KR100814077B1 - An ip address lookup method using priority-trie and apparatus for relaying packets therefor - Google Patents

Abstract

An IP(Internet Protocol) address lookup method using priority-trie and a packet relay apparatus providing the same are provided to replace empty nodes with priority prefix on the basis of a priority-trie structure, thereby improving a search rate and extension. An IP address search method comprises the following steps of: extracting a destination IP address of a received packet(S100); searching for a network address matching the extracted destination IP address for a longest time from a routing table in which an output port according the network address is designated(S102); and designating the output port corresponding to the searched network address as the output port to transmit the received packet(S104).

Description

An IP address lookup method using priority-trie and apparatus for relaying packets therefor}

1 is an example of a prefix set for explaining an IP address searching method;

2 is an example of building a conventional binary tri structure using the prefix set;

3 is a block diagram of a packet relay apparatus according to an embodiment of the present invention;

4 is a detailed block diagram of an interface included in the packet relay apparatus;

5 is a flowchart illustrating a method of searching for an IP address using priority trie according to an embodiment of the present invention;

6 is an example of constructing a priority tri structure according to the present invention using the prefix set shown in FIG. 1;

7 is an example of a routing table having a priority based tri structure according to the present invention;

8 is a view comparing the number of memory accesses of the conventional binary tri structure and the priority tri structure according to the present invention;

9 to 11 are examples of pseudo code implementing a method of constructing, searching, and updating a priority trie according to the present invention;

12 to 14 show experimental results of evaluating the performance of the priority tri structure according to the present invention.

<Description of the symbols for the main parts of the drawings>

10: interface 20: switch fabric

11: packet transceiver 12: frame analyzer

13: Address lookup processing unit 14: Packet classification processing unit

15: transmission control unit

The present invention relates to an IP address lookup method in a packet relay apparatus, and proposes a method of effectively searching for an IP address using a priority tree.

Recently, with the explosive increase in the use of the Internet, Internet traffic is increasing rapidly. Accordingly, it is required that a packet relay device such as a router performs packet forwarding at high speed. Improving the router's IP address lookup efficiency is essential for high speed packet forwarding. This is because each packet to be input is required to be processed at wire-speed under an environment in which the arrival rate of the packets and the size of the routing table increase rapidly.

Here, the IP address lookup means that the output port is determined using the destination IP address (Internet Protocol address) of the received packet so that the incoming packet is forwarded to the final destination. An IP address has a two-level hierarchy of network parts and host parts, and the dual network part is called a prefix. Hereinafter, the network portion of the IP address is collectively called a prefix.

Conventionally, a prefix having a certain length for each class is used, but there is a problem in that an IP address is wasted due to a class structure having a fixed length. Improvements have recently been made to allow combinations of arbitrary length prefixes and addresses using classless inter-domain routing (CIDR). Therefore, as the length of the prefix in the IP address varies, the router needs to find the longest prefix in the routing table that matches the destination address of the input packet in order to find the specific network path of the IP address. At this time, it is necessary to determine the longest matching prefix (abbreviated as LMP) or the best matching prefix (abbreviated as BMP) for IP address retrieval. To take into account both factors, the length of the prefix and the value.

The conventional technique most representatively used for IP address search of the CIDR structure is a binary search method using a trie structure. A tri is a tree-based data structure that uses linear search along its length. Tri-structures are used to represent prefix lengths and values in CIDR-type IP addresses. That is, each prefix exists in one node of the tri and the level of that node corresponds to the length of the prefix. The search in the tri-structure is checked sequentially from the Most Significant Bit (MSB) of the destination IP address of the incoming packet, starting from the root node of the tri, and if the checked bit is 0, it follows the pointer pointing to the left child node. Follow the same pointer down to the right child node and repeat the process until you meet the leaf of the tri.

Meanwhile, FIG. 1 shows an example of a prefix set for explaining an IP address searching method, and FIG. 2 shows an example of building a conventional binary tri structure using the prefix set.

Nodes marked in black in FIG. 2 represent prefixes, and nodes marked in white represent empty internal nodes with no prefix assigned. For example, the prefix 1000 * that is P2 in the prefix set of FIG. 1 is represented by a black node denoted by P2 of FIG. The P2 node has a value of 1000 * and its length is 4, so it is located at the fourth level from the root node.

The first problem with this tri structure is that short length prefixes are stored at the top level of the tri and long length prefixes are stored at the lower level of the trie because the length of the prefix corresponds to the level of the tri. That is, shorter prefixes are stored at a higher level and compared before longer prefixes. Therefore, even when a matching node is found during a search, the search must continue until a leaf node is reached. This is because even if a matching prefix node is encountered in the middle, there may be a longer matching prefix at a lower level. Therefore, since the search must be performed every leaf node, the number of memory accesses increases and the search time becomes long. In addition, the tri node includes a large number of empty internal nodes, which increases the number of memory accesses, thereby increasing search time and wasting memory space.

Accordingly, an object of the present invention to solve the above problems is to provide an effective IP address search method that can reduce the search time and the required memory size.

In addition, another object of the present invention is to provide a packet relay apparatus applying an effective IP address search method that can reduce the search time and the required memory size.

According to the present invention, the above object is achieved by the method of IP address retrieval for packet forwarding, comprising: (a) extracting a destination IP address of a received packet; (b) searching for a network address whose output port according to the destination address is the longest match with the destination address extracted from the designated routing table; And (c) designating an output port corresponding to the retrieved network address as an output port to which the received packet is to be sent, wherein the retrieving comprises priority priority generated based on the network addresses included in the routing table. -trie) is achieved by the IP address retrieval method characterized in that it is made using the structure.

Here, the priority tri structure is preferably constructed by assigning a long prefix to a high level using empty internal nodes of the general tri structure so that the longer prefix is compared first.

Further, the priority tri structure is particularly preferably constructed by replacing each bean inner node of the general tri structure with the longest one of the prefixes belonging to the subtree whose bin inner node is the root node.

Further, retrieving the network address may include (b1) comparing the extracted destination address with each node of the priority tri structure; (b2) If the destination address extracted as a result of the comparison matches the general node, the prefix of the matching general node is stored as the longest prefix matched so far, and the search continues, (b3) The destination address extracted as a result of the comparison takes precedence. Terminating the search if it matches a priority prefix or replaces a leaf in the rank tri structure with an empty prefix; And (b4) if the search is terminated by matching the priority prefix, the priority prefix is returned as the longest matching prefix with the extracted destination address; It is preferable to include the step of returning the prefix to the longest matching the extracted destination address.

Meanwhile, according to an aspect of the present invention, in the method for constructing a routing table for IP address retrieval, (a) storing the prefixes representing network addresses as nodes having a level corresponding to their length, respectively, Building a tri structure; And (b) reconstructing the priority-based tri-structure structured so that the longer prefix from the general tri-structure is compared first.

Meanwhile, according to an aspect of the present invention, the above object is a method for updating a routing table for IP address retrieval (a) priority-trie generated based on network addresses included in the routing table. Making a node to be deleted an empty internal node when deleting one prefix from the structure; And (b) reconstructing the priority based tri structure for the routing table at predetermined intervals to remove empty internal nodes.

Meanwhile, according to another aspect of the present invention, the above object is a method for updating a routing table for IP address retrieval (a) priority-trie generated based on network addresses included in the routing table. When inserting one prefix in the structure, arranging the prefix to be inserted occupies the position of the priority prefix when the length of the prefix to be inserted is longer than the length of the priority prefix corresponding to the prefix; And (b) relocating the prefix to be inserted to occupy the position of another priority prefix if the general position of the prefix to be inserted is already occupied by another priority prefix. do.

On the other hand, according to an aspect of the present invention, the above object is a packet relay apparatus for IP address search, the frame analysis unit for extracting the destination IP address of the received packet; And an address lookup processing unit for retrieving a network address whose longest output port is matched with a destination IP address extracted from a specified routing table and specifying an output port corresponding to the retrieved network address as an output port to which a received packet is to be transmitted. And the address lookup processing unit is configured to perform an IP address search using a priority-trie structure generated based on network addresses included in a routing table.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the present invention; In the following description of the present invention, if it is determined that detailed descriptions of related well-known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, terms to be described below are terms defined in consideration of functions in the present invention, which may vary according to intention or custom of a user or an operator. Therefore, the definition should be made based on the contents throughout the specification.

First, the structure of the packet relay apparatus will be described, and the IP address searching method according to the present invention will be described in detail.

3 is a block diagram of a packet relay apparatus according to an embodiment of the present invention. A packet relay device is a relay device that connects one of a plurality of input ports and one of a plurality of output ports to output an input packet to a designated output port, and a router is representative.

Referring to FIG. 3, the packet relay apparatus according to the present invention includes a plurality of interfaces 10 and a switch fabric 20. In addition, a controller such as a management controller or a route controller may be further included. The packet relay apparatus may include a plurality of input ports and a plurality of output ports to output a packet input to one of the plurality of input ports to the designated output port via the switch fabric 20. The IP address retrieval function according to the invention is responsible for the interface 10.

4 is a detailed block diagram of an interface included in the packet relay apparatus. Referring to FIG. 4, the interface 10 includes a packet transmitter / receiver 11, a frame analyzer 12, an address lookup processor 13, and a transmission manager. (15) is provided. In addition, a packet classification processor 14, a CPU, and a memory are further provided.

The packet transmitter / receiver 11 receives a packet through an input port and transmits a packet through an output port. The frame analysis unit 12 detects the start and end of the input packet, analyzes the packet, separates the header of each layer, and analyzes the source address, the destination address, the source port, and the destination port. In particular, it extracts the destination IP address needed for packet forwarding.

The address lookup processor 13 designates an output port to which the received packet is transmitted by referring to a destination IP address of the packet received through the input port. The address lookup processor 13 searches for a network address whose output port corresponding to the network address is the longest match with the destination IP address included in the packet received from the specified routing table, and prioritizes the priority-trie according to the present invention. IP address lookup using the A specific priority based tri structure will be described later.

The traffic manager 15 controls the output port corresponding to the network address retrieved by the address lookup processor 13. In addition, the packet classification processor 14 searches for a rule matching the flow condition of the received packet from the packet classification table in which the rules for classifying the packet are stored. Whether to include a packet classification processing unit in a packet relay device such as a router is a designer's option.

An IP address retrieval method using a priority trie according to the present invention will be described based on the structure of the packet relay apparatus described above.

Referring to FIG. 5, first, a packet received by the packet relay apparatus is interpreted to extract a destination IP address (S100), and a network that matches the longest destination IP address extracted from a routing table having a priority-based tri-structure described later. Search for the address (S102). The routing table is a table designating an output port according to a network address, and each entry is stored in a priority based tri structure according to the present invention. That is, each entry is stored according to a priority-based tri-structure, in which long prefixes are assigned to high levels using empty internal nodes of a general tri-structure so that longer prefixes are compared first. Accordingly, even when a node matching the priority prefix is found even before reaching the leaf node of the tree at the time of IP address search, the search can be terminated. A more specific search method will be described later. Now, referring to the routing table, the output port corresponding to the retrieved network address is designated as a port to output the received packet (S104).

Hereinafter, the priority tri structure according to the present invention will be described in more detail.

The present invention proposes to use an empty internal node present in the regular tri to assign a longer prefix to a higher level. That is, the bin node is combined with the longest prefix among prefixes belonging to the subtree whose root node is the root. This eliminates all empty internal nodes, reduces the depth of the trie, and makes searching more efficient. At this time, the prefixes bound to the bean internal nodes are called priority-prefixes. In this way, the prefixes do not need to be duplicated according to the present invention because the empty internal nodes are allocated to prefixes that are equal to or longer than their own lengths.

6 illustrates an example of constructing a priority tri structure according to the present invention using the prefix set shown in FIG. 1. Referring to FIG. 6, black nodes represent ordinary prefixes placed at their level, and white nodes use a priority internal prefix to assign a longer prefix to a higher level using an empty internal node. ).

2 illustrates an example in which a conventional general tri structure is constructed using the same prefix set shown in FIG. 1. First, the longest prefix in the general tri in FIG. 2 has a value of 100110 * as P4 and is stored at the lowest level. The result of reconstructing this into the priority tri structure according to the present invention is shown in FIG. 6. Referring to FIG. 6, it can be seen that the prefix P4 stored at the lowest level is now assigned to the root node which is an empty internal node and becomes a priority prefix.

As such, when the empty node is reconstructed in the normal tri as shown in FIG. 2 by constructing the priority node by combining the empty node with the longest prefix among the prefixes belonging to the subtree whose root node is the root, the result is as shown in FIG. Is obtained. That is, according to the present invention, a longer prefix is stored in an empty node of a higher level. 2 and 6, the prefixes P0, P2, P3, P4, P5, and P7 are allocated to empty internal nodes to become priority prefixes. Here, if two prefixes having the same length exist, it is assumed that the left prefix has priority. As a result, all empty internal nodes are removed and the depth of the trie is reduced, thereby enabling more efficient searching.

FIG. 7 illustrates an example of a routing table having a priority based tri structure shown in FIG. 6. Referring to FIG. 7, the first column is a memory address and may be any value. The second column indicates whether the prefix is a priority prefix (1) or a general prefix (0). The prefix and its length are stored in the third and fourth columns, respectively. The fifth and sixth columns represent the memory addresses where child nodes on the left and right of the current node are stored. The last column is the output port that corresponds to the prefix. For this example, the prefix name is simply stored for ease of explanation. As shown, in the proposed algorithm, the number of entries in the routing table is N, and N entries are used to store N prefixes. This indicates that there are no empty nodes inside.

8 is a view comparing the number of memory accesses of the conventional binary tri structure and the priority tri structure according to the present invention.

Referring to FIG. 8, it can be seen that the number of memory accesses of the priority tri structure shown in FIG. 6 is much reduced compared to the number of memory accesses of the general tri structure shown in FIG. 2. In particular, it can be seen that as the prefixes P0, P2, P3, P4, P5, and P7 marked with the priority prefix are allocated to the higher level empty internal nodes, the number of memory accesses is significantly reduced. As described above, using the priority tri structure according to the present invention, all empty internal nodes are removed and the depth of the trie is reduced, thereby saving memory space and reducing the number of memory accesses, thereby enabling more efficient searching.

9 to 11 are examples of pseudo code implementing the method of constructing, searching for, and updating the priority trie according to the present invention.

Referring to FIG. 9, a method of constructing a priority-based trie according to the present invention includes a process of building a general binary trie whose length of a first prefix corresponds to a level of a node (BuildBinaryTire), and a second longer prefix. This is done by rebuilding the priority tri (BuildPriorityTrie) using an empty internal node that exists in the regular binary tri to assign to.

Here, the process of rebuilding the priority tris will be described in detail with reference to FIG. 9. First, if a general non-empty node is found, the general binary tri is traversed. If it finds an empty node, it sorts the prefixes belonging to the subtree rooted at that root, E ( x ), and removes the longest prefix of that sorted prefix, P ( x ), from the trie. We then store P ( x ) in that empty node. The node where P ( x ) is stored is now marked as a priority node. Note that P ( x ) must move to a level higher than its current prefix length. This process is repeated until all nodes in the regular binary tree are retrieved or until the regular binary tree no longer contains empty nodes. Since all empty nodes are removed when the priority tri rebuild process is complete, the number of nodes in the tri is equal to the number of prefixes.

On the other hand, Figure 10 shows an example of the pseudo code implementing the method for searching the priority tri in accordance with the present invention.

Referring to FIG. 10, first, an IP address inputted at a root node is compared with nodes stored in a priority tri. If the address entered matches the normal node, the prefix of the matching general node is stored as the longest prefix matched so far and the search continues.

The search is terminated when the extracted address matches the priority prefix replaced by an empty node in the priority tri-structure or meets a leaf.

In this case, when the search is terminated in accordance with the priority prefix, the corresponding priority prefix may be returned to the network address that has the longest match with the extracted address. Since the priority based tri according to the present invention is rebuilt to store the long prefix at a higher level, matching with the priority prefix during the search ensures that it is the longest matching node in the subtree. Therefore, it is no longer necessary to compare with subordinate nodes. As a result, the number of memory accesses for searching can be significantly reduced.

On the other hand, when the end of the leaf meets, the longest prefix that has been matched so far may be returned as the network address that has the longest match with the extracted address.

Meanwhile, FIG. 11 shows an example of pseudo code implementing the method of updating the priority trie according to the present invention.

For reference, although not shown, a node to be deleted is made an empty internal node to delete one prefix in a priority tri according to the present invention, and the priority is given to the routing table at predetermined intervals to remove the empty internal node. It is desirable to rebuild the base tri structure. Specifically, the prefix to be deleted is found, and the first three fields of the entry of the routing table shown in FIG. 7, that is, the memory address, the mark information indicating whether the node is a general node or the priority node, and the prefix value are deleted to make the empty node. You must keep this node because it still stores a pointer to the child node so that you can continue searching down to the lower level. If we assume that we rebuild the entire routing table at a reasonable interval, the number of free nodes will not matter as we delete the prefix.

Meanwhile, the update procedure shown in FIG. 11 is performed to insert one prefix in the proposed algorithm.

Referring to FIG. 11, when a prefix is inserted in a priority-trie structure generated based on network addresses included in a routing table, two cases where a plurality of nodes are affected by the insertion There is. The first is when the prefix to be inserted matches the priority prefix and its length is longer than the priority prefix. The second is when another priority prefix already occupies the general position of the prefix to be inserted. In these cases the prefix to be inserted occupies the position of the priority prefix. The same procedure is repeated for priority prefixes deprived by the prefix to be inserted.

 In all other cases, the prefix to be inserted is stored in the leaf node or the existing prefix in the routing table is replaced.

12 to 14 show experimental results of evaluating the performance of the priority tri structure according to the present invention.

The experimental results shown represent simulation results for the actual routing table of the backbone router. 12 and 13 show the performance evaluation results of the proposed priority binary tri and priority based multi-bit tri (in case of 2 bits).

Each N is the number of routing prefix, N _P is a routing prefix of the number of the priority prefix, D is the maximum prefix length, D _p is the depth of the proposed priority tri, T a is the average memory access for address searches, M represents the size of memory required for routing data of various sizes. As indicated by the number of priority prefixes in FIG. 12 for the first three routing tables, more than 90% of the prefixes are stored as priority prefixes. This means that a regular binary tree has many empty nodes.

According to the proposed algorithm, the more priority prefix nodes, the better the search performance. The average number of memory accesses is about 16-23, and the size of the routing table does not increase significantly. In FIG. 13 two bits are considered to be the same as in a two bit tri. It can be seen that two nodes are created with separate nodes. In terms of tri depth, the performance of the proposed algorithm, that is, the average number of memory accesses and the required memory size, does not deteriorate significantly as the size of the routing table increases. In other words, the proposed priority tri-structure has good scalability for large routing data.

On the other hand, Figure 14 shows the results of simulating a binary search structure with 112K entries and 225K entries. Performance comparisons are given in terms of worst case memory access number T _max , average memory access number T a, required memory size M , and number of extra nodes needed ( N _extra ). As shown, the proposed priority-based multi-bit trial exhibits the best performance in terms of the worst-case memory access times ( T _max ), and is based on the conventional binary search on range (BSR) and the priority-based proposed in the present invention. The multi-bit tris of Cs show the best performance in terms of average memory access times ( T a). In terms of the size of memory required ( M ), the BSR and the proposed priority based binary tris show the best performance.

On the other hand, the conventional BSR algorithm does not provide incremental update because it calculates the best matching prefix for each separation section in advance. On the other hand, the proposed algorithm can provide incremental update.

The above-described IP address retrieval method according to the present invention can be implemented in hardware in a packet relay apparatus. Since the IP address search in the packet relay apparatus must be processed in real time for all the input packets, it is appropriate to design and implement the present invention in a hardware structure to perform the IP address search in real time.

On the other hand, the IP address retrieval method according to the present invention can also be created by a computer program. Codes and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the program is stored in a computer readable media, and read and executed by a computer to implement the IP address retrieval method. The information storage medium includes a magnetic recording medium, an optical recording medium, and a carrier wave medium.

As described above, the present invention provides an effective IP address retrieval method and a packet relay apparatus using the same, which can reduce retrieval time and reduce the size of memory required.

According to the present invention, a new IP address retrieval method using a priority trie is provided. That is, the proposed algorithm is based on the tree structure, but empty inner nodes are replaced with priority prefixes. Thus, the search for the longest matching prefix (LMP) in the proposed algorithm is performed more efficiently because the input prefix ends earlier if it matches the priority prefix node than the normal prefix node. The results of the efficiency assessment show that the structured priority tree has very good performance in terms of memory, search speed and scalability required.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

Claims (13)

In the IP address retrieval method for packet forwarding, (a) extracting a destination IP address of the received packet; (b) retrieving a network address whose output port according to the network address is the longest match from the extracted destination IP address; And (c) designating an output port corresponding to the retrieved network address as an output port to which the received packet is to be sent; The searching is performed by using a priority-trie structure generated based on network addresses included in the routing table. The method of claim 1, Wherein the priority tri structure is constructed by allocating a long prefix to a higher level using empty internal nodes of a general tri structure such that longer prefixes are compared first. The method of claim 1, Wherein the priority tri structure is constructed by replacing each bean inner node of the general tri structure with the longest one of the prefixes belonging to the subtree whose bin inner node is the root node. According to claim 1, wherein step (b), (b1) comparing the extracted destination IP address with respective nodes of the priority tri structure; (b2) if the extracted destination IP address matches the general node, as a result of the comparison, the prefix of the matching general node is stored as the longest prefix matched so far, and the search continues; (b3) terminating the search if the extracted destination IP address matches a priority prefix substituted for an empty node or meets a leaf in the priority tri-structure; And (b4) If the search is terminated by matching the priority prefix, the corresponding priority prefix is returned to the network address that has the longest match with the extracted destination IP address; comprising the step of returning the longest prefix match with the network address to the longest match with the extracted destination IP address, IP address search method. In the method for building a routing table for IP address lookup, (a) storing a prefix representing a network address as nodes of a level corresponding to a length of each to construct a general tri-structure; And and (b) reconstructing the priority-based tri-structure structured so that the longer prefix from the general tri-structure is compared first. The method of claim 5, wherein step (b) comprises: Storing the longest prefix in the first empty node in the path of the general tri structure constructed in step (a) and marking the prefix stored in the empty node as a priority prefix in the empty tri-structure. Routing table construction method characterized in that iterate until the node disappears. A method of updating a routing table for IP address lookup, the method comprising: comprising the steps of: (a) creating a prefix to be deleted by the blank internal node when deleting a prefix from the network address of the first generation on the basis of the ranking tri (priority-trie) structures included in the routing table; And (b) reconstructing a priority-based tri-structure for the routing table at predetermined intervals to remove the empty internal node. The method of claim 7, wherein step (b), The longest prefix included in the routing table is configured to reconstruct a priority-based tri-structure by allocating a long prefix to a high level using the empty internal node so as to be compared first. A method of updating a routing table for IP address lookup, the method comprising: (a) When a prefix is inserted in a priority-trie structure generated based on network addresses included in the routing table, the prefix to be inserted coincides with the priority prefix and its length is equal to the priority prefix. Arranging the prefix to be inserted occupies the position of the priority prefix when it is longer than a length of the priority prefix; And and (b) arranging the prefix to be inserted occupy the position of the other priority prefix when the position of the prefix to be inserted is already occupied by another priority prefix. In the packet relay device for retrieving an IP address, A frame analyzer extracting a destination IP address of the received packet; And An address lookup that retrieves a network address whose longest output port matches a extracted destination IP address from a specified routing table and designates an output port corresponding to the retrieved network address as an output port to which the received packet is to be sent A processing unit, And the address lookup processing unit performs the network address search using a priority-trie structure generated based on network addresses included in the routing table. The method of claim 10, And the priority tri structure is constructed by allocating a long prefix to a high level using empty internal nodes of a general tri structure so that longer prefixes are compared first. The method of claim 10, And the priority tri structure is constructed by replacing each bean inner node of a general tri structure with the longest prefix among prefixes belonging to a subtree whose bin inner node is a root node. The method of claim 10, wherein the address lookup processing unit, The extracted destination IP address is compared with each node of the priority tri structure, and when the extracted destination IP address matches a general node, the prefix of the matching general node is stored as the longest prefix matched so far. And continue the search and terminate the search if the extracted destination IP address matches the priority prefix replaced by an empty node or meets a leaf in the priority tri-structure, If the search is terminated by matching the priority prefix, the corresponding priority prefix is returned to the network address that matches the extracted destination IP address for the longest time. And return a long prefix to a network address that matches the extracted destination IP address longest.

Images