(19)
(11)EP 2 117 166 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
09.01.2019 Bulletin 2019/02

(21)Application number: 08856338.2

(22)Date of filing:  27.10.2008
(51)International Patent Classification (IPC): 
H04L 12/24(2006.01)
H04L 12/751(2013.01)
H04L 12/703(2013.01)
H04L 29/08(2006.01)
(86)International application number:
PCT/CN2008/072836
(87)International publication number:
WO 2009/071008 (11.06.2009 Gazette  2009/24)

(54)

METHOD AND SYSTEM FOR UPDATING ROUTING TABLES AFTER NODE FAILURE IN PEER-TO-PEER NETWORK

VERFAHREN UND SYSTEM ZUM AKTUALISIEREN VON ROUTING-TABELLEN NACH EINEM KNOTENAUSFALL IN EINEM PEER-TO-PEER NETZ

PROCÉDÉ ET SYSTÈME DE MISE À JOUR DE TABLES DE ROUTAGE APRÈS UNE DÉFAILLANCE DANS UN RÉSEAU PEER-TO-PEER


(84)Designated Contracting States:
AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MT NL NO PL PT RO SE SI SK TR

(30)Priority: 22.11.2007 CN 200710188100

(43)Date of publication of application:
11.11.2009 Bulletin 2009/46

(73)Proprietor: Huawei Technologies Co., Ltd.
Longgang District Shenzhen, Guangdong 518129 (CN)

(72)Inventors:
  • SHI, Guangyu
    Guangdong 518129 (CN)
  • CHEN, Jian
    Guangdong 518129 (CN)
  • GONG, Hao
    Guangdong 518129 (CN)

(74)Representative: Isarpatent 
Patent- und Rechtsanwälte Behnisch Barth Charles Hassa Peckmann & Partner mbB Postfach 44 01 51
80750 München
80750 München (DE)


(56)References cited: : 
EP-A1- 1 802 070
CN-C- 1 283 076
US-A1- 2005 063 318
CN-A- 1 681 257
US-A1- 2005 063 318
  
  • GISCARD WEPIWE: "HiPeer: An Evolutionary Approach to P2P Systems" INTERNET CITATION, [Online] 1 January 2006 (2006-01-01), pages 1-196, XP007911547 Retrieved from the Internet: URL:http://deposit.ddb.de/cgi-bin/dokserv? idn=980482771&dok_var=d1&dok_ext=pdf&filen ame=980482771.pdf> [retrieved on 2010-02-05]
  • ZHUANG S Q ET AL: "On failure detection algorithms in overlay networks" INFOCOM 2005. 24TH ANNUAL JOINT CONFERENCE OF THE IEEE COMPUTER AND CO MMUNICATIONS SOCIETIES. PROCEEDINGS IEEE MIAMI, FL, USA 13-17 MARCH 2005, PISCATAWAY, NJ, USA,IEEE, PISCATAWAY, NJ, USA, vol. 3, 13 March 2005 (2005-03-13), pages 2112-2123, XP010829335 ISBN: 978-0-7803-8968-7
  • HUNG-CHANG HSIAO ET AL: "Mobility Churn in DHTs" DISTRIBUTED COMPUTING SYSTEMS WORKSHOPS, 2005. 25TH IEEE INTERNATIONAL CONFERENCE ON COLUMBUS, OH, USA 06-10 JUNE 2005, PISCATAWAY, NJ, USA,IEEE, 6 June 2005 (2005-06-06), pages 799-805, XP010808144 ISBN: 978-0-7695-2328-6
  
Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


Description

FIELD OF THE INVENTION



[0001] The present invention relates to communication technologies, and in particular, to a method, a device, and a system for updating routes after a node fails in a Peer-to-Peer (P2P) network.

BACKGROUND



[0002] Unlike the traditional client/server system, a P2P system is characterized by operations between peers. Each peer in the P2P system can serve other peers and be served by other peers. Depending on the topology, P2P systems are sorted into these types: centralized topology, decentralized unstructured topology, decentralized structured topology (also known as DHT network), and hybrid topology.

[0003] The current prevalent topology of P2P systems is a Distributed Hash Table (DHT) network. A majority of applications in the P2P system are based on the DHT network. In such systems, a unique identifier (node ID) of a node is obtained through Hash operation according to unique attributes such as an IP address of the node. The data item corresponding to the node ID is expressed by a key pair <key, value>, where a key is an index to a data item, and a value may a locating address (for example, IP or URL) of the data item. Through Hash operation, a unique identifier is assigned to a data index key, and the key value pair corresponding to this key is stored into the node nearest to this key identifier. The unique identifier may be obtained through Hash operation after the key value pair is found. Through the unique identifier, the nearest node is found (the nearest node stores the address of the data item).

[0004] Meanwhile, the P2P system is a self-organized network. In this network, a node can join or quit the network randomly. The randomness leads to inaccuracy of locating the resources and churning of the network. The extent of network churn directly affects the efficiency of the route discovery method. Network churn is caused by joining, quitting, failure, or migration of a node; or by a concurrent joining process or network splitting. The way of handling different network churns in the DHT route searching method of the P2P network directly affects the route efficiency and the load overhead of the whole P2P network.

[0005] Two solutions to P2P network churn are provided in the prior art.

Prior art 1



[0006] In the process of handling network churn in the DHT network, three major elements are considered: fast replacement of the node, timeout detection, and the nearest neighboring node. This technology resists network churn directly in this way: K sequential nodes in the P2P node ID space ping each other to keep connection between nodes.

[0007] In the operation process of the P2P network, each node sends a Ping maintenance message to K sequential nodes around at intervals. The K sequential nodes receive the Ping message from the source node, and then feed back a message to the source node, indicating that the node is still alive and the routing table of the source node needs no change. When a node in the network fails, the neighboring nodes around can discover the node churn through proactive ping, find the failed node, and broadcast the failure information to the K sequential nodes, which therefore update their routing table links.

[0008] In the process of developing the present invention, the inventor finds at least the following defects in the prior art 1.

[0009] In the prior art 1, the K sequential nodes in the P2P node ID space ping each other to keep connection between the nodes, namely, the maintenance information is transferred only between K neighboring nodes. When a node fails in a network, the routing table information keeps continuity of only K nodes. In the attempt to search for a node in the network, "crawling" may occur. That is, the routing table information of the node which is originally directed to the failed node has failed, and the remote node finds the substitute node of the failed node only through a second recursive search. If the two nodes are farther away from each other, the problem is more serious, and the recursive search occurs more frequently between the nodes. When the network churns drastically, if multiple key routing nodes fail in the P2P path, the routing table information crawls more slowly in the network, and node has to search for the next step of the route at an interval of K hops.

[0010] In the P2P network, the churn and the crawl accrue. With increase of the failed nodes in the network, the failure information is not broadcast effectively, and most nodes keep the original failure routing table. Consequently, the crawl occurs more frequently and lasts for longer periods. At worst, the time complexity of searching for a node reaches O(N) times (N is the number of nodes in the P2P Overlay), which affects the efficiency of route searching in the P2P network seriously.

Prior art 2



[0011] The current survival probability of a node is predicted according to the history lifecycle information of the nodes throughout the network. Specifically, the history lifecycle distribution probability of nodes throughout the P2P network is collected, the survival probability of a node in the next period is deduced, and prediction information about the actions of the nodes throughout the network is generated for deciding the next-step action of the network. When a node is predicted to be stable in the next period, its neighboring node reduces the Ping maintenance operation for this node, thus reducing the overhead of the maintenance bandwidth.

[0012] In the process of developing the present invention, the inventor finds at least the following defects in the prior art 2.

[0013] The more sophisticated the P2P network is, the less accurate the prediction of the node lifecycle will be. In the prediction process, many system parameters need to be set, and the selection of the parameter values is specific to each situation. Different parameter settings may lead to drastic inaccuracy of the prediction.

[0014] Besides, the prior art 2 reduces the system maintenance overhead, but makes no improvement for the efficiency of the route searching method of the P2P network. When the network churns seriously, the complexity of searching for a node is still up to O(N) times, and the crawl still exists in the process of searching for the route, thus leading to low efficiency of searching for the route. A document "Hipeer: An Evolutionary Approach to P2P Systems" INTERNET CITATION, 1 January 2006 (2006-01-01), pages 1-196, XP007911547 discloses technical schemes to solve the problem of resource management in P2P systems. An evolutionary concept that enables self-management of nodes and resources in P2P networks is proposed. Routing operation can be performed in three different directions: on_ring_forwarding, outward_forwarding and inward_forwarding.

SUMMARY



[0015] The invention provides a method according to claim 1 and a network according to claim 4. Further advantageous embodiments and improvements are listed in the dependent claims. Hereinafter, before coming to the detailed description of the embodiments of the invention with reference to the drawings, some aspects associated with the invention are discussed below and they will enable a better understanding of what follows thereafter.

[0016] Compared with the prior art which notifies the failure information to only K neighboring nodes in the case of node failure, the embodiments of the present invention enhance the capability of the whole P2P network in perceiving the node churn and improve the route search efficiency and system stability of the whole P2P network.

BRIEF DESCRIPTION OF THE DRAWINGS



[0017] 

FIG. 1 is a flowchart of updating routing table information after a node fails in a P2P network according to an embodiment of the present invention;

FIG. 2 shows an instance of updating routing table information after a node fails in a P2P network according to an embodiment of the present invention;

FIG. 3 shows an instance of updating routing table information after a node fails in P2P network architecture according to an embodiment of the present invention;

FIG. 4 shows a schematic view of node failure in a Kademlia DHT network according to an embodiment of the present invention;

FIG. 5 shows a schematic flow of node failure in a Kademlia DHT network according to an embodiment of the present invention;

FIG. 6 shows a schematic structural view of a Chord network according to an embodiment of the present invention;

FIG. 7A, FIG. 7B, FIG. 7C, FIG. 8A and FIG. 8B show a schematic structural view of a network device according to an embodiment of the present invention; and

FIG. 9 shows a schematic structural view of a communication network according to an embodiment of the present invention.


DETAILED DESCRIPTION



[0018] The embodiments of the present invention are detailed below by reference to accompanying drawings.

[0019] As shown in FIG. 1, the process of updating routing table information after a node fails in a P2P network is as follows:

Step 11: The range of the nodes whose route is directed to the failed node is determined according to the distance between the failed node and the neighboring node of the failed node.

Step 12: The failure information of the failed node is sent to the LDN of the failed node within the node range. There are i LDNs (i is a natural number greater than 0). If the node ID of node B differs from the node ID of node A only in the i bit (i is an ordinal number) and all other bits are the same or nearest to each other, then node B is LDN[i] of node A.

Step 13: The LDN that receives the failure information updates the routing table according to the failure information, and directs the routing table information, which is originally directed to the failed node, to the substitute node of the failed node.



[0020] In the flow shown in FIG. 1, before the range of the nodes whose route is directed to the failed node is determined, the state of the neighboring node in the network may be probed periodically to determine the failed node. Depending on the set period of probing, the failed node may be a node that fails abruptly in the current network, or a node that fails in the network a period ago. In different networks, the range of the nodes whose route is directed to the failed node (which is determined according to the distance between the failed node and the neighboring node of the failed node) differs. In an embodiment, the range of the nodes whose route is directed to the failed node may be determined according to the distance between the failed node and the neighboring node of the failed node and the routing features of the network.

[0021] In step 12, according to the key value feature stored in the network, the substitute node is selected among the neighboring nodes of the failed node to send failure information. The substitute node sends the failure information of the failed node to the LDN of the failed node within the range of the nodes whose route is directed to the failed node.

[0022] The failure information includes the address of the substitute node, and optionally, may further includes a least one of an identifier of the failed node, an address of the failed node, and a key value of the failed node. In an embodiment, the failure information further includes the impact range within which the LDN is impacted by the failed node. After receiving the failure information, the LDN may forward the failure information to the nodes within the impact range.

[0023] When the LDN determines that the impact range does not exceed the maintenance range of the neighboring nodes of the local node, the LDN may forward the failure information to the nodes within the impact range directly. Alternatively, when the LDN determines that the impact range exceeds the maintenance range of the neighboring nodes of the local node, preferably, the LDN forwards the failure information to the neighboring node nearest to the impact range, and the neighboring node further forwards the failure information to other nodes in the impact range. Nevertheless, in the case, the LDN may forward the failure information to other nodes, and the node which receives the failure information further forwards the failure information to other nodes in the impact range. After receiving the failure information, the node in the impact range updates the routing table, namely, directs the routing table information, which is originally directed to the failed node, to the substitute node of the failed node.

[0024] FIG. 2 is a flowchart of updating routing table information after a node fails in a P2P network in an embodiment of the present invention. As regards the implementation of this flow in the P2P network, the process of updating routing table information shown in FIG. 3 serves as a reference. The updating process is as follows:

Step 21: A node probes the survival state of the neighboring node proactively. For example, node B probes the neighboring node at intervals, and obtains the survival state of the neighboring node F.

Step 22: When a node (node F) fails, the neighboring node (node B) discovers failure of node F through step 21, and calculates the range of notifying failure information. The neighboring node (node B) calculates the range of the nodes whose route is directed to the failed node throughout the network based on different DHT network routing method features according to the distance between the node ID of the neighboring node (node B) and the node ID of the failed node (node F).

Step 23: The neighboring node (node B) determines a substitute node (node A) for sending the failure information. Step 23 may occur before, during, or after step 22.



[0025] After calculating the impact range of the failed node, the neighboring node may select a neighboring node as the substitute node for sending the node failure information according to the key value features stored in different DHT networks. In the Kademlia network or Pastry network, the stored key value is selected based on the distance of the node ID of the node, the substitute node is the neighboring node at the minimum XOR distance away from the failed node. In the Chord network or Koorde network, the stored key value is selected based on the nearest successor node, the substitute node is the node subsequent to the failed node.

[0026] Step 24: The substitute node (node A) searches for the LDN in the impact range of the failure. The neighboring node that serves as a substitute node finds the LDN[i] node corresponding to the failed node in each range through the route search method. The LDN[i] node may be the node nearest to the failed node in each range.

[0027] Step 25: The substitute node (node A) sends the failure information to the LDNs of the failed node consecutively. After the LDN(i) of each impact range of the failed node is found, the substitute node sends the node failure information to the LDNs consecutively. Failure information includes the address of the substitute node, and optionally, may further include at least one of the identifier of the failed node, the address of the failed node and the key value of the failed node.

[0028] The neighboring node can also notify the impact range of the failed node to each LDN. For example, the failure information includes the range of impact caused by the failed node onto the LDNs, and notifies the LDNs to update the corresponding routing table information and forward the failure information among the neighbors of the LDNs according to the impact range of the failure information.

[0029] Step 26: After receiving the failure information, the LDN updates the routing table, and directs the routing table information, which is originally directed to the failed node, to the substitute node (node A). Further, the LDN node forwards the failure information to its neighboring node according to the received impact range of the failed node.

[0030] Each LDN node calculates how many neighboring nodes around need to know the node failure information according to the received impact range of the failed node, and forwards the information to such nodes. If the range of nodes which need to receive the information exceeds the range of neighboring nodes, the LDN notifies the neighbor nearest to the range to further notify (for example, multicast) the information to other nodes within the range. Through multiple hops, all the nodes in the range are notified. Nevertheless, in the case, the LDN may also forward the failure information to other nodes, and the node that receives the failure information further forwards the failure information to other nodes in the impact range.

[0031] The nodes in the impact range update their routing tables. After receiving the failure information, the node in the impact range directs the pointer, which is originally directed to the failed node, in the routing table to the substitute node of the failed node, thus updating the routing table.

[0032] The embodiments of the present invention are detailed below, taking different P2P networks as examples.

I. Taking a Kademlia DHT network as an example



[0033] Supposing that node 10110 in the Kademlia network shown in FIG. 4 fails, according to the method under the present invention, its neighboring nodes 1010 and 10111 discovers failure of the node 10110 through proactive probe, and can calculate the range of impact of the failure information according to the distance between the node ID of the neighboring node and the node ID of the failed node, when the node 10110 in the network fails. According to the features of the Kademlia network, when node F fails, all the nodes whose routing table pointer has been directed to node F are covered in the following range:

where:

node F fails and churns, node X is the predecessor node of node F, node Y is the successor node of node F, and node N needs to be notified of the failure information; and

m is the number of bits of the node ID, the m in the Kademlia network is 160, and i ranges from 1 to m.



[0034] By reference to the foregoing formula, the range of impacting node N needs to fulfill the following:

The XOR distance between node N and the failed node F is greater than 2m-I and less than or equal to 2m-i+1;

the XOR distance between node N and node F is less than the XOR distance between node N and node X; and

the XOR distance between node N and node F is less than the XOR distance between node N and node Y. If node N fulfills the three expressions, N is the node that needs to be notified of the failure information.



[0035] According to the difference between their own node ID and the node ID of the failed node 10110, node 1010 and node 10111 determines the substitute node of the failed node. Because the node ID difference between node 10111 and the failed node 10110 is less than that between node 1010 and the failed node 10110, the neighboring node 10111 is determined as the substitute node. The neighboring node 10111 finds a LDN node in each impact range determined through the foregoing formula, and notifies the failure information to all the LDN nodes. As shown in FIG. 5, in one embodiment, according to the symmetry of the Kademlia routing table, the LDN[1] of the node 10110 is 00110 because all the bits of the node 00110 are the same as those of the node 10110 except the initial bit. According to the composition of the Kademlia K bucket routing table, the node 00110 is in the K bucket routing table of the node 10110, and vice versa. Besides, according to the foregoing formula, the LDNs of the node 10110 further include: LDN[1]=00110, LDN[2]=111, LDN[3]=10010, LDN[4]=1010, and LDN[5]=10111.

[0036] After the range of the nodes to be notified is worked out and the LDN nodes in the range are found, the LDN node searches for the node N in the respective range and transmits the churn information to such node Ns, and the node Ns update their routing tables.

[0037] Through proactive notification, when the remote node needs to search for the failed node, the remote node can know the information state of the node and the neighboring substitute node directly, thus avoiding the trouble of crawling slowly to search for the failed node like the practice in a traditional P2P network.

II. Taking a Pastry DHT network as an example



[0038] The routing table structure of a Pastry network is similar to that in the Kademlia method, but differs in: The Pastry network changes the binary node ID to a Multi-ary node ID on the basis of the Kademlia network. In the implementation, the node ID may be changed from multi-ary to binary first, and then apply the Kademlia method on the Pastry network. The detailed process is as follows:

[0039] When node F churns, its nearest neighboring node X detects the churn. Afterward, the node X may work together with another neighbor of the failed node to find the node LDN[1] according to the range determined by the following formula, where the initial bit of LDN[1] is different from that of node F and the XOR difference between the remaining bits of LDN[1] and those of node F is the minimum, and notifies the churn information to the LDN[1]. The notification indicates that node F has failed, and all the routes subsequently directed to node F are directed to node X instead, and indicates that the churn information is forwarded among the nodes within the range. By analogy, the neighboring node X notifies other LDN[i] determined in this way: The i bit of LDN[i] is different from that of node F and the XOR difference between the remaining bits of LDN[i] and those of node F is the minimum. The neighboring node X tells the LDN[i] that node F has failed, and instructs the LDN[i] to notify other nodes in their respective ranges to update the routing table.

where:

k is a k-ary radix, and i ranges from 1 to m; and

X is the predecessor neighbor of the failed node F, Y is the successor neighbor of F, and N is the node that needs to be notified.



[0040] By reference to the foregoing formula, the range of impacting node N needs to fulfill the following:

The XOR distance between node N and the failed node F is greater than 2m-I and less than or equal to 2m-i+1;

the XOR distance between node N and node F is less than the XOR distance between node N and node X; and

the XOR distance between node N and node F is less than the XOR distance between node N and node Y. If node N fulfills the three expressions, N is the node that needs to be notified of the failure information.



[0041] In the implementation, Pastry differs from Kademlia in the following:

The Kademlia employs the binary node ID, and therefore, the Kademlia uses XOR operation to calculate the difference between two nodes; and

the Pastry employs the multi-ary node ID, and needs to convert the multi-ary node ID in the Pastry network to the binary node ID first.


III. Taking a Chord DHT example as an example



[0042] As shown in FIG. 6, in a Chord network, n nodes is Hashed to the token ring with the "logn" bits. Each node x is directed to its next successor node (the nearest node in the clockwise direction of the ring). It maintains a pointer table composed of m=logn entries. The i entry (i is an ordinal number) stores the subsequent ID of the x+2i-1.

[0043] As shown in FIG. 5 and FIG. 6, when node F churns, according to the method of one embodiment, the neighboring node of F discovers the churn of F through proactive probe, and calculates the range of all the nodes (node N) whose routing table pointer has been directed to node F in the whole Chord ring according to the difference between the node ID of the neighboring node and the node ID of the failed node F.

where predecessor(F) is the predecessor node of node F. According to the foregoing formula, the node ID of node N is added to 2i to obtain a sum, and the sum modulo 2m is greater than the node ID of the predecessor node of the failed node F and is less than or equal to the node ID of F. The node Ns in the range defined by the foregoing inequality are the nodes that need to be notified.

[0044] After the impact range of the churn is obtained, the substitute node of the failed node is determined as the successor node according to the features of the Chord network. The substitute node may find the LDN nodes in each range according to the calculated range. For example, LDN[m-1] of node N8 ought to be a node whose routing table contains a last entry directed to N8, that is, the node defined by: The successor corresponding to the (node ID +2m-1) of the node is directed to N8. LDN[i] are the nodes defined by: The successor corresponding to the (node ID +2i) of the nodes is directed to N8, where i ranges from 0 to m-1. The substitute node of F notifies all the LDNs in each impact range in the Chord ring, and tells the node failure information and the churn range. The LDNs multicast the information within their ranges, and send a notification to the nodes whose routing table pointer has been directed to node F. The notification indicates that node F has failed and the substitute node is the successor node of F. The nodes in the impact range update their own routing table according to this notification.

IV. Taking a Koorde DHT as an example



[0045] The Koorde DHT introduces the conception of the de Bruijn diagram on the basis of the Chord, with a view to improving the Chord route.

[0046] In a Koorde network, when node F fails, its predecessor node X and its successor node Y discover the failure information through probing. According to the features of the Koorde network architecture, the key value of a node in the Koorde network is stored in the nearest successor node. Therefore, when node F fails, the substitute node of the failed node is specified to be the successor node Y, and the range of the nodes (node N) which need to know the failure information is:



[0047] By reference to the foregoing equation, when node N fulfills:

The predecessor node of the failed node 0 is X and the successor node is Y;

node F is outside N and the successor node range of N (the range includes the successor node of N (successor (N)) but does not include N); and

the de Bruijn virtual node i of F after being shifted through topBit(kshift) is also outside N and the successor node range of N (the range includes the successor node of N (successor (N)) but does not include N). If node N fulfills the three expressions, N is the node that needs to be notified of the failure information.



[0048] Node Y multicasts the failure information to the LDN nodes within the range, and instructs the LDN nodes to forward the failure information of the node within the range.

[0049] To sum up, the technical solution under the present invention makes use of the symmetric mechanism implemented on the existing DHT routing table, and calculates the range of impact of the churn information quickly according to the distance between node ID of the churn node and the node ID of the neighboring nodes after a node churns in the network. The neighboring node notifies the churn information to the remote node that desires the information within the range proactively, and maintains the hop of route between the original two nodes, thus preventing the remote node from crawling to search for the churn node. When a node fails, as against notifying only K neighboring nodes, notifying the remote node in the routing table makes the network more perceptive to the node churn. The notification information magnitude is O(logN) or O(1), depending on the size of the routing table of each node in the DHT. In this way, the efficiency of the whole P2P network in searching for routes in a churn environment is improved, the system maintenance overhead is cut back, and the system stability is improved.

[0050] Based on the same conception, a network device is disclosed in an embodiment of the present invention. As shown in FIG. 7A, the network device includes: a first determining module 71, adapted to determine the range of the nodes whose route is directed to the failed node according to the distance between the failed node and the neighboring node of the failed node; and a first sending module 72, adapted to send the failure information of the failed node to the LDN of the failed node within the node range.

[0051] As shown in FIG. 7B, the network device shown in FIG. 7A in an embodiment may further include a probing module 73, adapted to probe the state of the neighboring node in the network at set intervals and determine the failed node.

[0052] In an embodiment, the first determining module 71 may be further adapted to determine the range of the nodes whose route is directed to the failed node according to the distance between the failed node and the neighboring node of the failed node as well as the network routing features.

[0053] As shown in FIG. 7C, the network device shown in FIG. 7A in an embodiment may further include a second determining module 74, adapted to determine the substitute node for sending the failure information among the neighboring nodes of the failed node according to the key value features stored in the network.

[0054] If the network is a Kademlia DHT network or a Pastry DHT network, the second determining module 74 may be further adapted to determine the substitute node to be the neighboring node at the minimum XOR distance away from the failed node. Optionally, if the network is a Chord DHT network or a Koorde DHT network, the second determining module 74 may be further adapted to determine the substitute node to be the successor node of failed node.

[0055] The failure information includes the address of the substitute node, and may further include at least one of the identifier of the failed node, the address of the failed node, and the key value of the failed node. The failure information may further include the range of impact caused by the failed node onto the LDN.

[0056] Based on the same conception, a network device is disclosed in an embodiment of the present invention. The network device is the LDN of the failed node. As shown in FIG. 8A, the network device includes:

a receiving module 81, adapted to receive the failure information of the failed node, where the failure information includes the address of the substitute node of the failed node; and

a processing module 82, adapted to update the routing table according to the failure information.



[0057] In an embodiment, the failure information may further include at least one of an identifier of the failed node, an address of the failed node, and a key value of the failed node.

[0058] The failure information may further include the range of impact caused by the failed node onto the local node. In this case, as shown in FIG. 8B, the network device shown in FIG. 8A may further include a second sending module 83, adapted to forward the failure information to the nodes within the impact range after the receiving module 81 receives the failure information.

[0059] In an embodiment, the failure information includes the range of impact caused by the failed node onto the local node. In this case, the second sending module 83 may be further adapted to forward the failure information to the nodes within the impact range directly after the receiving module 81 receives the failure information if the impact range does not exceed the maintenance range of the neighboring nodes of the local node. Alternatively, when the impact range exceeds the maintenance range of the neighboring nodes of the local node, the second sending module 83 may be further adapted to forward the failure information to the neighboring node nearest to the impact range, and instruct the neighboring node to forward the failure information to other nodes in the impact range. The determining of the impact range here is similar to the determining of the impact range in a different P2P network mentioned above (for example, Kademlia network, Pastry network, Chord network, Koorde network).

[0060] Based on the same conception, a P2P communication network is disclosed in an embodiment of the present invention. As shown in FIG. 9, the P2P communication network includes: a failed node 91, a neighboring node 92 of the failed node, and an LDN 93. The neighboring node 92 of the failed node is adapted to determine the range of the nodes whose route is directed to the failed node according to the distance between the failed node and the neighboring node of the failed node, and send the failure information of the failed node to the LDN of the failed node within the node range. The LDN 93 is adapted to receive the failure information, and update the routing table according to the failure information.

[0061] It is understandable to those skilled in the art that all or part of the steps in the foregoing embodiments may be performed through hardware instructed by a program. The program may be stored in a computer-readable storage medium such as the ROM/RAM, magnetic disk, and compact disk.

[0062] In the embodiments of the present invention, the range of the nodes whose route is directed to the failed node is determined according to the distance between the failed node and the neighboring node of the failed node; the failure information of the failed node is sent to the LDN of the failed node within the node range; and the LDN updates the routing table according to the failure information, and notifies the neighboring nodes within the node range to update the their respective routing table. Compared with the prior art which notifies the failure information to only K neighboring nodes in the case of node failure, the embodiments of the present invention enhance the capability of whole the P2P network in perceiving node churn, and improve the route search efficiency and system stability of the whole P2P network.

[0063] It is apparent that those skilled in the art can make various modifications and variations to the present invention without departing from the scope of the present invention. The present invention is intended to cover such modifications and variations provided that they fall in the scope of protection defined by the following claims or their equivalents.


Claims

1. A method for updating routing table information after a node fails in a Peer-to-Peer, P2P, network which is a Distributed Hash Table, DHT, network, comprising the steps of:

detecting, by a neighboring node of a node, that the node fails by probing the node at periodic intervals;

determining, by the neighboring node, a substitute node for sending failure information of the failed node, wherein the substitute node has a minimum node ID distance from a node ID of the failed node;

the neighboring node determining (11) a plurality of ranges of nodes whose routes are directed to a failed node according to respective distances between the node ID of the failed node, the node ID of the neighboring node of the failed node, the node ID of the substitute node and the node IDs of the nodes included in each range;

sending, by the neighboring node, the failure information to the substitute node;

searching. by the substitute node, for a long distance neighbor, LDN, in each range, wherein the LDN is a node nearest to the failed node in each range;

sending (12), by the substitute node, the failure information of the failed node to the LDNs within each range of the nodes;

the LDNs updating their corresponding routing table information according to the failure information and transmitting the failure information to the nodes of their ranges to update their routing tables, wherein information updated in the routing tables directs to the substitute node.


 
2. The method according to claim 1, wherein the method further comprises the steps of:

determining (23) the substitute node among neighboring nodes of the failed node to send the failure information, according to key value features stored in the network; and

instructing (25) the substitute node to send the failure information to the LDN within each range of the nodes.


 
3. The method according to claim 1, wherein the failure information includes the range caused by the failed node onto the LDN, and the method further comprises the step of:
forwarding, by the LDN, the failure information to the nodes within the range, after receiving the failure information.
 
4. A P2P network which is a Distributed Hash Table, DHT, network, comprising a plurality of interconnected nodes, wherein
a neighboring node of a node is adapted to detect that the node fails by having a probing module (73) thereof probing the node at periodic intervals;
the neighboring node is adapted to determine a substitute node for sending failure information of the failed node, wherein the substitute node has a minimum node ID distance from a node ID of the failed node;
the neighboring node is further adapted to determine (11) a plurality of ranges of nodes whose routes are directed to a failed node according to respective distances between the node ID of the failed node, the node ID of the neighboring node of the failed node, the node ID of the substitute node and the node IDs of the nodes included in each range;
the neighboring node further adapted to send the failure information to the substitute node;
the substitute node further configured to search for a long distance neighbor, LDN, in each range, wherein the LDN is a node nearest to the failed node in each range;
the substitute node further adapted to send (12) the failure information of the failed node to
the LDNs within each range of the nodes; and
the LDNs adapted to update their corresponding routing table information according to the failure information and to transmit the failure information to the nodes of their ranges to update their routing tables, wherein information updated in the routing tables directs to the substitute node.
 
5. The network according to claim 4, further comprising:
the neighboring node further adapted to determine the substitute node for sending the failure information among neighboring nodes of the failed node according to key value features stored in the network.
 
6. The network according to claim 5, wherein
the neighboring node is further adapted to determine the substitute node that is the neighboring node at the minimum XOR distance away from the failed node, if the network is a Kademlia DHT network or a Pastry DHT network; or
the neighboring node is further adapted to determine the substitute node to be a nearest successor node of the failed node if the network is a Chord DHT network or a Koorde DHT network.
 


Ansprüche

1. Verfahren zum Aktualisieren von Routing-Tabellen-Informationen, nachdem ein Knoten in einem Peer-to-Peer-, P2P, -Netz ausfällt, welches ein verteiltes Hashtabellen-, DHT, -Netz ist, umfassend folgende Schritte:

Erkennen durch einen Nachbarknoten eines Knotens, dass der Knoten ausfällt, indem der Knoten in regelmäßigen Intervallen untersucht wird;

Bestimmen durch den Nachbarknoten eines Ersatzknotens zum Senden von Ausfallinformationen des ausgefallenen Knotens, wobei der Ersatzknoten eine minimale Knoten-ID-Distanz von einer Knoten-ID des ausgefallenen Knotens aufweist;

Bestimmen (11) durch den Nachbarknoten mehrerer Bereiche von Knoten, deren Routen auf einen ausgefallenen Knoten gerichtet sind, gemäß jeweiligen Distanzen zwischen der Knoten-ID des ausgefallenen Knotens, der Knoten-ID des Nachbarknotens des ausgefallenen Knotens, der Knoten-ID des Ersatzknotens und den Knoten-IDs der Knoten, die in jedem Bereich enthalten sind;

Senden durch den Nachbarknoten der Ausfallinformationen zu dem Ersatzknoten;

Suchen durch den Ersatzknoten nach einem "Long Distance Neighbor", LDN, in jedem Bereich, wobei der LDN ein Knoten ist, der am nächsten bei dem ausgefallenen Knoten in jedem Bereich liegt;

Senden (12) durch den Ersatzknoten der Ausfallinformationen des ausgefallenen Knotens zu den LDNs innerhalb jedes Bereichs der Knoten;

Aktualisieren der LDNs ihrer entsprechenden Routing-Tabellen-Informationen gemäß den Ausfallinformationen und Übertragen die Ausfallinformationen zu den Knoten ihrer Bereiche, um deren Routing-Tabellen zu aktualisieren, wobei Informationen, die in den Routing-Tabellen aktualisiert werden, auf den Ersatzknoten gerichtet sind.


 
2. Verfahren nach Anspruch 1, wobei das Verfahren ferner folgende Schritte umfasst:

Bestimmen (23) des Ersatzknotens unter Nachbarknoten des ausgefallenen Knotens, um die Ausfallinformationen zu senden, gemäß Schlüsselwertmerkmalen, die in dem Netz gespeichert sind; und

Anweisen (25) des Ersatzknotens, die Ausfallinformationen zu dem LDN innerhalb jedes Bereichs der Knoten zu senden.


 
3. Verfahren nach Anspruch 1, wobei die Ausfallinformationen den Bereich aufweisen, der von dem ausgefallenen Knoten auf dem LDN hervorgerufen wird, und das Verfahren ferner folgenden Schritt umfasst:
Weiterleiten durch den LDN der Ausfallinformationen zu den Knoten innerhalb des Bereichs nach dem Erhalten der Ausfallinformationen.
 
4. P2P-Netz, welches ein verteiltes Hashtabellen-, DHT, -Netz ist, das mehrere verbundene Knoten aufweist, wobei

ein Nachbarknoten eines Knotens eingerichtet ist, um zu erkennen, dass der Knoten ausfällt, indem er ein Untersuchungsmodul (73) davon den Knoten in regelmäßigen Intervallen untersuchen lässt;

der Nachbarknoten eingerichtet ist, um einen Ersatzknoten zum Senden von Ausfallinformationen des ausgefallenen Knotens zu bestimmen, wobei der Ersatzknoten eine minimale Knoten-ID-Distanz von einer Knoten-ID des ausgefallenen Knotens aufweist;

der Nachbarknoten ferner eingerichtet ist, um mehrere Bereiche von Knoten, deren Routen auf einen ausgefallenen Knoten gerichtet sind, gemäß Distanzen zwischen der Knoten-ID des ausgefallenen Knotens, der Knoten-ID des Nachbarknotens des ausgefallenen Knotens, der Knoten-ID des Ersatzknotens und den Knoten-IDs der Knoten, die in jedem Bereich enthalten sind, zu bestimmen (11);

der Nachbarknoten ferner eingerichtet ist, um die Ausfallinformationen zu dem Ersatzknoten zu senden;

der Ersatzknoten ferner konfiguriert ist, um nach einem Nachbar mit großer Distanz, LDN, in jedem Bereich zu suchen, wobei der LDN ein Knoten ist, der am nächsten bei dem ausgefallenen Knoten in jedem Bereich liegt;

der Ersatzknoten ferner eingerichtet ist, um die Ausfallinformationen des ausgefallenen Knotens zu den LDNs innerhalb jedes Bereichs der Knoten zu senden (12); und

die LDNs eingerichtet sind, um ihre entsprechenden Routing-Tabellen-Informationen gemäß den Ausfallinformationen zu aktualisieren und die Ausfallinformationen zu den Knoten ihrer Bereiche zu übertragen, um ihre Routing-Tabellen zu aktualisieren, wobei Informationen, die in den Routing-Tabellen aktualisiert sind, auf den Ersatzknoten gerichtet sind.


 
5. Netz nach Anspruch 4, das ferner Folgendes aufweist:
der Nachbarknoten ist ferner eingerichtet, um den Ersatzknoten zum Senden der Ausfallinformationen unter Nachbarknoten des ausgefallenen Knotens gemäß Schlüsselwertmerkmalen, die in dem Netz gespeichert sind, zu bestimmen.
 
6. Netz nach Anspruch 5, wobei
der Nachbarknoten ferner eingerichtet ist, um den Ersatzknoten zu bestimmen, der
der Nachbarknoten mit der minimalen XOR-Distanz von dem ausgefallenen Knoten weg ist, wenn das Netz ein Kademlia-DHT-Netz oder ein Pastry-DHT-Netz ist; oder der Nachbarknoten ferner eingerichtet ist, um den Ersatzknoten als einen nächsten Nachfolgerknoten des ausgefallenen Knotens zu bestimmen, wenn das Netz ein Chord-DHT-Netz oder ein Koorde-DHT-Netz ist.
 


Revendications

1. Procédé de mise à jour d'informations de table de routage après une défaillance de noeud dans un réseau pair à pair, P2P, qui est un réseau à table de hachage distribuée, DHT, comprenant les étapes consistant à :

détecter, par un noeud voisin d'un noeud, une défaillance du noeud par sondage du noeud à des intervalles périodiques ;

déterminer, par le noeud voisin, un noeud de substitution pour envoyer des informations de défaillance du noeud défaillant, le noeud de substitution ayant une distance d'ID de noeud minimale à partir d'un ID de noeud du noeud défaillant :

le noeud voisin déterminant (11) une pluralité de plages de noeuds dont les itinéraires sont dirigés vers un noeud défaillant selon des distances respectives entre l'ID de noeud du noeud défaillant, l'ID de noeud du noeud voisin du noeud défaillant, l'ID de noeud du noeud de substitution et les ID de noeud des noeuds inclus dans chaque plage ;

envoyer, par le noeud voisin, les informations de défaillance au noeud de substitution ; rechercher, par le noeud de substitution, pour un voisin de longue distance, LDN, dans chaque plage, le LDN étant un noeud le plus proche du noeud défaillant dans chaque plage ;

envoyer (12), par le noeud de substitution, les informations de défaillance du noeud défaillant aux LDN à l'intérieur de chaque plage des noeuds ;

les LDN mettant à jour leurs informations de table de routage correspondantes selon les informations de défaillance, et transmettant les informations de défaillance aux noeuds de leurs plages pour mettre à jour leurs tables de routage, les informations mises à jour dans les tables de routage dirigeant vers le noeud de substitution.


 
2. Procédé selon la revendication 1, le procédé comprenant en outre les étapes consistant à :

déterminer (23) le noeud de substitution parmi des noeuds voisins du noeud défaillant pour envoyer les informations de défaillance, en fonction des caractéristiques de valeur clé stockées dans le réseau ; et

ordonner (25) que le noeud de substitution envoie les informations de défaillance au LDN à l'intérieur de chaque plage des noeuds.


 
3. Procédé selon la revendication 1, les informations de défaillance comprenant la plage provoquée par le noeud défaillant sur le LDN, et le procédé comprenant en outre l'étape consistant à :
transférer, par le LDN, des informations de défaillance aux noeuds dans la plage, après la réception des informations de défaillance.
 
4. Réseau P2P qui est un réseau à table de hachage distribuée, DHT, comprenant une pluralité de noeuds interconnectés,
un noeud voisin d'un noeud étant adapté pour détecter une défaillance du noeud en ayant un module de sondage (73) de celui-ci, sondant le noeud à des intervalles périodiques :
le noeud voisin étant adapté pour déterminer un noeud de substitution pour envoyer des informations de défaillance du noeud défaillant, le noeud de substitution ayant une distance d'ID de noeud minimale à partir d'un ID de noeud du noeud défaillant :

le noeud voisin étant en outre adapté pour déterminer (11) une pluralité de plages de noeuds dont les itinéraires sont dirigés vers un noeud défaillant en fonction des distances respectives entre l'ID de noeud du noeud défaillant, l'ID de noeud du noeud de substitution du noeud défaillant, l'ID de noeud du noeud de substitution et les ID de noeud des noeuds inclus dans chaque plage ;

le noeud voisin étant en outre adapté pour envoyer les informations de défaillance au noeud de substitution ;

le noeud de substitution étant en outre configuré pour rechercher un voisin de longue distance LDN, dans chaque plage, le LDN étant un noeud le plus proche du noeud défaillant dans chaque plage ;

le noeud de substitution étant en outre adapté pour envoyer (12) les informations de défaillance du noeud défaillant aux LDN à l'intérieur de chaque plage des noeuds ; et les LDN étant adaptés pour mettre à jour leurs informations de table de routage correspondantes selon les informations de défaillance et pour transmettre les informations de défaillance aux noeuds de leurs plages pour mettre à jour leurs tables de routage, les informations mises à jour dans les tables de routage dirigeant vers le noeud de substitution.


 
5. Réseau selon la revendication 4, comprenant en outre :
le noeud voisin étant en outre adapté pour déterminer le noeud de substitution pour envoyer les informations de défaillance entre des noeuds voisins du noeud défaillant conformément à des caractéristiques de valeur clé stockées dans le réseau.
 
6. Réseau selon la revendication 5,
le noeud voisin étant en outre adapté pour déterminer le noeud de substitution qui est
le noeud voisin à la distance XOR minimale à l'opposé du noeud défaillant, si le réseau est un réseau DHT Kademlia ou un réseau DHT Pastry ; ou
le noeud voisin étant en outre adapté pour déterminer que le noeud de substitution doit être un noeud défaillant successeur le plus proche du noeud défaillant si le réseau est un réseau DHT Chord ou un réseau DHT Koorde.
 




Drawing





























Cited references

REFERENCES CITED IN THE DESCRIPTION



This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

Non-patent literature cited in the description