TW508926B - Method and system for efficient routing table compression and fast routing lookups - Google Patents

Abstract

A system for IP routing lookup in a routing table, comprising entries of arbitrary length prefixes with associated next-hop information in a next-hop table, for determining where IP datagrams are to be forwarded, and data structure means for representing the set of dominating points in the IP address apace for each prefix in the routing table, said means comprising three levels of compressed pointer arrays (CPA), each including a chunk and an array of associated pointers, wherein each pointer comprises a next hop index or a pointer to the next level of chunk.

Description

508926 Printed by A7 B7, Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs V. Description of the Invention (I) [Field of the Invention] The present invention is related to the method and system of IP path selection lookup table in the path selection table. The table contains any length prefix items related to the sub-hop information in the sub-hop table to determine where the IP network packets should be forwarded. [Background of the Invention] The Internet is a collection of interconnected roads, and each of the constituent networks still retains its identity and requires specific mechanisms to communicate across multiple networks. These constituent networks are referred to as subnetworks. Each subnet in the Internet can support communication operations between devices connected to the subnet. In addition, the sub-networks are connected to each other through a device called Γ network connection units (IWUs). There is a specific type of IWU called a router that can be used to connect two networks that are similar or very different. Routers use Internet Protocol (IP), which represents each router and each host on the network. IP can provide connectionless or network packet services between sites. Generally, a route selection table is established in each station and router to complete the route selection operation. For each possible destination network, the table can indicate the IP network packet to be sent below. A router. The IP router performs a path selection lookup table in the path selection table to determine where the IP network packets should be forwarded. The result of this operation is a secondary jump on the path to that destination. Fortunately, an item in the path selection table has an arbitrary length prefix related to the sub-jump information. 3 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) I ---- I — II — I ---— IIII ^ 1 ------ t I (Please read the back first (Notes on this page, please fill in this page) 508926 A7 B7 V. Description of the Invention (z) The path selection table lookup task must find the path selection item with the longest matching prefix. Due to the existing delay and complexity of the IP path selection and table lookup operation, the previous technical solutions have actually become complicated and unwilling to be applied. Most of the current IP router designs use caching technology, among which the path selection items that have recently used the destination address will be maintained in the cache. However, this technology depends on the sufficient positioning of the traffic to enable the cache to have a high hit rate, and the cost of path selection and table lookup can be shared among many packets. These cache technologies have worked well in the past. However, due to the rapid growth of the Internet today, the size required for address caching has increased, making hardware caching uneconomic. IP routers are designed to use specific purpose hardware for IP processing. This is an inelastic solution. Any IP format or protocol change will cause the design to fail. Therefore, software flexibility and high-speed performance improvement of general-purpose processors will be more suitable for this solution. · [Description of the present invention] Therefore, an object of the present invention is to provide an improved 1p path selection table lookup method and system, so as to perform a complete path selection table lookup operation for each 1p packet up, and the method and system are sufficient Many of the aforementioned inconveniences can be overcome. Another purpose is to increase the speed of path selection look-up table with traditional microprocessor. Yet another object is to minimize the lookup time in the forward table. 4 This paper size applies to China National Standard (CNS) A4 (210 x 297 public love) " " " " " " ". (Please read the precautions on the back before filling this page) Binding- -------- · 508926 A7 B7 V. Explanation of the invention ()) Another object of the present invention is to provide a data structure that can be completely adapted to the cache memory of a conventional microprocessor. Therefore, it is possible to greatly speed up the memory access in a stepwise manner than by resident the data structure in a memory composed of, for example, a slower DRAM. The disclosure purpose can be achieved by the IP path selection table lookup method and system consistent with the present invention. The system is a data binding that can represent many large path selection tables in a simplified form and uses only a few memory references. It can be searched quickly. One advantage of the present invention is the efficient implementation of table lookup, which results in better instruction cache behavior and more stable table lookup costs. [Brief description of the drawings] In order to explain the present invention in detail, the advantages and characteristics of the preferred embodiments of the present invention will be detailed in the following. The accompanying figures are provided with reference numbers in parallel. A schematic diagram of the design of a router; FIG. 2 is an explanatory diagram of a direct sub-hop encoded into a digital character at level 1 of a data structure according to the present invention; FIG. 3 is a level 1 of a data structure according to the present invention. An explanatory diagram of the direct secondary jump encoded as an index; FIG. 4 is an explanatory diagram of the index of the CPA that is encoded into digital characters and points to level 2 according to the present invention, and the index of the level 1 index in the data structure, this paper Standards are applicable to China National Standard (CNS) A4 specifications (210 X 297 mm) (Please read the precautions on the back before filling out this page) Packing ---- 11 --- Order ---- # · Ministry of Economy Intellectual Property Printed by the Bureau's Consumer Cooperatives 508926 A7

______BL V. Description of the invention (40 FIG. 5 is an explanatory diagram of direct sub-jumps encoded into digital characters at level 2 of the data structure according to the present invention; FIG. 6 is at level 2 of the data structure according to the present invention FIG. 7 is an explanatory diagram of the direct sub-hop encoded as an index; FIG. 7 is an explanatory diagram of the index of the CPA directed to level 3 and the index of level 2 in the data structure according to the present invention; 8 is an explanatory diagram of direct sub-jumps encoded into digital characters at level 3 according to the data structure of the present invention; FIG. 9 is sub-jumps encoded as an index at level 3 according to the data structure of the present invention FIG. 10 is a high-level view of a method for generating a data structure device according to the present invention, so as to represent a set of dominant points for each prefix in an IP address space in a path table; FIG. 11 is According to the present invention, a block diagram of a first embodiment of a system hardware layout for performing a path selection lookup operation in a path selection table is shown. FIG. 12 is a diagram for performing the path selection in a path table according to the present invention. Figure 2 is a block diagram of a second embodiment of a system implementation layout of a look-up table; FIG. 13 is an explanatory diagram of digital characters used as a third embodiment of the system hardware for performing a path look-up table in a path table according to the present invention; 14 is a block diagram of a third embodiment of a hardware implementation layout of a system for performing a path lookup table in a path table according to the present invention; FIG. 15 is a 32-bit level 1 digital word used for software implementation of the present invention The explanatory diagram of Yuan. 6 This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 mm) ~ (Please read the precautions on the back before filling out this page) Install i I 1

«— — III Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 B7 V. Description of the invention (s) [Explanation of component symbols] 1. Network interface interface 2. Network interface interface 3. Fronthaul engine 4. Network Road processor 5. Connection facilities / [Detailed description of the present invention] Referring now to FIG. 1, this is a router, which includes multiple network inbound interfaces 1, network outbound interfaces 2, fronthaul engine 3, and a network. The processor 4 is connected to each other via the connection facility 5. The inbound interface 1 sends the packet header to the fronthaul engine 3 via the connection facility 5. The fronthaul engine 3 then decides which outgoing interface 2 the packet should be sent to. This information is sent back to the incoming interface 1, and the person forwards the packet to the outgoing interface 2. The only job of the fronthaul engine 3 is to process the packet header. All other tasks, such as participation in routing protocols, resource reservation, packet processing requiring special attention, and other administrative tasks, are the responsibility of the network processor 4. Each fronthaul engine 3 uses a routing table and a regional version of the fronthaul table, which are downloaded from the network processor 4 and stored in the storage device of the fronthaul engine 3 to make routing decisions. It is not necessary to download a new forward list each time the path selection is updated. The path selection update operation may be recurring, but because the path selection agreement takes some time to complete, the front-end table is slightly increased and does not need to be updated more than once per second. This website is 7 packs (please read the precautions on the back before filling this page) Order -------- Printed by the Consumers' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs This paper applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 508926 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention () The processor 4 needs a dynamic path selection table which is designed to quickly update and quickly generate a prequeue table. On the other hand, these forward tables can also be optimized according to the lookup speed and need not be dynamic. In order to reduce the table lookup time to the minimum, two parameters, that is, the number of memory accesses required during the table lookup process, and the size of the data structure must be minimized at the same time as the data structure in the forward table. It is important to reduce the number of memory accesses required during the table lookup process, because memory access is quite slow and often becomes a bottleneck in the table lookup process. If the data structure can be made small enough, it fits perfectly into the cache memory of a conventional microprocessor. Compared to the fact that the data structure needs to be resident in a memory composed of relatively slow DRAM, this means that such memory access can be accelerated by orders of magnitude, as in the case of the Patricia tree structure. However, if the prequel table is not completely fit into the cache memory, if most of the prequel tables can reside in the memory, the other is still beneficial. The regionality of the traffic model can retain the most commonly used data structure fragments in the cache, so most table lookup operations are still fast. In addition, it may be more practical to use fast SRAM as the small amount of additional memory required. SRAM is more expensive, and the more expensive it is, the faster it becomes. Therefore, for a given cost, if the amount is less, a faster SRAM can be applied. The second design goal is that the data structure should require fewer instructions. During the table lookup process, each item should be aligned as naturally as possible to avoid expensive instructions and verbose bit fetch operations. 8 (Please read the precautions on the back before filling out this page) Installation specifications · This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs Consumer Cooperatives 508926 A7 B7 5 2. Description of the invention (^) The road selection table lookup table is a type of prefix database, in which each prefix is related to the secondary jump information displayed according to the index form referred to in the secondary jump table. The prefix is basically a one-digit string containing 0 to 32 bits, which is the prefix length. The first bit is numbered 31 and the second is numbered 30. Continue until the end of the bit string. Take the prefix of length 5 as an example, that is, 檩 is shown as bits 31, 30, 29, 28, and 27. This means that the prefix will match any of the bits 3 彡, 30, 29, 28, and 27 of its IP address to the IP address of the corresponding bit of the prefix. Since only the first (most significant) five bits of the prefix are marked, the bits 0 to 26 of the HP addresses are irrelevant. A prefix of length 0 can be matched to any IP address. The result of the path selection lookup table is generally the index 値 related to the longest matching prefix in the path selection table. For example, for an IP address starting with 101…, bit 31 is the leftmost bit, and they are combined in the prefix 10 * and the prefix 101 *, where * indicates the end of the prefix. However, the prefix 101 * is longer than 10 *. Therefore, the better result of the path selection lookup table between these two prefixes is the index 値 related to the prefix 101 *. The prefix can also be IP address space. · The size of 4294967295 is represented by a range of 2i, where i is a non-negative integer. The starting point of this range must be an integer multiple of 21. Otherwise, this is not a prefix. This is called "square power." For example, the prefix 10 * starts at 2147483648 and ends at 3221225472. The size of the range is 1073741824, which is equal to 23G. The starting point is 2 * 23 (). 9 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)

508926 A7 B7

Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economics. Inventive Note (g) Each prefix is a range with a power of two, and vice versa. Therefore, we do not need to distinguish between the range of the power of second power and the difference between the two. The prefix length is defined as the number of bits marked, and the size of the prefix can be defined as the size of the IP address space range covered by the prefix. The binary representation of the starting point of the prefix can be calculated by zero-filling the bit string expression to the right until the length of the bit string becomes 32. The prefix size of length i is set to 232 ". In the path selection table, there are often many large prefixes that will completely cover the smaller ones. To reduce the amount of memory required to represent the path selection table, Redundant information can be removed by generating an auxiliary path selection table that contains exactly the same path selection information but does not have a covering prefix, that is, prefix division. Covering prefixes can be recursive Divided in half until the size of the remainder divided by the large prefix is equal to the size of the smaller prefix covered by it, which means that they have indeed been removed. Then these are left from the covering prefix Any method except the prefix can be discarded. This method can be applied repeatedly starting from the smallest prefix, and then continuing to the larger prefix as long as there is a larger prefix that will cover the smaller prefix. The result is one. Groups can form a set of prefixes for IP address space partitioning. For example, consider the following path selection table: prefix first IP address final IP address secondary hop index 値 氺 0 · 0 · 0.0 255.255.255.255 1 0 * 0 · 0 · 0 · 0 127.255.255.255 2 001 * 32.0.0.0 ^ ~ 63.255.255.255 3 10 * 128.0.0.0 191.255.255.255 4 The prefix 001 * will be covered by the prefix 0 *. Therefore, the 10 paper sizes with larger prefixes will apply the Chinese National Standard (CNS ) A4 specification (21〇X 297) (Please read the precautions on the back before filling in this page)

Installation——I order --- 砉 · 508926 A7B7 V. Description of the invention (?) Divide it; i: Jieli County 00 * 0 · 0 · 0.0 one one ^ 63.255.255.255 2 01 * 64.0.0.0 one _ 127.255 .255.255 2 By dividing 〇〇 *, m to 000 * 0.0.0.0-31.255.255.255 2 001 * 32.0.0.0 __ 63.255.255.255 Remove 01 * 64.0.0.0 one one ^ 127.255.255.255 2 Therefore the path selection table The first IP bit in the end of the word here, the last IP address, and the second hop index 値 氺 0 · 0 · 0 · 0 a__ 255.255.255.255 1 000 * 0 · 0 · 0 · 0 ^ a ^ a 31.255.255.255 2 001 * 32.0.0.0 __ 63.255.255.255 3 01 * 64.0.0.0 _ one 127.255.255.255 2 10 * 128.0.0.0 one 191.255.255.255 4 The prefix * will still cover the first combination of medical treatment 1 "to divide by 0 * 0 · 0 · 0 · 0 127.255.255.255 _ 1 * 128.0.0.0 255.255.255.255 1 t. -------- Difficult to write this page) Borrow and recurse to cut 1 * to get 10 * 128.0.0.0 191.255 .255.255 Remove 01 * 192.0.0.0 255.255.255.255 1 Finally, you can get the following path selection table, in which the prefixes are indeed customizable! Suspect. After all the wisdom to discuss the bombers!-肖 费 合 Cha i. Prefix first IP address final > Address Secondary Jump Index 値 000 * 0.0 · 0.0 31.255.255.255 2 001 * 32.0.0.0 63.255.255.255 3 01 * 64.0.0.0 127.255.255.255 2 _ 10 * 128.0.0.0 1 191.255.255.2551 4 11 * 192.0. 0.0 255.255.255.255 1 The path selection table can also be expressed in the form of a binary tree, where each node can have zero, one, or two children. A node with zero children is called an external node, or Leaf point, and a paper with one or two 11 paper sizes applicable to Chinese National Standards (CNS) A4 (210 X 297 mm) 508926 A7 B7 V. Description of the invention (/ 〃) The nodes of the children are called Are internal nodes. In this tree, the subjump index 次 of the prefix 01 * is stored in the left (0th) child and the right (Ist) child of the root point. The sub-jump index 値 of the prefix * will be stored at the root point. The tree-like representation of the limited prefix division derived from the application of the aforementioned method has the following important properties: • all internal nodes have exactly two children • the only external node contains sub-hop information It is called a tree-like perfection method. The tree-like representation of its original path selection table can be constructed as a complete binary number. Represent the prefix segmentation as a set of prefixes. Since the prefix segmentation operation will generate a set of prefixes, each of the prefixes will be related to the sub-hop information and the IP space can be appropriately segmented. The term segmentation is indeed sufficient with a set of starting points (the first IP address) of these prefixes. Therefore, the information required to present the form in the example above can be as follows. Please read the memorandum first.

Fill in 1 Sheet of nested book I Order the consumer spending agreement of the Intellectual Property Bureau of the Ministry of Economic Affairs. Print the first IP address jump index 値 0 · 0 · 0 · 0 2 32.0.0.0 3 64.0.0.0 2 128.0. 0.0 4 192.0.0.0 1 Find the nearest dominant point in the prefix set. Given a set of points in the real number set, here are two points defined by a distance above the norm L0〇 (that is, L-infinity). For lim Bu. odpi-p / y / k = jpi-pj (I-distance will be equal to Euclidean distance in one dimension). In addition, we can apply 12 Chinese paper standards (CNS) A4 specifications (210 x 297 public narrative) # 508926 A7 B7 V. Description of the invention (il) Say point P2 leading point Pl (or point P1 is determined by point P2 (Dominant) if and only if ρι $ ρ2. When a path selection table is represented as a set of starting points of a set of prefixes, that is, as described above, the problem of the path selection table is reduced to the problem of finding the nearest dominant point P. For example, in order to perform the path selection table lookup operation of the IP address 172.16.0.20 in the table above, only the leading points 〇〇〇〇〇〇, 32.〇 · 〇 · 〇, 64.0.0.0, 128.0. In the set of 0.0, the point with the shortest Loo distance from ι72 · 16 · 20 · 20 can be found. Obviously, this point is 128 · 〇 · 〇 · 〇, and its path selection lookup table result is the sub-hop index 4. XTC data structure XTC is a data structure that uses three levels of "compressed index array" to effectively represent the set of dominant points. Each compression indicator array (CPA) contains a block and an array related to a 20-bit index and / or a 15-bit sub-hop index frame. The CPA constituting level 1 represents an uncompressed index array (bits 31 ·· 16 of the IP address) with a size of 65536 pointers, and the CPA constituting levels 2 and 3 each represent an index of 256. Compression indicator array and dominant point set Compression indicator array is represented by a block and many indicators. The block contains digital characters, and each digital character represents a 16-bit character representing a recessive bit array of the block. The size of the recessive bit array is equivalent to the largest number of indicators in the CPA. In the recessive bit array, the bit with index 値 i can be set if and only if a prefix begins at the position i. The so-called starting point, meaning 13 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 x 297 mm) (please read the precautions on the back first to write this page) -Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 B7 V. Invention Description (A) is the inspected bits in the leading point (3L.16 of class 1, 15..8 of class 2 and class) 3 · 7 · 0), when indexed into the CPA, will be equal to i. The recessive bit array is represented by 16 recessive bit masks. There are three types of bit masks: • Empty bit masks without a bit set • Bit masks with only one bit set • Bit masks with two or more bit sets by The power of the second power ^ A bit mask with a bit set indicates that the initial of a starting point in the block is a multiple of 16. He can completely cover the entire bit mask, as well as all the empty bit masks that follow. For example, if the first jump index of a prefix starts at position 64 = 4 * 16 (indicating that the leading point of the prefix is 16) and ends at position 127 = 7 * 16 + 15, the bit mask 4 Only the first set of bits is available, and masks 5, 6, and 7 are empty. For bit masks that have a set of bits and then follow the empty set of bits, the secondary skip index 値 can be directly encoded as a digital character. Each of these digital characters can represent 16 indicators required when the indicator array is not compressed. In the example above, the numeric characters 4..7 will be the same. The bitcoin mask printed by the employee's consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, which has more than two bit sets, indicates that the prefix of a bit mask is not completely covered. For such digital characters, the starting point and the ending point are located in the same bit mask. By the nature of the second power, the number of different types of bit masks with a length of at least one bit set and a length of 21 can be given by the following recursive equation: A (0) = 1 14 This paper scale applies Chinese national standards (CNS) A4 Regulations (210 X 297 mm) 508926 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of Invention (ο) A (i) = 1 + A (il) This means that for i = 1, 2 and 3, the bit masks can be as shown in the following table. For i = 4, there are 677 possible situations, which are not explained here. (See the attached table on page 13.) Therefore, by the power of 2 power, there will be 676 different 16-bit masks with more than two bit sets. Such a bit mask and its corresponding indicator can be represented by a combination of a digital character and a list of indicators related to the digital character. Each indicator consists of a 15-bit jump index, or it consists of a 20-bit indicator that is oriented towards the next level of CPA. Initially, each bit set in the bit mask will have an index. However, in the XTC algorithm, the number of indicators (and bits) will increase slightly during the encoding process, thereby reducing the complexity as in the FTC algorithm. In FTC, 10 bits are used to encode the bit mask type as an index frame into a map. This map requires memory access, which slows down the lookup of the table. To improve its speed, XTC uses a similar encoding method of 2 bits instead of 10 bits. In addition, the mapping table is no longer needed at all, thus significantly reducing the number of memory accesses. By checking the set of prefixes whose starting points are located in the same bit mask, it can be identified that the size of the prefix is indeed the shortest (for prefixes whose starting point and ending point are in the same bit mask) In terms of size, it is · 1). Each prefix in the set can be 1, 2, 4, or 8. If the size is 16, he can completely cover the bit mask, and can apply the aforementioned direct encoding operation (the bit mask happens to be a bit set). Any paper with a size greater than the shortest of 15 sheets is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm). ^ —----- ^ ------ (Please read the notes on the back first (Fill in this page again) # Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 ____B7___ 5. The initials of the size of the invention description (wow) can be repeated in half until all the prefixes are the same size. Whenever the prefix is divided, a zero bit (overlap) is set in the bit mask, and a copy of the index related to the set of bits closest to the left end of the dialed bit is' directly Added after the duplicated indicator in the indicator list. Each split operation will increase the number of indicators one by one. After each step, you can get the situation shown in the following table. Digital description of large and small bit masks 8 1000000010000000 3 Two prefixes, each prefix will cover half of the bit mask 4 1000100010001000 2 Four prefixes, each prefix will cover a 1/4 bit mask 2 1010101010101010 1 18 prefixes, each prefix will cover 1/8 of the bit mask 1 11111111111111111 0 16 or more prefixes, each element of each mask can correspond to one or more prefixes here Instead of processing 676 different kinds of bit masks like the FTC algorithm, only four kinds of bit masks need to be processed. If the number of bits initially set is 5, it can be easily shown that the number of additional indicators introduced under this law is 11. Therefore, in the worst case, the number of indicators required by XTC notation will increase by a factor of 16/5 compared to FTC notation. The experiment shows that the general increase is a factor of two. In the lookup table, a part of the IP address is used to index to the CPA. The high 12 (at level 1) or 4 (other) bits are the index of the digital characters. The lower (that is, the least significant) 4 bits will be used to index to the bit mask of the digital character (assuming that the digital character does not include the coded direct next hop 16) This paper standard applies Chinese national standards (CNS) A4 specifications (210 x 297 mm) .—— 1-— J ^ T -------- 1 (Please read the notes on the back before filling this page) 508926 A7 B7 V. Description of the invention (6) If you jump). By indexing to a bit mask, you can either index a bit that has been set to (out of the set bits) to capture an indicator associated with that bit, or index an unset bit . If the bit is not set, then an indicator related to the set bit closest to the left end of the indexed bit needs to be retrieved. That is, as mentioned above, there is an index list related to each bit mask / digital character. These indicator lists can contain 2, 4, 8 or 16 indicators. The bit mask is encoded in the previous table as described above. Among the set bits, the index closest to the left end of the bit set (and the index of its corresponding index in the index list) can be obtained by a simple shift operation. Calculated. In other words, when looking up the table, the lower 4 bits that are used to index into the bit mask will be shifted to the right by the "digit" indicated bits. It can be compared with the expensive mapping table lookup operation used in the FTC algorithm. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs In the XTC data structure, this block contains a fixed-size array of digital characters (the size of which depends on the class). In order to reduce the number of bits required to indicate the "first" index index 値 (the beginning of the index list related to the digital character) related to each digital character, the indicators in the XTC data structure Will be stored in φ of the same array as the block. After the last digital character of the block, the indicator array related to a block is stored immediately. That is, as mentioned above, the index array includes a 20-bit index and / or a 15-bit sub-hop index. This is an array represented by a 16-bit integer. The most significant bit indicates whether the lower 15 bits indicate a sub-jump index, or the lower 15 bits of the 20-bit index. In order to show a 20-digit indicator that points to the next level, you can use the $ digits of the digital character to 17 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 508926 A7 B7 5 2. Description of the invention (屮) It indicates the upper 5 bits of the 20-bit indicator. When captured, these bits are simply appended to the indicator. In this way, it is obviously required that in any list of indicators directed to the next level associated with the same bit mask, their bits 15..19 must be exactly the same. This can be achieved by moving the block at the next level during the construction process if the lower 15 bits of the indicators can be hidden, and the bit mask associated with the current level at the construction level The block is constructed at the next level of the (see section 3 for details). In summary, each digital character is composed of the following: • 1 bit, which can indicate whether the other is a direct sub-jump. If this is a sub-jump, there are 15 bits to represent the sub-jump index. If not, the following are: • 2 bits for digital use (displacement 値) • 5 bits for high index use of related indicators • 16 (Tier 1) or 8 (Tier 2 and 3) The single bit is used to indicate the index displacement 内 in the index array directly stored behind the block, that is, the index of the "first" index related to the digital character. χ) Level 1: A 65536-element pre-existing index array The first level (level 1) of the XTC data structure can be represented by a CPA with 4096 digital characters and 0 ... 65536 indicators. The first step in the table lookup operation is to use bits 31 ··· 20 of the IP address as the index to the 24-bit digital character that has been retrieved (in software implementation, it can be aligned The reason is to store the 24-bit digital character in a 32-bit integer). If the most significant bit of the digital character is set (indicated by the white characters on the black background in Figure 2), the next 15 bits represent the sub-jump index, and 18 paper sizes can be captured immediately. Standard (CNS) A4 specification (210 x 297 mm) (Please read the notes on the back before filling out this page) i Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 Α7 Β7 5. The invention description (η) came out and completed Lookup table job. In this case, bits 7… 0 ° are not used. In the second step, bits 19… 16 of the IP address are used to search the digital characters to find the indicator in the indicator list. This can be done by shifting the IP address by 4 bits with a 2-bit integer captured from a digital character (digital), and adding 16 bits that are also captured by the digital character. The displacement of the element is achieved (see Figure 3). The result number is the index number of the index stored in the index list related to the block. However, if the most significant bit of the digital character is not set, the lower I5 bit will form a jump index and complete the table lookup operation (see Figure 3). If the indicator does not represent a direct jump, the lower 15 bits can be captured and combined with the 5 bits stored in the digital character to form a 20-bit indicator, pointing to Level 2 CPA (see Figure 4), and continue to look up the table. The compressed consumer array printed by the Intellectual Property Bureau of the Ministry of Economic Affairs' employee co-operatives will be stored in a large array of 16-bit integers. Depending on who is immediately following the list of digital characters. Therefore, it is indeed impossible to index the i-th CPA at level 2. Conversely, the CPA at level 3 must be based on its starting point (the index of the first digital character), and this is why when representing a large path selection table, a 20-bit index is needed. Tier 2: — 256 element pressure index array The second level of XTC contains a large matrix of Tier 2 CPAs. Each such CPA includes 16 digital characters and 0 ... 256 indicators. After calculating the position of Level 2 CPA at Level 1, proceed as follows: 19 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 male β ~ -926 A7 B7 V. Description of the invention (β ) Perform a table lookup operation. The third step in the table lookup operation is to use bits 15 ... 12 of the IP address as an index to the 16-bit digital characters that have been retrieved. The most significant bit is set (indicated by the white characters on the black background in Figure 5), then the next 15 bits represent the secondary jump index, which can be immediately retrieved and completed the table lookup operation. In the fourth step , Bits 11 ·· 8 of the IP address are used to index to digital characters to find the index in the index list. This can be achieved by using the 2 bits captured from the digital character (digital) An integer to shift the IP address by 4 bits, and then add the 16-bit shift 値 also captured by the digital character (see Figure 6). The resulting number 値 refers to Index of the index stored in the index list related to the block. However, if the most significant bit of the digital character is After setting, the lower 15 bits will form a jump index and complete the look-up table operation (see Figure 6). If the indicator does not indicate a direct jump, the lower 15 bits can be retrieved and Combine with the 5 bits stored in the digital character to form a 20-bit index, point to the CPA of level 3 (as shown in Figure 7), and continue to look up the table. Follow the same way as in level 2. The compressed indicator array of this layer 3 will also be stored in a large array of 16-bit integers, where the size of each CPA is determined by the number of indicators in its indicator list. Therefore, one will be needed A 20-bit index to indicate the position of the next level of CPA. Level 3: The third level of the 256-element pressure index array XTC contains a large level 3 CPA matrix. Each of these 20 paper standards is subject to Chinese National Standards (CNS ) A4 size (210 X 297 mm) (Please read the precautions on the back first and write this page) rl pack

Order -------- I Printed by the Employees 'Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by 508926 A7 B7 V. Description of the Invention (q) CPAs include 16 digital characters and 0 … 256 indicators. After calculating the position of the CPA at level 3 at level 2, the table lookup operation is continued in a similar manner to that at level 2. The fifth step is to use bits 7 ··· 4 of the IP address as an index to the 16-bit digital characters that have been retrieved. If the most significant bit of the digital character is set (indicated by the black and white characters in Figure 8), the next 15 bits represent the sub-skip index, which can be immediately retrieved and completed the table lookup operation. In the sixth step, bits 3 ... 0 of the gray address are used to index to the digital characters, so as to find the indicator in the indicator list. This can be done by shifting the IP address by 4 bits with a 2-bit integer captured from a digital character (digital), and adding 8 bits that are also captured by the digital character. The displacement of the element is achieved (see Figure 6). The result number is the index of the index stored in the index list related to the block. At this point, all bits of the IP address are inspected. Therefore, the index retrieved from the index list must indicate a jump index. Therefore, this step is to complete the table look-up operation (as shown in Figure 9) 〇 As described in the previous four sections, I have fully described the data binding of XTC. The previous article explained in detail how blocks and digital characters are represented at various levels, and how information is encoded in them. Finally, it also explains in detail the table lookup operations performed in each hierarchy to obtain the secondary jump index or the index towards the next hierarchy. The lookup operation was performed in six steps to obtain results. In each step, except for the first access, which will access 24 bits (32 bits in software implementation), the rest will perform a 16-bit memory access. XTC method can be accessed from only 6 memories 21

This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 508926 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Invention description (w) and calculate the secondary jump index, and the worst The situation is a total of 13 (14) bytes accessed. XTC Construction In order to achieve the fastest possible construction method, XTC system is constructed by selecting the table items across the entire path, and making a jump table in a single review mode and constructing an overriding XTC notation. The only requirement for this construction operation is that the path selections are browsed in a certain order and a function is called for each item. Therefore, the representation of the dynamic path selection table will be completely independent of the XTC construction algorithm, and the XTC forward table can be created by each other. Path selection conversion and sub-jump table construction The complete construction process consists of three simultaneous steps (see Figure 10). In the first step, the dynamic routing table is viewed in the order of the IP address (starting point) and the IP mask length. Those with a smaller number of IP addresses are ranked before those with a larger number of IP addresses. For those with the same number of IP addresses, the one with the shorter IP mask length (shortest prefix) is ranked first. The tree structure view is closely related to the representation of the dynamic path selection table used, and in the case of Effnet, it is a Radix tree structure, which is not described here (the view path selection table is not an invention, and it is not Technology integration projects). Each path selection item will be sent for the path selection item conversion and sub-jump table generation procedure shown in FIG. 10. This program can effectively split the path choices into many prefixes, and discard the prefixes that cover other prefixes, that is, if 22 ^ paper size is applicable _ Chinese National Standard (CNS> A4 size (210 x 297 public issued ^ ~ " (Please read the notes on the back before filling this page)

Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs on consumer cooperation 508926 A7 B7 V. Description of the invention (> 1) As mentioned in the preamble. He is also responsible for generating the sub-hop table from the sub-hop information (interface, MAC address, ATM channel, etc.) appearing in the dynamic routing table. Finally, each prefix that originates from the split of the path option (many prefixes can be generated by splitting a path option) will be fed to the XTC builder program, which will be explained in detail in the next section. The same path selection conversion and sub-jump table generation program can be applied to both the FTC builder program and the XTC builder program. Therefore, I will not repeat the details here. They are not separate patent applications. Depth-first left-to-right XTC construction work After browsing through the path selection items described in Section 3.1 and splitting them into prefixes, each existing prefix is passed to the ongoing XTC construction work. That is, as mentioned in the preamble, each prefix will have a starting point and an ending point. With the conversion operation described in the previous section, these prefixes will be fed to the XTC construction operation in the same order. In addition, the endpoint of the next prefix to be processed will be equal to the current prefix plus one. All three levels of XTC are built at the same time. Depending on the length of the prefix that is being processed at this time, construction can be performed at different levels. Sometimes, the construction of a certain prefix is required at all levels. This will require a large number of accounting operations during the construction operation. I am following the following events ...

Expect expected the starting point of the next pending prefix has el address of construction class 1 23 This paper size applies to China National Standard (CNS) A4 specification (210 X 297 mm) 508926 Employees of Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the cooperative V. Description of the invention (/) A7 B7 base2 The address of the construction level 2 (point to the first level 2 block) base3 The address of the construction level 3 (to the first level 3 block) Address) chunk1 points to the address of layer 1 in the current construction chunk2 points to the address of layer 2 in the current construction chunk3 points to the address of layer 3 in the current construction pointer 1 points to layer 1 in the current construction The address pointer2 of the index list points to the address pointer 3 of the level 2 index list under construction. Pointer 3 points to the address f pos2 of the level 3 index list currently under construction relative to the position pos3 from 2 of the level 2 block currently under construction in base2. The position of the layer 3 block in the base3 project is related to the first layer 2 of the current layer 1 digital character. The position of the block 2 is related to the first layer 3 of the current layer 2 digital character. The bit of the block fromS is set to bucket 1. The bucket contains 16 indicators related to the current level 1 digital characters bucket2 [] contains 16 indicators related to the current level 2 digital characters bucket3 [] contains 16 The offset of the index related to the current level 3 digital characters offsetl The offset of the index related to the current level 1 digital characters 値 offset2 The index displacement of the current level 2 digital characters 値 offsets the index of the current level 3 digital characters Displacement 値 maskl is related to the bit mask of the current level 1 digital character mask2 is related to the bit mask of the current level 2 digital character maskS is related to the bit mask of the current level 3 digital character 24 (please read the back first Please note this page before filling in this page) * I ------ Order * ------ · This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) Employees of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the Consumer Cooperative 508926 A7 B7 V. Description of the Invention (岣) For each processed prefix, the expected variable can be checked based on the starting point to ensure that the prefix stream is valid. The processing operations of each prefix include "initialization phase", "construction phase" and "completion phase". In the initialization phase of the first prefix (starting point is zero), the entire construction operation is performed by setting basel, base2, and base3 to the memory address of each hierarchy, and setting p0s2, p0s3, and from2 It is initialized with zero from3 and so on. Here, chunkl can also be set to basel, and pointl to the memory address of the layer 1 index list to be stored. If the index list needs to be stored immediately after the block, then 値In order to set to basel + 8192, use this to initialize the construction operation of level 1. In addition, offsetl is also set to zero. Then, for all prefixes (including the first one), the initialization job executes the following items. First, the "depth" of the construction operation (which can be 12, 16, 20, 24, 28, or 32) can be calculated by analyzing the network mask stored in the prefix. To calculate the depth 値 for the network mask m, the following formula can be used: 32- ((! (M & 0xf) +! (M & 0xf0) +! (M & 0xf00) +! (M & 0xf000) + ! (m & 0xf 0000)) «2) where & means bitwise operation ,! Represents the logical "no" operation, that is, if x = 0, then x = l 'and 0 is the opposite, and the < < table τρ; bitwise shift operation to the left. It should also be noted that this bit mask is usually indicated by an integer in the hexadecimal notation that starts with 0x. By processing the specific depth of the calculated depth, 25 paper sizes are applicable to China National Standard (CNS) A4 (210 X 297 mm).

508926 A7 B7 5. Description of the Invention (蚪) Continue the initialization operation. However, for many deep single-word construction tasks that will need to be initialized, the deepest can be started. Depth 32 determines whether the construction of the new level 3 digital characters begins with the following formula:! (Expect & ((1 < <: 4)-1)) The result of the above formula is "exposure" (ie zero) ), There is no need to perform the initialization operation of the level 3 digital characters' and proceed to the construction stage. Otherwise, 'by setting mask3 to 0', set bucket3 [0], ..., ^) 11〇1 (^ 3 [15] is "EMPTY" to start the construction of digital characters of layer 3 "and then proceed to the initialization of depth 28 Operation. Initialization of Zeyan 2828 Determine whether the construction of the new three-level block begins with the following formula:! (Expect & ((1 < < 8)-1)) Employees ’Consumption of Intellectual Property, Ministry of Economic Affairs The result printed by the cooperative as above is "false" (ie, zero), so there is no need to initialize the block of level 3 and proceed to the construction stage. Otherwise, compare bits 15 ... 19 of from3 and P0s3 to ensure that They are equal. Remember that fr0m3 and Pos3 are related to the position of the first level 3 block and the current level 3 block of the current level 2 digital characters under construction, respectively, and remember that these points point to each other. The 20-bit indicator of the equal block will be stored in level 2, where 5 high-positioned elements are in the digital character, and 15 low-positioned elements are in the index related to the digital character (by fr0m3 and pos3 (Represented). Therefore, it may be necessary to 26_ this paper size applies Chinese National Standard (CNS) A4 Grid (210 x 297 public love) Coffee 508926 A7 B7 V. Description of the invention (4) Find the bits 15 ... 19 of from3 and pos3 are equal. If they are the same, you can set chunk3 to base3 + P〇s3, set pointed to chunk3 + 16 (this is the size of the block), and set 0ffset3 to zero 完成 to complete the initialization of depth 28. But if they are not the same, you need The compression index array built in layer 3 is adjusted starting from3, so that the 5 high-level elements become equal, and then the 20-bit index coding operation can be performed. This can be calculated by calculating p0s3 & OxfSOOO and値 Assign the f wrap variable to achieve. Then the delta of the adjustment (that is, how many positions of the CPA need to be moved) can be calculated by the difference between wrap and from3. Then, repeatedly by base3 [index] Assigned to base3 [index + delta] to move the CPAs, where the index is from p0s3-l to from3. The rest is stored in the bucket related to the digital characters of layer 2 under construction Indicators, because the position of CPAs related to their Tier 3 CPAs is indeed This can be achieved by adding a delta 値 to each index stored in the bucket2 array (skipping the EMPTY bucket), and then using a simple bit mask operation to retain only the lower 15 bits. Finally, from3 is set to wraP 'and P〇s3 adds a delta 値, and before completing the depth 28 initialization operation, set chunk3 to base3 + pos3 and p〇inter3 to chunk3 + 16 and ffff3 to zero 値' then The initialization operation of the depth 24 is resumed. Depth 24 initialization operation_ Determine whether the construction work of the new level 2 digital characters is opened by the following formula 27 This paper size applies to China National Standard (CNS) A4 specification (210 X 297 public love) (Please read the precautions on the back first Fill out this page again) --I-- till! Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 —_______ E .____ V. Description of the Invention (4) Start:! (Expect & ((1 «12)-1 ))

If the result of the above formula is "false" (that is, zero), it is not necessary to perform the initialization operation of the digital characters of level 2 and proceed to the construction stage. Otherwise, by setting mask2 to 0 and bucket2 [0], ..., bucket2 [15] to "EMPTY", start the layer 2 digital character construction operation, and then continue the initialization operation of depth 20. The initial operation of degree 20 determines whether the construction of the new layer 2 block starts by the following formula:! (Expect & ((1 < < 16)-1)) The result of the above formula is "false" (that is, zero)値), there is no need to carry out the initialization of Level 2 blocks and proceed to the construction stage. Otherwise, bits 15 ... 19 of from2 and Pos2 are compared to ensure that they are equal. Remember that from2 and Pos2 are related to the position of the first layer 2 block and the current layer 2 block of the current layer 1 digital characters under construction, respectively. Also, remember that the 20-bit indicators that point to their blocks will be stored in level 1, of which 5 high-position elements are in the digital character, and 15 low-position elements are in the index related to the digital character. (Represented by fr0m2 and pos2). Therefore, it is required that the bits 15 ... 19 of fr0m2 and poS2 are equal. If they are the same, you can complete the initialization of depth 28 by setting chunk2 to base2 + pos2, pointer2 to chunk2 + 16 (this is the size of the block), and offset2 to zero. operation. But if they are not the same, the compression index array built in layer 2 needs to be 28. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 public love) 508926 A7 ____ __B7____ V. Description of the invention (State ) The adjustment starts from fr0m2 so that the five high-level elements become equal, and then a 20-bit index coding operation can be performed. This can be achieved by calculating p0s2 & 0xf8000 and assigning this unit to the wrap variable. Then the delta of this adjustment (that is, how many CPA positions need to be moved) can be calculated by the difference between wrap and from2. Then repeatedly move the CPAs by repeatedly specifying base2 [index] to base2 [index + delta], where the index loop is from pos2-l to from2. What remains is to update the various indicators stored in the set of digital characters related to the stratum 2 under construction. This is because / the position of the stratum 2 CPA related to them has indeed changed. This can be achieved by adding a delta 对 to each indicator stored in the bucket array (skipping the EMPTY bucket), and then using a simple bit mask operation to retain only the lower 15 bits. Finally, set from2 to wrap and add pos2 to deltadel, and before completing the depth 28 initialization operation, set chunk2 to base2 + pos2 and pointer2 to chunk2 + 16 and set ffset2 to zero, and then Continue the initialization operation at depth 16. Depth 16 initialization operation Intellectual property staff of the Ministry of Economic Affairs X. Consumption Kettle · fi. Yinli determines whether the construction of the new class 1 digital characters is started by the following formula:! (Expect & ((1 < < 20 )-1)) If the result of the above formula is "false" (that is, zero), then it is not necessary to perform the initialization operation of the digital characters of level 1 and proceed to the construction stage. Otherwise 'by setting maskl to 0' bucketl [0], ..., 1 > 11〇1 < ^ 1: 1 [15] is "29 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 cm) 508926 508926 Intellectual Property Staff of the Ministry of Economic Affairs X Consumption Cooperation Du printed A7 B7 V. Description of Invention (β) EMPTY "to start the construction of digital characters at level 1, and then continue the initialization at depth 12. Depth 12 initialization operation There is no need to perform initialization operation here. When the entire construction operation is started, it will be arranged. After the initialization phase, the expect variable can be updated by setting it to the end point of the processed prefix by one. Then, proceed to the construction phase and perform the construction operation at the depth calculated at the beginning of the processing procedure. Depth 32 construction operation The direct sub-jumps stored in the level 3 index can be constructed as follows. The bit index 値 bix can be calculated by extracting its bits 0 ... 3 from the starting point of the prefix. Then, set the bit bix of mask3, and store the sub-jump index 値 (uncoded) of the prefix into bucket3 [bix]. During the completion phase, the contents of these buckets will be moved to the indicator array related to the layer 3 digital characters under construction. Liangdu 28 construction operation The following method can be used to construct the direct jump directly encoded in the layer 3 digital characters. The indexes of the first and final digital characters to be constructed, namely start and stop, can be calculated by extracting the bits 4 ... 7 of the starting point and ending point of the prefix respectively. Then, set bit 15 to apply the Chinese National Standard (CNS) A4 specification (210 X 297 mm) to the 30 paper sizes of the prefix ----------- 丨 ----- ( Please read the precautions on the back before filling this page.) Order. 508926 Printed by A7 B7, Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs 5. Description of the invention (β) Jump index 値, and the encoded jump Index 値 is assigned to chunk3 [start], ..., chunk3 [stop] to store the chunk. Depth 24 construction operation The direct sub-jumps stored in the level 2 index can be constructed as follows. The bit index 値 bix can be calculated by extracting bits 8 ... 11 from the starting point of the prefix. Then, bit 2 of mask2 is set, and by setting bit 15, the "jump index" of the prefix is encoded and stored in bucket2 [bix]. During the completion phase, the contents of these buckets will be moved out into the indicator array related to the layer 2 digital characters under construction. Depth 20 construction operation The following method can be used to construct a direct jump directly encoded in the layer 2 digital characters. The indexes of the first and final digital characters to be constructed, namely start and stop, can be calculated by extracting the bits 12 ... 15 of the starting point and ending point of the prefix respectively. Then, bit 15 is set to encode the sub-jump index 値 of the prefix, and the encoded sub-jump index 値 is assigned to chunk2 [start], ..., chunk2 [stop] to store the block. Depth 16 construction operation The direct sub-jumps stored in the level 2 index can be constructed as follows. The bit index 値 bix can be calculated by extracting its bits 16 ... 19 from the starting point of the prefix. Then, set bit 2 of mask2, and set bit 15 of this prefix by encoding bit 15, and then save it into 31. This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ) ------------ · Install -------- 丨 Order i (Please read the notes on the back before filling out this page) # 508926 Printed by the Employees ’Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs System A7 B7 V. Invention description (π) Within bucket2 [bix]. During the completion phase, the contents of these buckets will be moved out into the index array related to the stratum 1 digital characters under construction. Depth 12 construction operation The following steps can be used to construct a direct jump directly encoded in the layer 3 digital characters. The indexes of the first and final digital characters you want to build, namely start and stop, can be calculated by extracting the starting point and ending point of the word at 20 ... 31, respectively, and then counting to the left. Shift 8 bits and set bit 23 to encode the first jump index 値 of the word and assign the encoded second jump index 编码 to chunkl [start], ..., diunkl [St Op] to store the block. Remember that the first block of the hierarchy contains 24-bit integers, and the lower 8-bits are unused here. In software implementation, the 24-bit digital character can be stored in a 32-bit integer for alignment reasons, and only a slight modification of the encoding method is required. The prefix processing operation can be completed within the completion phase, which starts at the same depth as the construction operation. For a certain prefix, completion work of multiple depths may require starting from the deepest position. The completion of the depth 32 operation can be determined by the following formula whether the construction of the digital characters of level 3 is completed (remember that the expect variable has been updated after the initialization stage): '· (expect & ((1 < < 4) -1)) If the result of the above formula is "false" (that is, zero), then one or more prefixes are required to complete the level 3 digital characters, and the current processing of prefixes is indeed 32 (CNS) A4 specifications (210 X 297 mm) ------------- (Please read the precautions on the back before filling out this page) 508926 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 5. The invention description (μ) is completed. Otherwise, the index tix of the digital character to be completed can be calculated by extracting the bits 4 ... 7 'of the starting point of the prefix. Next, analyze mask3 to calculate the two-bit shift 値 shift. This can be obtained by the following simple program: mask3 & Oxaaaa > 0 shift = 0 mask3 & 0x4444 > 0 — > shift = 0 mask3 & 0x1010 > 0 > shift = 0 mask3 & 0x0100 > 0 shift = 0 This can calculate the shift 値 and the code of the digital character (shift < < 8) + offset3, and store it in 'chunk3 [tix]. In order to complete the digital character, the indicators stored in the bucket must be moved out to the index list related to the digital character. Depending on the displacement, we can extract indicators from 2, 4, 8 or 16 buckets. Displacement unit set buckets 0 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 1 0, 2, 4, 6, 8, 10, 12 , 14 2 0,4,8, 12 3 0, 8 The indicators can be retrieved from the buckets, and borrowed from the buckets marked in the table above, and stored them in the relevant digital Character array. For each non-EMPTY bucket, the contents of the bucket will be assigned to pointed [offset3], and for each EMPTY bucket, the pointer3 [offset3-l] will be assigned to p0inter3 [offset3]. Therefore, offset3 is incremented by one before proceeding to the next bucket. After processing the buckets listed in the table above, the digital character is completed and the deep 28 completion operation is continued. 33 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ------------ · I ------ Order --------- ( Please read the precautions on the back before filling in this page) 508926 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention (p) The completion of the depth 28 can be determined by the following formula: Whether to end (remember that the expect variable has indeed been updated after the initialization phase):! (Expect & ((1 < < 8)-1)) If the result of the above formula is "false" (that is, zero), then One or more prefixes are required to complete the three blocks of the hierarchy, and the current processing of prefixes has been completed. Otherwise, after setting bit bix of mask2, by extracting bits 8 ... U of the starting point of the prefix, calculate the level 2 bit index bix. Finally, before proceeding to the depth 24 to complete the operation, the lower part of the P0s3 is 15 bits. This is the position of the layer 3 block that is being completed at the moment. After adding 16 + offset3 to pos3, it is stored bucket2 [bix] to prepare for the construction of the next layer of 3 blocks. The completion of the depth 24 operation can be determined by the following formula whether the construction of the digital characters of level 2 is completed (remember that the expect variable has been updated after the initialization stage):! (Expect & ((1 < < 12)- 1)) If the result of the above formula is "false" (that is, zero), one or more prefixes are required to complete the level 2 digital characters, and the current prefix processing operation is confirmed to be completed. Otherwise, you can calculate the index tix of the digital character you want to complete by extracting bits 12 ... 15 of the starting point of the prefix. Next, analyze mask2 to calculate the displacement of two bits; shift. This can be obtained by the same simple procedure as in the above-mentioned depth 32 completion operation. Therefore, the displacement 値 and the number of digital characters can be calculated. 34 This paper size is applicable to the Chinese National Standard (CNS) A4 (210 X 297 public love) (Please read the precautions on the back before filling this page) · n I nn · Ϋ i > — n ″, I nmn ϋ i 508926 A7 B7 V. Description of the invention (c) ((from3 & 0xf8000) »5) + (shift« 8) + offset2 and store it in chunk2 [tix] Inside. In addition to the coding of the depth 32 completion operation, the index bits 15 ... 19 pointing to the level 3 CPAs related to the digital character will be extracted from from3, and from3 is set to pos3 to prepare for the next level 2 digital After the character construction operation, it is encoded into the digital character bit 10. ·· 14. In order to complete the digital character, the indicators stored in the bucket must be moved out to the index list related to the digital character. Depending on the displacement, we can retrieve indicators from 2, 4, 8 or 16 buckets in the same way as in the previous depth 32 completion. The indicators are obtained from the buckets, and the buckets marked in the above table are borrowed from the tour, and stored in the indicator array related to digital characters. For each non-EMPTY bucket, the contents of the bucket will be assigned to point2 [offset2], and for each EMPTY bucket, the pointer2 [offset2-l] will be assigned as pointer2 [offset2]. Therefore, 0ffset2 will be incremented by one before proceeding to the next bucket. When the buckets listed in the above table are processed, the digital character is completed and the depth 20 completion operation is continued. Please read it first

Remarks (reload this book. Page I Order # The consumer cooperation of the Intellectual Property Bureau of the Ministry of Economic Affairs, the printing depth 20 is completed. The operation can be determined by the following formula whether the construction of the 2 blocks of the stratum is completed (remember that the expect variable is already in the Updated after the initialization stage):! (Expect & ((1 «16)-1)) As the result of the above formula is" false "(that is, zero), one or more prefixes are required to complete the level 2 block Area, and the processing of the current prefix has been completed. 35 This paper size is applicable to the Chinese National Standard (CNS) A4 (210 X 297 mm) 508926 Employees' cooperation in intellectual property bureau of the Ministry of Economic Affairs, printed A7 B7 V. Description of the invention (Bait). Otherwise, you can calculate the level 1 bit index bix by extracting the bit 16 ... 19 of the initial point of the maskl bit bix. Finally, go to depth Before the 16 operation is completed, the lower part of the P0S2 is 15 bits. This is the position of the block 2 that is currently being completed. It will be added to bucketl [bix] after adding 16 + offset2 to pos2. In preparation for the construction of the next level of 2 blocks. Industry, you can determine whether the construction of the digital characters of level 1 is completed by the following formula (remember that the expect variable has been updated after the initialization stage):! (Expect & ((1 «20)-1)) as above The result of the formula is "pseudo" (that is, zero), then one or more prefixes are required to complete the level 1 digital characters, and the current prefix processing operation is confirmed to be completed. Otherwise, the prefix can be retrieved by extracting Point bit 20 ... 31, to calculate the index tix of the digital character you want to complete. Then, analyze maskl to calculate the displacement of two bits 値 shift. This can be done by the same operation as in the previous depth 32 completion operation. It can be obtained by a simple program. Therefore, the code of the displacement 値 and the digital character ((from2 & 0xf8000) «3) + (shift« 16) + offsetl can be calculated and stored in chunkl [tix]. Except the foregoing In addition to the coding of the completed operation at depth 32, the index bits 15 ... 19 pointing to the level 2 CPAs related to the digital character will be retrieved from from2, and from2 is set to pos2 to prepare for the next level 1 digital character After encoding, encode it into bits of a digital character 20 ... 23. In order to complete this number, 36 paper sizes are applicable to China National Standard (CNS) A4 (210 X 297 mm) ----------------- (please first (Please read the notes on the back and fill in this page) 508926 A7 ___B7_ V. Description of the invention (5) The code character 'that is, the indicators stored in the bucket must be moved out to the index list related to the digital character. Depending on the displacement, we can retrieve indicators from 2, 4, 8 or 16 buckets in the same way as in the previous depth 32 completion. The indicators are obtained from the buckets, and the buckets marked in the above table are borrowed from the tour, and stored in the indicator array related to digital characters. For each non-EMPTY set bucket, the contents of the set bucket will be assigned to pointerl [0ffSetl], and for each EMPTY set bucket, the pointerl [offsetl_l] will be assigned as p〇interl [0ffsetl]. Therefore, offsetl is incremented by one before proceeding to the next bucket. After processing the buckets listed in the above table, the digital character is completed and the depth 12 completion operation is continued. Depth 12 Completion The gXI XTC construction operation was confirmed to be complete. In the construction operation software, some statistics about the construction operation can be updated, but it is not necessary to perform the operation completed in depth 12. In this section, we will explain the complete method of processing the route selection items that are repeatedly fed into the construction operation, compressing them and converting them into XTC representations. xtc lookup table implementation This section describes many possible XTC lookup table hardware and software implementations. Some of these are intuitive, but some of them are slightly different from the previous narratives. The details of the structure of the data and the procedures of the builder need to be changed in detail. 37 This paper size is applicable to Chinese National Standard (CNS) A4 specification (210 X 297 meals) ^ (Please read the precautions on the back before filling this page) r. 丨 Binding --------. Etc. 508926 Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs' Consumer Cooperatives A7 B7 V. Description of Invention (4) Intuitive hardware implementation This XTC lookup table is very suitable for hardware implementation. Figure 11 shows the layout of an intuitive hardware implementation based on the algorithm described in the previous section. ° The IP address can be fed from the left end for table lookup, and the entire table lookup operation is processed in six steps. , Until the jump index 値 is available at the right end. It is assumed that the data structure is stored by individual strata in individual records. In other words, one is for a single "level 1 block, one is for an array of indicators related to node level 1, one is for node level 2 and the index 'and one is node 3 for the level and index (these also Can be mapped into a single memory). If the table lookup operation is indeed completed in six steps, the result will be “locked” and the 15 bits of the result will be fed to the multiplexer chain sequence. Pressing this' will come from other parts of the table lookup operation, but not necessarily discard it for simultaneous results. If you have used four memories that can be accessed at the same time, the table lookup operation will be pipelined in four steps ’to speed it up. Because of its simplicity, it does not include detailed descriptions of the necessary control logic. Cost-Effective Hardware Implementation Figure 12 shows a more cost-effective hardware implementation layout, which has a slightly lower performance but a higher compression. The idea is to have four rather than three levels, where each level is represented exactly as levels 2 and 3 in the original data structure. This simplifies the design, but adds two extra memory accesses to the lookup table. 38 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----------------- (Please read the precautions on the back before filling this page) Order --- # 508926 Employees' cooperation in the Intellectual Property Bureau of the Ministry of Economic Affairs printed A7 B7 V. Description of Invention (VI) It is also assumed that only one memory is used for access, which means that the time spent on a certain lookup table is It is proportional to the time taken to perform eight memory references. Here, a multiplexer is used to pick up four bits from the IP address at the same time, and to interpret the results of each memory access as digital characters or indicators. If the table lookup operation is completed before all eight steps, a multiplexer is also used here to lock the previously accessed memory address. This may be a better way to lock results, even for other designers, because it prevents illegal memory access and improves the stability of this implementation to some extent. Real-time hardware implementation To increase the lookup speed, it is indeed possible to use six memories that can be accessed simultaneously and change these digital characters into 24-bit digital characters at all three levels. In this way, a 21-bit indicator can be stored in the indicator array, which means that we can store the indicator in a separate memory without having to store the indicator related to a block directly behind the block. As shown in Fig. 14, compared with the original design method, the design can be simplified slightly. The reason is that each block will occupy exactly 16 24-bit characters in the block memory in the hierarchy to which it belongs, which means that the starting point of the block will always be a multiple of 16. Therefore, 4 bits will be stored for reference of the block, so the 15 bits of the indicator are enough to refer to the block of the next level. Therefore, it is not necessary to encode the high-level elements of these indicators, nor is it necessary to extract their bits to obtain an "enough" indicator. This lookup table can be used to schedule 39 pipelines. The paper size is applicable to the Chinese National Standard (CNS) A4 (210 X 297 mm) ttw ------— ^ ------— (Please read the back first Note: Please fill in this page again) 508926 Printed by A7 B7 of the Consumer Cooperatives of Intellectual Property Bureau of the Ministry of Economic Affairs 5. The invention description (β) is in six steps, which means that each time you complete a memory access, you also complete a Path selection lookup job. In this way, the simple look-up table design method can achieve the performance of more than 100 million packets per second. Intuitive software implementation In terms of software implementation, we can't easily add extra memory access, but we can use 24 bits to represent this level 1 digital character. Therefore, they use 32-bit characters and store the index shift 値 in the upper 16-bits (see Figure 15). Except for the 8 least significant bits, if the digital character does not indicate a direct jump, these are not used. The lower 15 bits are in the same way as levels 2 and 3. Apply it. In addition, the overall data structure is stored in a sufficiently large contiguous block of memory, which is calibrated by a constant number of spatial motions allocated to each of the three levels. In addition to these minor changes, the software version of this lookup operation is the result of intuitive implementation of the algorithm written in Section 2. Performance improvement

Compared to the FTC data structure, the performance improvement results of the XTC data structure are derived from exchanging memory access, simplicity, and speed with memory usage. It is a well-known fact that if the amount of memory is really irrelevant, then someone can use a single memory access to perform a very simple path selection lookup operation. This can be achieved by preparing only an array of 232 elements, where the sub-hop index of each possible IP address is stored in a paper with the same IP 40 size as the Chinese National Standard (CNS) A4 (210 X 297) (Mm) --- I ------ ^ --- I-- (Please read the notes on the back before filling out this page) Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 B7 V. Description of the invention (4) In the slot with the same number of addresses. In order to obtain this exchange result, if only a single layer with 228 digital characters is considered, an XTC lookup table can be performed with only two memory accesses, but it will still consume a very large amount of memory . One possible way is to reduce the number of digital characters by allowing each digital character to represent more than 16 bits. For example, if each digital character represents 256 bits instead of 16, then an 8-bit digital character must be used as a displacement 値. In the worst case, the number of "externally increased" indicators generated by the simplified XTC encoding method will increase with the worst factor being > 56/9. The total number of digital characters would be 224, and the number would still be too large. A better way to improve performance without wasting too much memory is to reduce the lookup table to four memory accesses. This can be easily achieved by building an XTC with only two levels. , Where, for example, level 1 indicates the upper 20 bits, and each level 2 blocks indicates the lower 12 bits. According to this, each block in level 1 will be increased by a factor of 16, which can also provide a rough estimate of the overall growth of the data structure. And this can be combined with the idea of using “larger” digital characters at level 1 as described above. The most intuitive way to increase speed is to use XTC technology only at levels 2 and 3, while using direct addressing at level 1. This will remove a memory access operation. This section focuses on the idea of how to increase performance by consuming more memory. If the goal is to reduce the representation as much as possible, the anti-terms of these thoughts can be adopted. That is, "smaller" digital characters and more levels 41 This paper size applies to the Chinese National Standard (CNS) A4 specification (21〇X 297 public love) I -1 -------- ^ ------ --- (Please read the notes on the back before filling in this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 508926 A7 B7 V. Description of the invention (drawing) 〇 Obviously, the present invention can provide an improved IP Address look-up table method and system 'can fully meet the goals and advantages set by the previous disclosure. Although the present invention has been described in terms of specific embodiments, the alternatives, modification results, and changes in transformation are obvious to those skilled in the art. 42 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm 7. I -------- ^, I -------- (Please read the precautions on the back before (Fill in this page)

Claims (1)

508926 A8 B8 C8 D8 6. Application for Patent Scope 1 · —A system for IP path selection look-up table in the IP path table, which contains many prefix items of any length related to the sub-hop information to determine these. Where the IP network packet should be sent is characterized by a data structure device used to represent the set of dominant points in the IP address space of each prefix in the routing table, and the device contains a three-level array of pressure indicators ( CPA), each of which includes an array of blocks and related indicators, where each index includes a jump index 値 or an indicator pointing to the next level of the block. 2. The system according to item 1 of the scope of patent application is characterized by 20 bits in the index, and the jump index 跳跃 contains 15 bits. 3. If the system of claim 1 or 2 is applied for, it is characterized in that the CPA at the first level represents an uncompressed index array with a size of 65536. The IP address bits 31..16 are composed The CPAs of levels 2 and 3 each indicate an index of 256, which are bits 15 · 8 and 7 · 0 of the IP address, respectively. For example, the system of the first scope of the patent application is characterized in the block It contains digital characters, where each digital character represents the 16 bits of the recessive bit array that represents the block, and the size of the recessive bit array is equal to the largest number of indicators in the CPA. 5. If the system of item 4 of the scope of patent application is characterized in that the bit with index 値 i in the recessive bit array is set if and only if a certain prefix begins in the At position i. 6. The system according to item 5 of the patent application, characterized in that the recessive bit array is represented by a recessive 16-bit mask, wherein each bit mask corresponds to a digital word Yuan, including blank paper size without bit set. Applicable to China National Standard (CNS) A4 specification (21 × 297 cm) (please read the notes on the back before filling in. Page, 1T Employees of Intellectual Property Bureau, Ministry of Economic Affairs) Printed by a consumer cooperative 508926 A8 B8 C8 D8 Printed by a consumer cooperative of employees of the Intellectual Property Bureau of the Ministry of Economic Affairs 6. Scope of patent application \ \ 1 yuan mask, a bit mask with only one bit set, and / or two or Bit masks for more bit sets. 7. If the system of claim 6 of the patent application is characterized by a multiple of 16 at the starting point of the block, and it can completely cover the entire bit mask, And all the following empty masks. 8. If the system of the scope of patent application No. 7 is characterized in that for the bit masks with a set of bits and the following set of empty bits, it can jump directly Index 値 encoded as digital characters 9. If the system of item 8 of the patent application scope is characterized by 676 different 16-bit masks with a set of two or more bits, among which the bit mask and its corresponding index can be represented by digital characters. And the index list related to the digital character. 10. The system of item 9 in the scope of patent application is characterized by a device sufficient to recognize that the size of the prefix set is indeed the shortest. Prefixes, and their starting points are located in the same bit mask, and the prefixes with a size larger than the shortest prefix are divided in half until all the prefixes have the same shortest size, Each time the prefix is divided, the zero bit is dialed into the bit mask, and a copy of the index related to the set of bits closest to the left end of the dialed bit is directly increased. It is listed after the duplicated index in the index list. 11. The system of item 10 in the scope of patent application is characterized in that the index list includes two, four, eight, or sixteen relevant digital characters. Indicators 12. If applying for a patent The system around item 10 is characterized in that each ____ 2 __: __ I paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) (please read the precautions on the back before filling in the pages). 508926 Α8 Β8 C8 D8 6. The digital characters in the patent application scope include: 1 bit, which can indicate whether the other is a direct sub-jump. If this is a sub-jump, there are 15 bits to indicate the sub-jump index, and If this is not a sub-jump, there are 2 bits for the digital displacement 値, 5 filTT: for the high-level elements of the relevant index, 16 bits for the level 1 or 8 bits for the level 2 and 3 are used to indicate the index offset 値 ', that is, the index 値 of the "first" index related to the digital character in the index array directly stored in the block. H. A method for performing IP path selection lookup in the IP path selection table 'contains an arbitrary-length prefix entry in the skip table related to the sub-hop information to determine where the IP network packets should be forwarded to It is characterized by the following steps: 'Access the first-level digital characters in the block in the first-level pressure index array at the local (31..20) position corresponding to the first index of the IP address (23..0), and if the most significant bit of the digital character is set, the sub-skip index bit (22 ·· 8) in the digital character can be extracted. I4. If the method of claim 13 is applied, it is characterized by the following further steps: If the most significant bit of the digital character is not set, then the bit of the digital character is shifted (17..16) To shift the second index 値 part of the IP address (19 " 16), and append the offset bit (14 ·· 0) of the digital character to the shifted second index 値 part to form a Index index: This paper size applies Chinese National Standard (CNS) A4 specification (210 × 297 mm) (Please read the notes on the back before filling in I ^ page) Order _ Printed by the Intellectual Property Bureau Employee Consumer Cooperative of the Ministry of Economic Affairs 508926 A8 B8 C8 D8 6. The scope of patent application is to access the index of the first level at the index index position of the index list in the index index array of the first level, and if the most significant bit of the digital character is set, it can be used at This index extracts the secondary jump index bit (14..〇). 15. The method according to item 14 of the scope of patent application, which is characterized by the following further steps: If the most significant bit of the indicator is not set, then the remaining bits (14) 0) and The index bit of the digital character is partial (22 ·· 18), so that for the starting point of the first-level impulse index array in the set of second-level index impulse array set, a * 20-bit first Second-level indicator starting indicator. 16. The method according to item 15 of the scope of patent application, which further includes the following further steps: At the position of the third index corresponding to the IP address, at a local (15..12) position, access the second-level pressure index The second level of digital characters (15 · 0) in the block in the array, and printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. If the most significant bit of the digital character is set, the digital character can be extracted Index number of sub-hops within the element (14..0). 17. The method according to item 16 of the patent application, which is characterized by the following further steps: If the most significant bit of the digital character is not set, then the bit of the digital character is shifted (9..8) To shift the fourth index 値 partial (11 " 1) of the IP address, and append the shift bit (7 ·· 0) of the digital character to the shifted fourth index _______4 ____ — this paper Standards: Chinese National Standards (CNS) A4 specifications (210 × 297 mm) 508926 B8 C8 P8 ------- 7T, application for patent scope (partial) to form an index index, ^ in the second level At the index position of the index in the index list in the index array, access the second-level index, and if the most significant bit of the digital character is set, the next jump index bit can be extracted from the index (14. 〇). 18. The method according to item 17 of the scope of patent application, which is characterized by the following further steps: If the most significant bit of the second-level indicator is not set, then the remaining bits in the indicator are combined with the number The index of the characters is local, so that the starting point of the third-level index array in the third-level index array set constitutes a 20-bit third-level index. 19. The method according to item 18 of the scope of patent application, which is characterized by the following further steps: At a position corresponding to the fifth index 値 local (7 ·· 4) corresponding to the IP address, 'access the third-level pressure index The third-level digital characters (15..0) in the block in the array, and if the most significant bit of the digital character is set, the sub-skipping index bit (Η ·· 0). 20. The method of claim 19 in the scope of patent application, characterized by the following further steps: If the most significant bit of the digital character is not set, then the bit of the digital character is shifted (9 ·· 8) To shift the sixth index 値 part (3 " 0) of the IP address, and add the offset bit (7 ·· 〇) of the digital character to the shifted sixth index book! China National Standard (CNS) A4 specification (210X297 male sauce) (Please read the notes on the back before filling. Pack-page, tT Printed by the employee consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs __5 _; __ 508926 A8 B8 C8 D8 6. The scope of the patent application (partially to form an index index). At the index index position of the index list in the index list of the third-level pressurized index, access the third-level index, and extract the secondary jump index from this index. Bit (14..0). 21 · —A computer program product that can be directly loaded into the internal memory of a digital computer, which is characterized in that the product includes a software digital device to perform any of the items in the aforementioned patent application range 13-20 Steps 22 · —There are A computer program product of a brain-readable medium, which is characterized in that the computer program digital device is stored on the medium, and after being loaded into the computer, the computer can cause the computer to execute any of the items in the aforementioned patent application range 13-20 Step 23.-A system for IP path selection lookup in the IP path selection table, including a number of prefix entries of any length in the sub-hop table related to the sub-hop information to determine the IP network Road packets should be sent to a destination, which is characterized by a device that performs the steps of any of the items in the aforementioned patent application range 13-20. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 24.-Kind in the IP path selection table The method of creating a data structure device by performing IP path selection look-up table includes a number of arbitrary length prefix items related to the sub-hop information to determine where the IP network packets should be forwarded. It is characterized by the following steps : A device for generating a data structure, used to represent the set of leading points in the IP address space of each prefix in the path selection table, which contains three Hierarchical Pressure Indicator Array (CPA), each of which contains an array of blocks and a related index, and each index contains a sub-jump index. The paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm 508926 Printed by A8, B8, C8, D8, Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. 6. The scope of patent applications or indicators pointing to the next level of the block. 25 · —A computer program that can be directly loaded into the internal memory of a digital computer The product is characterized in that the product contains the software digital device, which can perform the steps in the scope of application for patent No. 24. 26. A computer program product containing a computer-readable medium, which is characterized in that the computer code device is stored on the medium and loaded into the computer, so that the computer can execute the scope of application for patent application No. 24 Item steps. 27 · —A system for generating a data structure device applied to an IP path selection lookup table in an IP path selection table, including an arbitrary-length prefix item related to the sub-jump information in the sub-jump table to determine Where the IP network packets should be sent is characterized in that the device can be used to perform step 24 of the scope of patent application. This paper uses Chinese National Standard (CNS) A4 (210X297 mm)