RU2304802C1 - Device for processing addresses of commutator in a local area network, operating according to transparent bridge principle - Google Patents

Abstract

FIELD: computer engineering, possible use during creation of active devices for local area networks.

SUBSTANCE: in accordance to invention, device includes address table, block for realization of hash function of first and second type, block for realization of orderly filling and binary search operations and control block. Address table is divided on two independent physical memory pages. Task queue is connected to block for realization of orderly filling and binary searching operations. Address table, blocks for realization of hash function of first and second types, task queue and block for realization of orderly filling and binary search operations are connected to control block.

EFFECT: possible usage of hash functions in conjunction with orderly filling and binary searching algorithms, increased speed of operation and increased reliability of commutator address processing.

1 dwg

Description

The present invention relates to computer technology and may find application in creating active devices of local area networks.

Switches use two main routing algorithms: a transparent bridge algorithm or a source routing algorithm [1].

In the source routing algorithm, the complete message route information is stored in the message itself. This information is used by the switches directly as a command to switch messages from the input port to a specific output port or group of output ports. The entire route is determined even at the beginning of the transmission of the message and is set by the transmitting station.

Transparent bridge switches are invisible to endpoints on the network. When using switches operating on the principle of transparent bridges, end stations work in the same way as in the absence of them, that is, they do not take any additional actions to ensure that the message passes through the switch. The switch, which works on the principle of a transparent bridge, independently builds a special address table, on the basis of which it decides which ports to send the received message to. A switch operating on the principle of a transparent bridge builds its address table based on passive monitoring of the schedule circulating in segments connected to its ports. At the same time, the switch operating on the principle of a transparent port takes into account the addresses of message sources arriving at the switch ports. At the source of the message, the switch concludes that the station with this address belongs to the network segment from which the message arrived. Each switch port acts as the endpoint of its segment with one exception - the switch port does not have its own address. The switch port operates in the illegible mode of message capture, when all messages arriving at the port are stored in the buffer memory. Using this mode, the switch monitors the entire schedule transmitted in segments attached to it and uses the messages passing through it to study the composition of the network.

Training for new addresses can be done in various ways.

The most direct way is to add a new address to the next cell. But such filling will be disordered, and the search for such an address in the address table will require large computational costs.

The most commonly used method is the use of special hash functions. The hash function maps a wide field of the station address in the network to a narrow field of the cell address in the address table where this station address is stored. The number of addresses that end stations in the network can have, as a rule, significantly exceeds the number of addresses that a switch operating on the principle of a transparent bridge can save in its address table. Therefore, collisions are not ruled out when the same cell in the address table corresponds to different addresses of stations in the network. Additional mechanisms are used to get rid of collisions, for example, several cells of the address table are mapped to each hash function value.

In addition to using hash functions, you can also use the method of orderly filling out the address table in ascending order of addresses of end stations in the network or in descending order. When a new address arrives, the address table is reordered. The address search in the table can be carried out by the binary search method. The binary search method is as follows. In an ordered data set, some arbitrary data is selected and this data is compared with the desired one. The address of the selected data is remembered. If the selected data is less than the desired, then the stored address is divided in half. This address is retrieved at the received address and again compared with the search. If it is larger than the desired one, then half the distance from this address to the address of the previous compared data is added to the current address. This address is retrieved at the found address and compared with the search. This continues until the desired data is found, or until the distance between the data involved in the next two comparisons is equal to unity, which means that there is no such data in the ordered data set. When using the binary search algorithm, the maximum time spent searching for a cell in the address table grows logarithmically at the base of two of the number of filled cells in the address table.

Closest to the proposed device is an address processing device used by Allayer in its local area network switches operating on the principle of a transparent bridge, which includes an address table, a block for implementing a hash function of the first type, and a control block [2]. As a hash function, the Allayer address processing device uses cyclic polynomial calculation algorithms [3], similar to calculating a cyclic redundancy code CRC (Cycle Redundancy Code), which is calculated over the entire field of a network address. Each hash function in the address table has multiple cells. After calculating the value of the hash function, the device selects a cell from the address table located at the address equal to the value of the hash function. By comparing the network address located in the cell with the network address, the cell with which you want to find, the device determines whether the cell is found correctly or a collision has occurred. If the searched network address and the found network address do not match, the device concludes that a collision has occurred and it is necessary to check the next cell corresponding to the same value of the hash function. For each hash function value, there can be 8 such cells in the internal address table, 4 in the address table using external memory, and up to 7 cells shared by all the hash function values. The minimum guaranteed selection of network addresses at the worst resolution for hash functions, when all the addresses of the sample lead to the same value of the hash function, contains 15 elements, which makes up a very small part of the total size of the address table (from 1024 cells and above). The time spent searching for network addresses that lead to the same value of the hash function grows linearly as the number of such addresses stored in the address table increases.

The objective of the invention is to increase the speed and reliability of address processing in switches operating on the principle of a transparent bridge and in which addresses would be processed quickly for arbitrary sets of addresses, while there would be a minimum guaranteed number of addresses that would be stored in the address table for the worst sample of these addresses in the sense of a hash function, and this minimum guaranteed number of addresses would be comparable with the size of the address table, and the maximum time to search for a set If the address table is full, it would increase more slowly than linearly.

The technical result is to ensure the use of hash functions in combination with an ordered filling algorithm and binary search.

The proposed device for processing addresses of a local area network switch operating on the principle of a transparent bridge uses both a hashing mechanism and an ordered filling and binary search mechanism.

The proposed device is illustrated in the drawing, which shows a functional diagram of the address processing device of a local area network switch operating on the principle of a transparent bridge.

The device contains the following functional blocks:

· Address table 1, consisting of:

- A physical page of memory 2, consisting of a hash function region of the first type;

- Physical memory page 3, consisting of

- Hash function areas of the second type 4,

- Regions of ordered filling and binary search 5,

· Block implementation of the hash function of the first type 6;

· Block implementation of the hash function of the second type 7;

· Block implementation of the operation of ordered filling and binary search 8;

· The job queue for the block implementing the operation of ordered filling and binary search 9;

· Block on the board 10.

The first physical page of memory 2 is connected with the implementation block of the hash function of the first type 6, the second physical page of memory 3 is composed of two areas, the first region 4 of the second physical page of memory 3 is connected with the implementation of the hash function of the second type 7, the second region 5 of the second physical memory page 3 is connected with the ordered filling and binary search block 8, job queue 9 is connected with the implementation of the ordered filling and binary search operation block 8, while the address table 1, the hash function blocks of the first 6 and the second 7 types, the job queue 9 and the implementation unit of the operation of ordered filling and binary search 8 are connected to the control unit 10.

The principle of operation of the address processing device of the local area network switch operating on the principle of a transparent bridge is as follows.

The address table 1 of the switch is physically divided into two pages of equal size 2 and 3. The filling of the first page 2 and the search for addresses in this page of the address table 1 are performed using a hash function of the first type. The hash function of the first type uses a hardware implementation of the Cycle Redundancy Code (CRC) algorithm, which maps a wide field of the station address in the network (for example, 48-bit) to a narrow field of the cell address in the address table (for example, 10- bit). The second physical page 3 is logically divided into two parts 4 and 5. The filling of the first half 4 of the second page 3 and the search for addresses in this part of the table are performed using a hash function of the second type. The hash function of the second type uses a hardware implementation of the algorithm for computing the cyclic redundancy code, which maps a wide field of the station address in the network (for example, 48-bit) to a narrow field of the cell address in the address table (for example, 9-bit). A hash function of the second type differs from a hash function of the first type in that it uses a different generating polynomial, and the bit depth of the result at its output is one bit less than the bit depth of the result at the output of the first hash function. The filling of the second half 5 of the second page 3 of the address table 1 is done in order of increasing station addresses. Addresses of stations in this part of the table are searched by binary search. The corresponding operations are implemented in the block of ordered filling and binary search 8.

The control unit 10 implements access control to the address table 1 from the side of the hash function implementation blocks of the first 6 and second 7 types and the ordered filling and binary search block 8, synchronizes the operation of the hash function implementation blocks of the first 6 and second 7 types, as well as the ordered block filling and binary search 8 and job queue 9. Also, the control unit 10 processes incoming requests from external devices for training or search for an address in the address table 1 and gives the external devices the results of the request processing. The control unit 10 can be implemented on a machine of finite state machines or on a microprocessor core.

Using two hash functions allows you to more evenly display the addresses of stations on the network in the addresses of the cells of the address table 1. Access to these areas is performed in one machine cycle.

Using the sequential filling and binary search areas allows the switch to work with some minimum guaranteed number of stations in the network, even when all these stations have the worst-case address selection with respect to both hash functions. The worst sample of addresses in this case is a sample of addresses in which all members of the sample lead to the same values of both hash functions, i.e. cause collisions. Accessing the ordered fill and binary search area is slower than it does when using hash functions. In this case, the access time increases linearly when filling out the table and logarithmically when searching for addresses in the table. Therefore, in order not to block the operation of devices from which address information is supplied to the switch address processing device, requests to this memory area are received through the buffer job queue, executed, for example, according to the principle of first-come-first-come out (FIFO - first input, first output) .

Both hash functions are used at the same time, and if a collision occurs on one hash function, then the second hash function passes control, if a collision occurs on the second hash function, control is transferred through the job queue to the block for implementing the operation of ordered filling and binary search 8. Management of the address processing device of the switch of the local area network is carried out by the control unit 10.

The advantage of the proposed device is that it allows you to quickly search for station addresses in the address table and quickly fill out the address table for random sets of station addresses and ensures that the switch can operate with a fixed minimum number of station addresses in the case of the worst selection of these addresses in the sense of hash functions. The minimum guaranteed number of station addresses is comparable to the size of the entire address table and is equal to a quarter of the size of the address table. For the worst sample addresses in the sense of hash functions, address processing is slower than for any other random sample. The maximum time spent searching for network addresses stored in the ordered filling and binary search areas grows only logarithmically for two in relation to the number of network addresses stored in this area of the address table.

Information sources

1. Olifer VG, Olifer N. A. Computer networks. Principles, technologies, protocols. - St. Petersburg: Peter, 2001 .-- 672 p.: Ill.

2. Understanding Address Hashing. Application Note AN002. Revision 1. Allayer Communications. August 2000. (www.allayer.com/pdf/an002.pdf)

3. Akritas A. Fundamentals of computer algebra with applications. Per. from English Pankratieva E.V. Publishing house "Mir". Moscow, 1994.

Claims (1)

An address processing device of a local area network switch operating on the principle of a transparent bridge, comprising an address table, a block for implementing a hash function of the first type, characterized in that it is equipped with a block for implementing a hash function of the second type, a block for implementing the operation of orderly filling in the address table and binary search of the address and the job queue for the implementation unit of the orderly filling of the address table and binary search of the address, as well as the control unit for chronization of the work of the implementation blocks of the hash function of the first and second types, and the implementation block of the operation of orderly filling the address table and binary address search, as well as for processing incoming requests from external devices and issuing the results of query processing to external devices, while the address table is divided into two independent physical memory pages, the first physical memory page is connected to the first type of hash function implementation block, the second physical memory page consists of two areas, the first area Torah physical memory pages associated with a block of the second type of hash functions, a second region of the second physical page associated with the memory block implementation of the operation table of addresses orderly filling and binary search addresses, job queue associated with implementation of the operation unit addresses orderly filling and binary search addresses.