CN107302500B

CN107302500B - Single-node fault protection method based on hop-by-hop mode

Info

Publication number: CN107302500B
Application number: CN201710436099.3A
Authority: CN
Inventors: 耿海军; 张举
Original assignee: Shanxi University
Current assignee: TAIYUAN QINGZHONGXIN SCIENCE AND TECHNOLOGY Co.,Ltd.
Priority date: 2017-06-12
Filing date: 2017-06-12
Publication date: 2020-06-12
Anticipated expiration: 2037-06-12
Also published as: CN107302500A

Abstract

The invention discloses a single-node fault protection method based on a hop-by-hop mode, belongs to the technical field of internet, and solves a route protection method capable of protecting all single-node faults in a network by 100%. The scheme comprises the following steps: node d calculates shortest path tree rspt (d) taking itself as root; the final bridge of the subtree is computed and then the backup next hops for all nodes are computed. The invention provides an effective single-node fault route protection scheme. The scheme is compatible with the intra-domain routing protocol deployed by the current Internet, so that incremental deployment is supported, and the scheme is easy to deploy in an actual environment.

Description

Single-node fault protection method based on hop-by-hop mode

Technical Field

The invention belongs to the technical field of internet, relates to an intra-domain route protection scheme, and particularly relates to a single-node fault protection method based on a hop-by-hop mode.

Background

In recent years, the development speed of the internet has been far beyond the expectations, and the range of applications supported by the internet is also expanding. The internet is rapidly developing and faces new challenges, and the Routing Availability (Routing Availability) is one of the problems to be solved urgently. Related research has shown that failures in the network occur frequently and are unavoidable. During the fault repair period, the routing protocol needs to go through a convergence process for a period of time, and a large amount of messages are lost in the convergence process of the routing protocol, so that the routing availability is greatly reduced. However, with the advent of new applications, such as VoIP, online gaming, video, these applications place more stringent requirements on network latency. Therefore, how to improve the network availability and reduce the message loss rate in the routing protocol convergence process is an important challenge for the internet. In order to solve the problem, the academic and industrial fields propose a route protection scheme, the scheme pre-calculates a backup route, and when a network fails, a pre-calculated backup path is used for forwarding a message affected by the failure, so that the message loss rate is effectively reduced.

The IP fast rerouting scheme is favored by academia and industry with its unique advantages, and it uses Loop Free rules (LFA) to calculate backup next hop, thereby implementing route protection. The scheme is small in computational complexity and easy to deploy, so that routers of most manufacturers support the scheme. However, studies have shown that only around 50% of single fault situations can be protected with this scheme. Based on an IP fast rerouting framework, an efficient route protection scheme is provided in an article. The algorithm of the scheme is low in complexity, dynamic updating is supported, incremental deployment is supported, and the fault protection rate of the scheme is close to that of IP fast reroute.

To further improve network availability, it is proposed to improve the failure protection rate using a U-turn scheme that can compute LFA next hops in its neighbor nodes. U-turn, while a significant improvement in the failsafe rate, still does not achieve the intended goal. Based on the defects of the fast reroute and the U-turn, a fast reroute scheme based on the Not-Via address is provided. The scheme utilizes a secondary mechanism Not-Via address to explicitly indicate a failure in the network, thereby effectively protecting all single failure situations in the network. Although the mechanism can greatly improve the availability of the network, the mechanism is complex to implement, high in cost and not easy to actually deploy.

Based on the defects of the scheme, a route protection scheme combining the IP-based fast rerouting and the Not-Via address-based fast rerouting is provided. The basic idea of this scheme is to use an IP fast reroute scheme when there is an IP fast reroute next hop for a node, whereas when the node is Not protecting the next hop as described above, the protected next hop is calculated using Not-Via address based fast reroute. Compared with a Not-Via address based fast rerouting scheme, the fusion scheme greatly reduces the complexity of the algorithm, but the scheme still needs to use Not-Via addresses, so that actual deployment is Not easy. In summary, the IP fast rerouting scheme is simple to implement, but has a low failure protection rate. Not-Via can protect 100% of single node failure situations in the network, however, the scheme is difficult to deploy. Therefore, the invention designs a route protection method which can protect all single-node failure situations in the network by 100 percent.

Disclosure of Invention

The technical problem to be solved by the invention is to provide a single-node fault protection method based on a hop-by-hop mode, and the method protects all single-node fault situations in a network by 100%. Since the present invention is a distributed solution, and the methods adopted by all nodes are the same, it is assumed that the computing node is node d in the following. For convenience of description, we first define some labels, which apply to the entire invention. The network topology may be represented by the graph G ═ V, E, where V represents the set of routers in the network and E represents the set of links in the network. For a certain link in the network, e is (x, y), and w (e) represents the weight of the link, which may be the number of hops, delay, bandwidth, energy consumption, etc., or a combination of these several metrics. Backup (s, d) represents the Backup next hop from node s to node d; cost (s, d) represents the minimum cost from node s to node d, rspt (d) represents a reverse shortest path tree with node d as the root, child (d, v) represents the child nodes of node v in rspt (d), and subtree (d, v) represents all the nodes of the subtree with node v as the root in rspt (d).

In order to solve the technical problem, the invention provides a single-node fault protection method based on a hop-by-hop mode, which comprises the following steps:

step 1: calculating a reverse shortest path tree rspt (d) taking the node d as a root node;

step 2: setting backup next hops of all nodes in a network to be null, setting access marks of all nodes to be unaccessed, and marking the colors of all nodes to be white;

and step 3: traversing nodes which are not accessed in the shortest path tree taking d as a root according to a depth-first algorithm; if the traversed node is the node which is not visited, executing the step 4; if all the traversed nodes of the shortest path tree rspt (d) which takes the node d as the root node are set as the visited nodes, the method is terminated;

and 4, step 4: representing the node traversed each time as a node v, and setting the access identifier of the node v as accessed;

and 5: marking all nodes in the subtree (d, v) of the subtree as red;

step 6: child nodes which are not accessed of the access node v are judged according to a method for calculating the first class of bridges; if the node u belongs to the subtree corresponding to the child (d, v) and only has one type of bridge, selecting the bridge as the final bridge of the child number, and marking as (x, y); if the node u belongs to a subtree corresponding to the child (d, v) and has a plurality of first bridges, calculating a bridge with the shortest rerouting path as the final bridge of the subtree according to a method for calculating the final bridge, and marking as (x, y);

the method for calculating the first type of bridge is as follows:

in a reverse shortest path tree taking d as a root, for any node V belongs to V-d in the tree, if a link exists, so that x belongs to a subtree (d, u) and y belongs to a V-subtree (d, V) -d simultaneously, the link (x, y) is a first type bridge of the subtree (d, u) when a node u belongs to child (d, V); if the node u belongs to a subtree corresponding to child (d, v), the first type of bridge is arranged, all nodes in the subtree (d, u) are marked to be red, all nodes in the subtree (d, u) are traversed according to a depth-first algorithm, each neighbor node y of the node x in the subtree is checked, and if the color of the node is white, the link (x, y) is the bridge of the subtree (d, u);

the method of computing the final bridge is as follows: calculating the cost of the rerouting path of the node u by the following formula, wherein r (u, d) is cost (u, x) + cost (x, y) + cost (y, d), and r (u, d) represents the cost of the rerouting path from the node u to the node d; the variables involved in the formula can be easily obtained from the link state routing protocol; the cost (x, y) can be obtained from a link state database, and the other variables can be obtained through rspt (d), so that the cost of the corresponding rerouting path of the bridge can be easily calculated; for all found bridges, selecting one bridge with the shortest reroute path as its final bridge;

and 7: if the subtree corresponding to the child node of the node v does not have the first class bridge, calculating the second class bridge of the subtree according to the calculation method of the second class bridge, and taking the second class bridge as the final bridge of the subtree and marking as (x, y);

the second type of bridge is calculated as follows:

in a reverse shortest path tree taking a destination address d as a root, if a link (p, q) exists for any node V belongs to V-d, when nodes u belongs to child (d, V) and w belongs to child (d, V), so that p belongs to subtree (d, u) and q belongs to subtree (d, w) are simultaneously established, the link (p, q) is a second type bridge of subtree (d, u) and subtree (d, w); if the node u belongs to a subtree corresponding to the child (d, v) and only has a second type of bridge, traversing all nodes in the subtree (d, u) according to a breadth-first algorithm, and searching for an edge (x, y) appearing for the first time, wherein x is red, y is green, and then a link (x, y) is the final bridge of the subtree;

and 8: calculating a backup next hop according to the selected final bridge as a corresponding node; the method comprises the following steps:

and calculating the rerouting path of the node u according to the selected final bridge, and representing the rerouting path of the node u by using (u, m.. x, y.. p, q), wherein the backup next hop of the corresponding node is as follows: backup (u, d) m, Backup (x, d) y …, Backup (p, d) q; if the node has the backup next hop, setting the access identifier of the node as accessed;

and step 9: if the child node u of the node v has a backup next hop, all nodes in the subtree corresponding to the child node are marked as green;

step 10: detecting the color of the node in the subtree (d, v), and if the color of the node in the subtree (d, v) has red, executing the step 6;

step 11: detecting the colors of the nodes in the subtree (d, v), if the colors of all the nodes in the subtree (d, v) are all green, marking the colors of all the nodes in the subtree (d, v) as white, and executing the step 3;

step 12: when a node u receives a message, detecting according to a fault detection method, and if the node u reaches the default next hop of a destination address without fault, directly forwarding the message to the default next hop; if the node u reaches the default next hop of the destination address and has a fault, directly forwarding the message to the backup next hop;

the fault detection method comprises the following steps: in an actual network, node faults in the network can be quickly detected through a bidirectional forwarding Detection mechanism (BFD), wherein the BFD is a high-speed independent HELLO protocol, and the BFD establishes sessions on two network devices, is used for detecting bidirectional forwarding paths between the network devices and serves upper-layer application; BFD does not have a neighbor discovery mechanism per se, but relies on the served upper layer application to inform the neighbor information thereof so as to establish a session; and after the session is established, the BFD message can be periodically and quickly sent, if the BFD message is not received in the detection time, the bidirectional forwarding path is considered to have a fault, and the upper layer application to be served is informed to carry out corresponding processing.

Compared with the prior art, the invention has the following advantages:

the invention provides a single-node fault protection method based on a hop-by-hop mode, which can protect all possible single-node fault situations in a network by 100 percent and can greatly improve the network availability, thereby better supporting the application sensitive to time delay. In addition, the invention fully utilizes the diversity of the paths in the network, thereby having higher reliability.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

FIG. 1 is a schematic flow chart of the protection method of the present invention.

Fig. 2 is a schematic diagram of a network topology according to an embodiment of the present invention.

Fig. 3 is a diagram illustrating the calculation of a reverse shortest path tree rooted at node d according to an embodiment of the present invention.

Fig. 4 is a schematic diagram illustrating the calculation of a protection path for a node e when accessing the node a according to the embodiment of the present invention.

Fig. 5 is a schematic diagram illustrating the calculation of a protection path for node c when node a is accessed in the embodiment of the present invention.

Fig. 6 is a schematic diagram illustrating the calculation of a protection path for node b when node a is accessed in the embodiment of the present invention.

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings.

FIG. 1 is a schematic flow diagram of a process according to the present invention.

The respective steps of the present embodiment are explained in detail below. The destination node is d, and the specific steps are as follows:

step 1: the node d constructs a reverse shortest path tree taking the node d as a root, and the reverse shortest path tree is shown in the figure 3;

step 2: setting backup next hops of all nodes in the network to be null, and marking the colors of all nodes to be white;

and step 3: traversing all nodes in the shortest path tree taking d as a root according to a depth-first algorithm;

and 4, step 4: representing the visited node as node v, where v is a;

and 5: marking all nodes subtree (d, v) ═ { b, c, e, f, g, i } in the subtree of v as red;

step 6: finding out a first type of bridge of subtrees of all child nodes of the node v, wherein the first type of bridge can be expressed as { (g, k), (i, l), (i, m) };

and 7: this step does not perform any operation because the subtree (d, e) corresponding to the child node e of node v has a first type of bridge;

and 8: when the first type bridge (g, k) of the subtree (d, e) is selected, the cost of the rerouted path of the node e is 29, when the first type bridge (i, l) of the subtree (d, e) is selected, the cost of the rerouted path of the node e is 32, and when the first type bridge (i, m) of the subtree (d, e) is selected, the cost of the rerouted path of the node e is 38; link (g, k) is therefore selected as the final bridge of the subtree (d, e) and all nodes in the subtree are marked green. According to the selected final bridge, the rerouting path of the node e is (e, g, k, j, d), so that back (g, d) is k, and back (e, d) is g;

and step 9: because the child node e of the current v already has the backup next hop, all subtrees corresponding to the child node e are marked as green, as shown in fig. 4;

step 10: because there is a red node in the subtree (d, v), step 6 is executed;

step 6: because the subtrees of the remaining child nodes of node v now have only bridges of type two, step 7 is performed;

and 7: selecting link (c, g) as the final bridge of the subtree rooted at node c;

and 8: according to the selected final bridge, the rerouting path of the node c is (c, g, k, j, d), so that Backup (c, d) is g;

and step 9: because the child node c of the current v already has the backup next hop, all subtrees corresponding to the child node c are marked as green, as shown in fig. 5;

step 10: because there is a red node in the subtree (d, a), step 6 is executed;

and 7: therefore, only one bridge is selected, and therefore link (b, c) is selected as the final bridge of the subtree rooted at b;

and 8: according to the selected final bridge, the rerouting path of the node b is (b, c, g, k, j, d), so that Backup (b, d) is c;

and step 9: because the child node b of the node v already has the backup next hop, all subtrees corresponding to the child node are marked as green, as shown in fig. 6;

step 10: because there is no red node in subtree (d, a), no operation is performed;

step 11: because all nodes in subtree (d, a) are green, marking all nodes subtree (d, v) { b, c, e, f, g, i } in the subtree of v as white, executing step 3;

and step 3: traverse the next node in the network rspt (d) that has not been visited.

The operation of the present invention will be described in detail below.

Definition 1: for any event (u, d, f), where u represents the source address, d represents the destination address, and f represents a single node failure on the u to d optimal path SP (u, d). When this event occurs, if u to d still have another path, the two remain connected. That is, there is no cut point in the network diagram, and when any node in the network topology fails, the connectivity of the network is not affected, so the network is said to have a robust topology.

Principle 1: for a robust network topology, suppose node f fails, and node f has k child nodes, each of which is designated as (f)₁,f₂...f_k) Meaning that there must be one child node f_xE.g. child (f), subtree (d, f) corresponding to the child node_x) At least one bridge of a first type; when a child node f_yE, child (f) when the corresponding subtree has no first kind bridge, the subtree (d, f) corresponding to the child node_y) There is at least one bridge of the second type.

And (3) proving that: the principle is demonstrated below using a counter-proof method.

The first half of the principle is first demonstrated. When node f fails, it is assumed that all child nodes of node f do not have a bridge of the first type. I.e. for an arbitrary node f_xE child (f) there is no link (p, q) so that p e subtree (d, f)_x) And q ∈ V-subtree (d, f) -d are true at the same time. Then for any node p ∈ subtree (d, f)_x) The other end q of the link connected to node p is only connected to a node in the set subtree (d, f), and the optimal path from q to d must pass through node f. As can be seen from the above description, when node f fails, node f_xE child (f) the nodes in the corresponding subtree will not reach the destination. This is in contradiction to the premise assumption of a robust network topology. The first half of the principle holds.

The latter half of the principle is demonstrated below. When node f fails, for node f_yE child (f), when the subtree corresponding to the node has no bridge of the first type, the subtree corresponding to the node is assumed to have no bridge of the second type. I.e. for node f_yE child (f), there is no link (m, n), so that m e subtree (d, f)_y) And n ∈ subtree (d, f)_k) At the same time, wherein f_yE child (f). Then m ∈ subtree (d, f) for any node_y) The other end n of the link connected to node m is simply summed with the set subtree (d, f)_y) The nodes in (b) are connected, and the optimal path from n to d must pass through the node f. As can be seen from the above description, when node f fails, node f_yE.g. child (f) pairThe nodes in the corresponding subtree will not reach the destination. This is in contradiction to the premise assumption of a robust network topology. The second half of the principle holds true.

From the above proof, this principle holds.

Principle 2: when all nodes in the network deploy the routing protection algorithm, any single node in the network can be protected from failure.

And (3) proving that: in the network topology, it is assumed that node f fails. For any source-destination (s, d), when the optimal path from the source to the destination passes through the node f, it can be known from principle 1 that at least one bridge exists in the subtree (d, f), and then the algorithm can protect the node f. Since the node s is a descendant node of the node f, when the node f is disconnected, the message sent from the source node s to the destination node d can arrive normally.

Claims

1. A single-node fault protection method based on a hop-by-hop mode comprises the following steps:

and 5: marking all nodes in a subtree (d, v) of the subtree as red, wherein the subtree (d, v) represents all nodes of the subtree taking the node v as a root in rspt (d);

step 6: child nodes which are not accessed of the access node v are judged according to a method for calculating the first class of bridges; if the node u belongs to a subtree corresponding to child (d, v) and has only one type of bridge, selecting the bridge as a final bridge of the subtree, and recording the final bridge as (x, y), wherein child (d, v) represents a child node of the node v in rspt (d); if the node u belongs to a subtree corresponding to the child (d, v) and has a plurality of first bridges, calculating a bridge with the shortest rerouting path as the final bridge of the subtree according to a method for calculating the final bridge, and marking as (x, y);

the method for calculating the first type of bridge comprises the following steps:

in a reverse shortest path tree taking d as a root, for any node V belongs to V-d in the tree, wherein V represents a set of routers in a network, when a node u belongs to child (d, V), if a link exists, so that x belongs to subtree (d, u) and y belongs to V-subtree (d, V) -d are simultaneously established, the link (x, y) is a first type bridge of the subtree (d, u); if the node u belongs to a subtree corresponding to child (d, v), the first type of bridge is arranged, all nodes in the subtree (d, u) are marked to be red, all nodes in the subtree (d, u) are traversed according to a depth-first algorithm, each neighbor node y of the node x in the subtree is checked, and if the color of the node is white, the link (x, y) is the bridge of the subtree (d, u);

the method for calculating the final bridge comprises the following steps:

calculating the cost of the rerouting path of the node u by the following formula, wherein r (u, d) is cost (u, x) + cost (x, y) + cost (y, d), wherein cost (u, x) represents the minimum cost from the node u to the node x, cost (x, y) represents the minimum cost from the node x to the node y, cost (y, d) represents the minimum cost from the node y to the node d, and r (u, d) represents the rerouting path cost from the node u to the node d; for all found bridges, selecting one bridge with the shortest rerouting path as the final bridge of the subtree;

the method for calculating the second type of bridge comprises the following steps:

and calculating the rerouting path of the node u according to the selected final bridge, and representing the rerouting path of the node u by using (u, m.. x, y.. p, q), wherein the backup next hop of the corresponding node is as follows: backup (u, d) m, Backup (x, d) y …, Backup (p, d) q, Backup (u, d) representing a Backup next hop from node u to node d, Backup (x, d) representing a Backup next hop from node x to node d, and Backup (p, d) representing a Backup next hop from node p to node d; if the node has the backup next hop, setting the access identifier of the node as accessed;

step 11: and detecting the colors of the nodes in the subtree (d, v), if the colors of all the nodes in the subtree (d, v) are all green, marking the colors of all the nodes in the subtree (d, v) as white, and executing the step 3.

2. The single-node fault protection method based on the hop-by-hop mode according to claim 1, characterized in that: further comprising the steps of:

step 12: when a node u receives a message, detecting according to a fault detection method, and if the node u reaches the default next hop of a destination address without fault, directly forwarding the message to the default next hop; and if the node u fails to reach the default next hop of the destination address, directly forwarding the message to the backup next hop.