兴通讯股份有限公司 | Network deployment system and mapping and data forwarding method thereof |
| WO2011006375A1 (en) * | 2009-07-14 | 2011-01-20 | 中兴通讯股份有限公司 | System for network deployment and method for mapping and data forwarding thereof |
| WO2013078611A1 (en) * | 2011-11-29 | 2013-06-06 | 华为技术有限公司 | Data processing method, device and client in distributed storage system |
| CN102799659B (en) * | 2012-07-05 | 2015-01-21 | 广州鼎鼎信息科技有限公司 | Overall repeating data deleting system and method based on non-centre distribution system |
| CN102799659A (en) * | 2012-07-05 | 2012-11-28 | 广州鼎鼎信息科技有限公司 | Overall repeating data deleting system and method based on non-centre distribution system |
| CN103747297A (en) * | 2014-01-09 | 2014-04-23 | 能力天空科技（北京）有限公司 | Video dynamic dispatching method, device and video playing system |
| CN109121436A (en) * | 2015-11-25 | 2019-01-01 | 蒂米菲尔股份有限公司 | Augmentation, exploration and Maintenance Significant Items method layered |
| CN109121436B (en) * | 2015-11-25 | 2022-06-21 | 蒂米菲尔股份有限公司 | Method for augmenting, exploring, and maintaining a hierarchy of projects |
| CN109862376A (en) * | 2019-02-28 | 2019-06-07 | 广州华多网络科技有限公司 | Live content jettison system, method, apparatus, listserv and storage medium |
| CN110932980A (en) * | 2019-11-27 | 2020-03-27 | 达闼科技成都有限公司 | Communication method, terminal, and computer-readable storage medium |
| CN110932980B (en) * | 2019-11-27 | 2022-09-20 | 达闼机器人股份有限公司 | Communication method, terminal, and computer-readable storage medium |
| CN111064641A (en) * | 2019-12-31 | 2020-04-24 | 上海焜耀网络科技有限公司 | Node performance detection system and method for decentralized storage network |
| CN111064641B (en) * | 2019-12-31 | 2021-07-02 | 上海焜耀网络科技有限公司 | Node performance detection system and method for decentralized storage network |
| CN114390338A (en) * | 2022-01-26 | 2022-04-22 | 湖南快乐阳光互动娱乐传媒有限公司 | Method and device for rapidly acquiring p2p video and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101378325A (en) | Superposition network and building method thereof | |
| US8279766B2 (en) | Interior-node-disjoint multi-tree topology formation | |
| RU2433461C2 (en) | Interaction between neighbourhoods within federation on rendezvous mechanism | |
| Xiang et al. | Stateless multicasting in mobile ad hoc networks | |
| WO2010127618A1 (en) | System and method for implementing streaming media content service | |
| JP2009508410A (en) | Parallel execution of peer-to-peer overlay communication using multi-destination routing | |
| Zulhasnine et al. | Towards an effective integration of cellular users to the structured peer-to-peer network | |
| Shi et al. | Tinymq: A content-based publish/subscribe middleware for wireless sensor networks | |
| CN1996931A (en) | Network multicast method | |
| Al-Adrousy et al. | A recommender system for team formation in MANET | |
| Baldoni et al. | Distributed event routing in publish/subscribe systems | |
| Baldoni et al. | A self-organizing crash-resilient topology management system for content-based publish/subscribe | |
| KR100921216B1 (en) | On the construction of multicast tree for multi-source service in wireless mesh networks | |
| Mundur et al. | Routing in intermittent networks using storage domains | |
| Son et al. | In-network processing for wireless sensor networks with multiple sinks and sources | |
| KR100960516B1 (en) | Routing recovery method for Ad-hoc Network | |
| Zheng et al. | Semanticcast: Content-based data distribution over self-organizing semantic overlay networks | |
| Lin et al. | An efficient anycast scheme for discovering K services in mobile ad-hoc networks | |
| Zheng et al. | Research on State Multicast Routing Algorithm Based on Distributed Sensing Network | |
| Naik et al. | An Introduction to Gossip Protocol Based Learning in Peer-to-Peer Federated Learning | |
| Ktari et al. | Symmetric routing in DHT overlays | |
| Tian et al. | Caching algorithms for broadcasting and multicasting in disruption tolerant networks | |
| Lee et al. | An efficient service propagation scheme for large-scale MANETs | |
| Xiaochuan et al. | A hypercubic event-dissemination overlay using structure-aware addressing for distributed xml-based pub/sub system | |
| Ktari et al. | Exploiting power-law node degree distribution in chord overlays |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20090304 |