KR100657107B1 - Method and apparatus for routing packet data - Google Patents

Abstract

A method and an apparatus for classifying packets are provided to simultaneously search protocol types and source/destination port numbers, store the shortest prefix with respect to various source/destination addresses having the same prefix, and perform the packet classification based on the stored prefix, thereby acquiring a fast packet classification time in comparison with existing one-dimensional processes. A method for classifying packets comprises the following steps of: searching protocol types and source/destination port numbers for packet classification in a router; searching source/destination IP(Internet Protocol) address based on a rule having the shortest address among addresses having the same prefix relation(S20,S40); performing sequential searching from a rule having high priority on rules satisfying both of the above search results, and searching a consistent rule; and classifying corresponding packets according to the final search result.

Description

Packet classification method and apparatus {METHOD AND APPARATUS FOR ROUTING PACKET DATA}

1 illustrates a routing system structure according to the present invention.

2 is a view showing an example of a port number search method according to the present invention;

3 is a view showing an example of a structure combining a port number and a protocol type in the present invention;

4 is a diagram showing an example of a table structure for storing information on a routing rule retrieved in the present invention.

5 is a flowchart of a packet classification method according to the present invention.

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

100: search unit 110: source port number and protocol memory

120: destination port number and protocol memory

130,170: origin rule memory 140,180: destination rule memory

150: source IP address memory 160: destination IP address memory

190: rule match table memory

The present invention relates to a method and apparatus for classifying a packet in a router. More specifically, the present invention simultaneously searches for the type of protocol and the source / destination port number when classifying packets in a router, and stores only the shortest prefix for several source / destination addresses having the same prefix. The present invention relates to a packet classification method and apparatus for improving packet classification speed and performance by preceding a search for a high priority rule.

As the services of the Internet are diversified, various and complex functions are required for Internet routers. One function is packet classification. Packet classification is to classify the packet flow into various types according to the information value of the packet and perform differentiated processing accordingly.

As is known, a router is a device that interconnects multiple telecommunication networks to send and receive data, and its function is basically a bridge, but the router selects a path at the network layer (third layer) of the OSI basic reference model. By doing so, the logical link control (LLC) protocol and the medium access control (MAC) protocol are interconnected with a plurality of information communication networks (LAN, etc.) different from each other. The bridge simply determines whether to pass the data, but the router interprets the protocol contained in the data and selects the optimal path to send the data.

The flow of a packet is defined by rules composed of several items of a packet header, and a set of rules determined by a specific application program is called a 'classifier'. When the rules constituting 'classifier' consist of a plurality of items, each item in the rule and the corresponding items in the packet are matched and the rule with the highest priority is selected, and the packet is selected from the selected rule. It is the process of processing as prescribed.

The packet classification function provides services such as packet filtering, traffic rate, limiting, policy based routing, accounting, and billing. Each item of a rule that constitutes a 'classifier' has various header items from the network layer header to the application layer header. Each entry uses either <value, length>, as used in IP address lookup, or restrictions such as 'greater than 1024', 'like 80', 'between 0 and 1024'. It also adds a wildcard operation that certain fields will not consider their values.

Conventional packet classification method searches the source address first and then the destination address to find a rule matching the input packet through the source address or destination address information. To this end, there are four methods that have been proposed.

The first method is to perform the search in order for the entire rule, which requires the number of O (N) searches for the N rules.

The second method is a sequential search that searches for a starting address and then searches for a destination address corresponding to a matching starting address. In this case, the search for each field is performed by applying a tree structure, but there is a disadvantage in that the backtracking should be performed due to the characteristics of the packet classification mentioned above.

The third way is to apply geometrical features to the rules. By dividing the range of each field and narrowing down the search, two fields are mainly used: the source address and the destination address. This method is difficult to extend to multidimensional retrieval.

The fourth method is to use empirical features. As a result of various approaches for packet classification, various methods are combined. In the case of existing methods having such characteristics, the performance is determined by the distribution or size of the rule.

This packet classification function has a disadvantage of high complexity and time since it has to search for several fields among header information of the packet. Conventional methods attempt to reduce the total search range by searching the source IP and destination IP addresses that have the most distribution among most rules, but the IP address search of packet classification has the disadvantage of having to trace back because of its characteristics or the number of searches. There was a downside to increasing.

In summary, conventionally, when classifying a packet in a router, the source / destination address is searched first, then the rule matching the input packet is searched through the searched address information, and the rule having the highest priority among the searched rules is selected. As soon as the packet classification process is performed. Therefore, since the search for several fields in the header information must be performed, the complexity is high, the classification processing time is long, and the number of searches is increased, thereby degrading packet classification performance.

SUMMARY OF THE INVENTION An object of the present invention is to provide a packet classification method and apparatus for improving packet classification processing performance by reducing the time for classifying input packets in a router, reducing complexity, and reducing the number of searches.

It is another object of the present invention to simultaneously search for the type of protocol and source / destination port number when classifying packets in a router, and store only the shortest prefix for several source / destination addresses having the same prefix. The present invention provides a packet classification method and apparatus for improving packet classification speed and performance by prioritizing a search for a high priority rule.

Packet classification method according to the present invention for achieving the above object, the step of retrieving the type of protocol and the source / destination port number for packet classification in the router; Retrieving a source / destination IP address based on a rule having the shortest address among addresses having the same prefix relationship; Searching for a matching rule by performing a sequential search starting from a rule having a high priority with respect to rules satisfying all of the search results; And classifying the packet according to the final search result.

In addition, in the packet classification method according to the present invention, the rule derived from the combination of the source port number and the destination port number is characterized in that only the rules satisfying both cases are stored through the 'AND' process.

In addition, in the packet classification method according to the present invention, a rule that satisfies both cases with respect to a rule derived from a combination of the source port number and the destination port number is stored as a 1-bit rule correspondence table for each rule. It features.

Also, in the packet classification method according to the present invention, the step of searching for the source / destination IP address may be performed by comparing the source / destination IP address of the rules in order to store only the rule having the shortest address among several addresses having a prefix relationship. Characterized in that.

In addition, in the packet classification method according to the present invention, for the packet classification, a rule having a high priority for the remaining items except for the items related to the searched source port number, destination port number, protocol type, source IP address, and destination IP address. To perform a sequential search.

In addition, the packet classification apparatus according to the present invention for achieving the above object, means for storing the type of protocol, the source and destination port number; Means for storing origin and destination addresses and corresponding rules; For the input packet, the protocol type, the source and destination port number, the source and destination address, and the corresponding rule are retrieved and stored, and the items related to the source and destination port number and protocol type, the source and destination address for the retrieved rules. Search means for searching for a matching rule for all items by sequentially searching from a rule having a high priority to the remaining items except for the above; Rule matching table storage means for storing the search results; Characterized in that comprises a.

In the packet classification apparatus according to the present invention, the means for storing the source IP address and the destination IP address is TCAM (Ternary Contents Address Memory).

In the packet classification apparatus according to the present invention, the protocol type, the source / destination port number, and the rule memory are composed of one memory, and the respective data are divided and recorded in the set area.

Hereinafter, a packet classification method and an apparatus thereof according to an embodiment of the present invention will be described with reference to the accompanying drawings. The present invention proposes a structure that reduces the overall search range by dividing the search for each item unlike the conventional approach.

1 shows a basic structure of a packet classification system in a router according to the present invention. As expressed herein, the present invention divides a search step to facilitate a search for each item. FIG. 2 illustrates the concept of port number search. As can be seen, if the port number of each rule is expressed in a straight line, it is as shown in FIG. 2, and the corresponding rules are included in each range. Since the present invention proposes a combination of protocol type and port number, the range of port numbers is arranged according to the protocol type. 3 simply shows how a combination of port number and protocol type is stored in the memory. 4 shows a state of a rule correspondence table for storing information on a rule retrieved at each step. 5 is a flowchart illustrating a packet classification method of the present invention to which the above characteristics of packet classification are applied.

1 is a diagram illustrating a routing system structure according to the present invention, a first memory 110 for storing a type of a source protocol and a source port number, and a second for storing a type of a destination protocol and a destination port number. A first rule memory 130 for storing the memory 120, a rule corresponding to the type and port number of the source protocol, and a second rule for storing a rule corresponding to the type and port number of the destination protocol. Memory 140, third memory 150 for storing the source IP address, fourth memory 160 for storing the destination IP address, and third rule for storing rules corresponding to the source IP address Memory 170, a fourth rule memory 180 for storing a rule corresponding to an destination IP address, a rule correspondence table memory 190 for storing a search result, and an input packet. It includes a search unit 100 for searching the type of the protocol, the source / destination port number, the source / destination IP address.

As shown in Fig. 1, the present invention is a memory 110, 120 for storing a protocol type, a source port number, and a destination port number, and a memory 150 for storing two SRAMs, a source and a destination IP address. Two 160 TCAM (Ternary Contents Address Memory), rule memory for storing the rules corresponding to each entry four SRAM (130, 140, 170, 180), one for storing the search results The rule correspondence table memory has one SRAM 190.

Since each of the memories 110, 120, 130, 140, 170, 180 and 190 requires a small size, the total memory size is not large and the remaining memories 110, 120, 130, 140, 170, 180 and 190 except for the TCAM 150 and 160 can be used as one memory. That is, the type of protocol, the source / destination port number, and the rule memory are composed of one memory and can be used by dividing each data into a predetermined area and recording the same.

As is known, Guy TCAM is a memory device that accesses the actual data as an address when storing or reading data in the memory device. It is more powerful and faster than general address memory.

The first memory (SRAM) 110 stores a result of searching for a source port number and a source protocol type, and the rule for the entry includes a first rule memory ( Rule Memory-Src., SRAM) 130. The second memory (SRAM) 120 stores a result of searching for a destination port number and a destination protocol type, and the rule for the entry is a second rule memory. Dest., SRAM) (140).

The third memory (TCAM) 150 stores a source IP address and the rules for the corresponding entry are stored in the third rule memory 170 (Rule Memory-Src., SRAM) 170. The fourth memory (TCAM) 160 stores a destination IP address, and the rules for the corresponding entry are stored in the fourth rule memory 180 (RRAM).

Search unit 100 retrieves the protocol type and port number of the source and destination, and stores it in the rule memory (130,140) and the rule from each combination of the source port number and the destination port number in the case of 'AND' process in two cases Extract only the rules that satisfy all of them and record them in the rule correspondence table memory 190. In addition, the search unit 100 performs a search for a source IP address and a destination IP address and stores them in the corresponding memory 150 and 160, respectively, and the rules are stored in the corresponding rule memory 170 and 180, respectively. For the selected rules, a search is performed sequentially from the high priority rule to the remaining items except for the source, destination port number, protocol type, and the source and destination address stored in the TCAM. If found, the packet classification process is terminated.

As shown in Fig. 1, the present invention divides a search step into an easy search for each item. The first is to search for a protocol type and a source and a port number of a destination, and the second is to search for an address of a source and a destination. The third step is to complete packet classification through priority-based sequential search based on the first and second search results.

In the step of searching for the first protocol type and the source and destination port numbers, two memories 110 and 120 are used to store the combination of the source port number and the protocol type and the combination of the destination port number and the protocol type. Done.

2 shows the concept of port number search. Port number retrieval in packet classification is characterized by range retrieval. Therefore, if the port number of each rule is expressed in a straight line, it becomes as shown in FIG. 2 and the corresponding rules are included in each range. Since the present invention proposes a combination of protocol type and port number, the range of port numbers is arranged according to the protocol type. The source port numbers represented by the straight lines are stored in the first memory 110 in the order, and the destination port numbers are stored in the second memory 120 in the order.

3 simply shows how the combination of the port number and the protocol type is stored in the memory. The protocol type and port number are shown with the rule memory pointer value, and the rule memory pointer is the rule memory 130. It is used as the pointer value of 140.

The rule from the combination of the source port number and the destination port number in the protocol type and the source / destination port number search step remains the rule that satisfies both cases through the 'AND' process. The structure is recorded in the rule correspondence table memory 190 of the structure. This table has 1 bit for each rule and stores the total number of rules.

The packet classification of the present invention performs a sequential search for the rules recorded in the rule correspondence table after completing the second step described below.

The second step of the present invention is the process of performing a search for a source IP address and a destination IP address. In the case of the IP address search, the TCAM 150, 160 is a process of finding the longest value (LPM) that matches the address of the input packet. In this case, the desired value can be found by one memory search. However, in case of packet classification, the method that finds the highest priority rule rather than the longest matching value cannot be used immediately. Therefore, in the final step, the rules having higher priority for the remaining items except for the source and destination port number, the protocol type, and the source and destination address stored in the TCAM 150, 160 are sequentially selected for the selected rules. Perform a search.

The rule memories 130, 140, 170, and 180 of FIG. 1 are used to store port numbers, protocol type combinations, or various rules of a source address. Rule memory is used because a combination or address can have multiple rules.

FIG. 5 is a flowchart of a method of configuring a source address lookup table applying the characteristics of packet classification as described above. The method is similarly applied to a case of configuring a destination address lookup table.

Summarizing the overall process, the source (destination) IP addresses of the rules are sequentially compared, and only the rule having the shortest address among the various addresses having the prefix relationship is stored in the TCAM 170 (180). By using this, it is possible to reduce the range in which the sequential search is to be performed by one TCAM 170, 180 search. The address stored in the TCAM 170, 180 stores information about the remaining rules in the rule memory 170, 180 as described above. After searching for the source address and the destination address, an AND process is performed on the results of the first step and the results of the second step to find a rule that satisfies both conditions. Lastly, the selected rules are sequentially searched from the high priority rules for the remaining items except for the source and destination port numbers and protocol types, and the source and destination addresses stored in the TCAM. Since the sequential search starts from the higher priority rule, the packet classification process is terminated when a matching rule is found for all items.

A method of constructing a source (destination) address lookup table of FIG. 5 will be described in more detail.

The first step S10 is an initialization process, in which i = 0 and j = 0 are set, 'A' is set as the i th prefix, and 'B' is set as the j + 1 th prefix.

The second step S20 is a determination step for searching for an IP address having the same prefix relationship with respect to the input packet, and compares the sizes of 'A' and 'B'. That is, it is determined whether or not the IP addresses currently compared with have the same prefix. If 'A = B', the process proceeds to the third step S30 to perform the fourth step S40 and the fifth step S50 to search for and store the shortest address among the IP addresses having the same prefix relationship. If it has a different value by comparing 'A' and 'B', and proceeds to a sixth step (S60).

The third step (S30) is a step for determining which of the 'A' and 'B' has a shorter IP address, if the length of the 'B' is longer than the length of the 'A' fourth step (S40) If the length of 'B' is shorter than the length of 'A', go to the fifth step (S50) and save 'A' in the starting rule memory. By exchanging the 'B' values with each other, the fourth step S40 is executed to store the shorter IP address 'A' in the source / destination rule memory 170 and 180.

The sixth step S60 is a step of determining whether the j value is smaller than the number N of all prefixes. In this step, if j <N, the process proceeds to the seventh step S70 to increase the j value, set it to 'B = j-th prefix', and repeat the process from the second step S20 to increase the j value while increasing the j value. The shortest IP address is retrieved and stored until the total prefix number N is reached.

When j≥N is repeated by repeating the above process, the process proceeds to the next eighth step S80 to increase the value of i and set 'A = i prefix' and 'B = j prefix', and then By repeating from step 2 (S20), while increasing the value of i, the process of searching for and storing the shortest IP address until the value of j reaches the total number N of prefixes.

In this case, only the rule having the shortest address among IP addresses having the same prefix relationship can be searched and stored in the rule memory 170 or 180.

Next, the 'AND' process is performed on the rules of the first step (protocol type and search result of source and destination port number) and the second step (result of source and destination IP address search). Find a rule that satisfies all four conditions.

Lastly, for all selected items, search is performed sequentially from the high-priority rules for the selected items except for the source and destination port number and protocol type, and the source and destination address stored in the TCAM. If a matching rule is found, the packet classification process according to the present invention is completed by classifying the input packet based on the rule.

Therefore, it is possible to reduce the overall search range by simultaneously searching for the protocol type and port number, and to search and search for all matching rules by storing only the shortest prefix for multiple source (destination) addresses with the same prefix. It is possible to overcome the limitations of conventional packet classification with a large number of times.

In particular, the memory burden can be reduced by storing information about the rule that the input packet matches by using a small size of memory that has one bit allocated to a rule. By finding them, the number of sequential searches can be reduced.

In addition, during the sequential search, the search for the rule having the highest priority is performed first, thereby reducing the total number of searches.

In the present invention, the protocol type and the port number of the source / destination destination are simultaneously searched, and only the shortest prefix is stored for several source / destination addresses having the same prefix, and the packet classification is performed based on this. Fast packet classification time can be obtained compared to the dimensional methods. In addition, the present invention is suitable for both hardware and software because it uses memory efficiently and uses a small size of memory. According to the present invention, it is possible to deal with the need for a complicated packet classification that increases with the structure that is not affected by the number of rules or items.

Claims (9)

Retrieving the type of protocol and source / destination port number for packet classification in a router; Retrieving a source / destination IP address based on a rule having the shortest address among addresses having the same prefix relationship; Searching for a matching rule by performing a sequential search starting from a rule having a high priority with respect to rules satisfying all of the search results; And classifying a corresponding packet according to the final search result. 2. The method of claim 1, wherein the rule derived from the combination of the source port number and the destination port number stores only the rules satisfying both cases through the 'AND' process. The packet according to claim 2, wherein a rule satisfying both cases with respect to a rule derived from a combination of the source port number and the destination port number is stored as a 1-bit rule correspondence table for each rule. Classification method. The method of claim 1, wherein the searching of the source / destination IP address comprises comparing the source / destination IP address of the rules in order and storing only the rule having the shortest address among several addresses having a prefix relationship. Packet classification method to do. The method according to claim 1, wherein the packet is classified in order from the high priority rule to the remaining items except for the searched source port number, destination port number, protocol type, source IP address, and destination IP address. Packet classification method, characterized in that performed. Means for storing the type of protocol, source and destination port numbers; Means for storing origin and destination addresses and corresponding rules; For the input packet, the protocol type, the source and destination port number, the source and destination address, and the corresponding rule are retrieved and stored, and the items related to the source and destination port number and protocol type, the source and destination address for the retrieved rules. Search means for searching for a matching rule for all items by sequentially searching from a rule having a high priority to the remaining items except for the above; Rule matching table storage means for storing the search results; Packet classification apparatus comprising a. 7. The packet classification apparatus of claim 6, wherein the rule correspondence table has one bit for each rule and is stored as many as the total number of rules. 7. The apparatus of claim 6, wherein the means for storing the source IP address and the destination IP address is TCAM (Ternary Contents Address Memory). 7. The packet classification apparatus according to claim 6, wherein the type of protocol, the source / destination port number, and the rule memory are composed of one memory and each data is divided and recorded in a set area.

Images