The router is a device of reference model OSI network level, using one or more metricss for definition of an
optimum way of the network traffic transfer on the basis of a network level information. From this definition
follows, that the router is necessary for definition of the further way of the data sent to a greater and complex network.
The user of such network sends the data in a network and specifies the address of the subscriber. Data pass on
a network and in points with a branching of routes act on routers which just and are established in such points. The
router chooses the further best way. What way is better, is defined by quantity indicators which refer to as metricss.
The best way is a way with the least metrics. In the metrics some parameters can be considered, for example, length of a
way, time of passage .
Routers divide into devices of the top, average and bottom classes.
High-efficiency switching routers of the top class serve for association of computer networks. They support set of
reports and interfaces. Devices of the given type can have up to 50 ports for association of local and global networks.
By means of middle class routers less large network associations of the enterprise scale are formed. The standard
configuration includes two-rub ports of local networks and from four up to eight ports global networks. Such routers
support the most widespread protocols of routing and transport protocols.
Routers of the bottom class intend for local networks of small offices. They connect small offices with a network of the
enterprise. A typical configuration: one port of a local network (Ethernet or Token Ring) and two ports of the global
network calculated on low-speed allocated lines. Nevertheless, similar routers are in great demand at managers it
is necessary for them to expand available gateway associations.
Routers for backbones and the removed offices have different architecture as different functional and operational
demands are made to them. Routers of backbones necessarily should have modular structure. It enables to increase
productivity of the equipment with the minimal expenses. Routers of class SoHo, contain only one processor operating
work of three or four interfaces. In them are used about the same reports, as in routers of base networks, however the
software is more directed on simplification of installation and operation.
The router of a backbone consists of following basic components: network adapters and the operating processor. After
receipt of a package on the interface module it analyzes the address of purpose and accepts commands of the operating
processor for definition of target port. Then the software package of the basic highway of a router is transferred in the
interface module serving for communication with the addressed segment of a local or global network.
The basic purpose of routers is a routing of the traffic of a network. Process of routing can be divided into two
hierarchically connected levels:
Level of routing. At this level there is a work to the table of routing. The table of routing serves for definition of
the address (a network level) a following router or directly addressee to an available address (a network level) and the
addressee after definition of the address of transfer the certain target physical port of a router gets out. This process
refers to as definition of a route of moving of a package. Adjustment of the table of routing is conducted by reports of
routing. At the same level the list of necessary given services is defined;
Level of transfer of packages. Before transferring a package, it is necessary: to check up the control sum of
heading of a package, to define the address (a channel level) the addressee of a package and to make directly sending a
package in view of sequence, fragmentations, a filtration, etc. These actions are carried out on the basis of the commands
acting from a level of routing.
Definition of a data route transmission occurs by program. Corresponding software carry names of routing reports.
The logic of their work is based on algorithms of routing. Algorithms of routing calculate cost of delivery and choose a
way with smaller cost. The elementary algorithms of routing define a route on the basis of the least number of
intermediate (transit) units on a way to the addressee. More complex algorithms in concept "cost" pawn some parameters,
for example, a delay by transfer of packages, throughput of liaison channels. The basic result of work of algorithm of
routing is creation and support of the table of routing in which all routing information enters the name. The maintenance
of the table of routing depends on the used report of routing.
Generally the table of routing contains the following information:
- The Valid addresses of devices in a network;
- The Service information of the report of routing;
- Addresses of the nearest routers.
The Basic requirements shown to algorithm of routing are:
The Optimality of a choice of a route;
Simplicity of realization;
Stability;
Fast convergence;
Flexibility of realization.
The optimality of a choice of a route is key parameter of algorithm that does not demand explanatories.
Algorithms of routing should be simple in realization and use as small as possible resources
Algorithms should be steady against refusals of the equipment on originally chosen route, to high loadings and mistakes
in construction of a network.
Convergence is a process of the coordination between routers information on topology of a network. If the certain event
in a network leads to that some routes become inaccessible or there are new routes, routers dispatch messages on it each
other on all network. After reception of these messages routers make reassignment of the optimum routes, hundred the new
stream of messages in turn can generate. This process should come to the end, and is fast enough, differently in network
topology there can be loops, or the network in general can cease to function. Algorithms of routing should consider
quickly and correctly changes in a condition of a network (for example, refusal of unit or a segment of a network).
Algorithms of routing can be:
- Static or dynamic;
- One routing or multirouting;
- Single-level or multilevel;
- Inside domain or between domain;
- Single-address or group.
For static (not adaptive) algorithms routes get out in advance and are brought manually in the table of routing where the
information on is stored on what port to send a package with the corresponding address. The reports developed on the basis
of static algorithms, name not routing. As examples not routing protocols can serve LAT (Local
AreaTransport, the transport report for channel areas) firms DEC, the report of connection of the terminal and NetBIOS.
Usually with these protocols bridges as they do not distinguish protocols of a network level work.
At use of dynamic algorithms the table of routing is automatically updated at change of topology of a network or the
traffic in it. Dynamic algorithms differ on a way of reception of the information on a condition of a network, time of
change of routes and used parameters of an estimation of a route.
One-routing protocols define only one route. It not always appears optimum. Multirouting algorithms offer some routes to
the same addressee. Such algorithms allow to transfer the information on several channels simultaneously, that means
increase of throughput of a network.
Algorithms of routing can work in networks with single-level or hierarchical architecture. In a single-level network all
its fragments have an identical priority, that, as a rule, is caused by similarity of their functional purpose. The
hierarchical network contains subnets (fragments of a network). Routers of the bottom level serve for communication of
fragments of a network. Routers of top level form the special part of a network named by a highway (a basic part).
Routers of the main network transfer packages between networks of the bottom level.
The hierarchical structure in greater and complex networks allows to simplify considerably managerial process by a
network, facilitates isolation of segments of a network, etc. For example, logic isolation of segments of a network
supposes installation of fireproof walls.
Some algorithms of routing operate only within the limits of the domains (inside domain routing), and others - both
within the limits of the domains, and in adjacent with them (interdomain routing). In this case the domain means area
of routing in which works one or several reports. In different domains different reports work. If communication of
domains is necessary, interdomain routing is used.
Single-address algorithms of routing are intended for transfer of the concrete information (on one or several routes)
only to one addressee. Multiple-address (or group) algorithms are capable to transfer the information to many addressees
simultaneously.
When the router receives a package, it reads out the address of purpose and defines, on what route to send a package.
Usually routers are stored by data about several possible routes. The choice of a route depends on several factors,
including:
- Applied system of measurement of length of a route (its metrics);
- The routing high level protocol;
- Topology of a network.
At a level of routing there are three basic groups of reports of routing (division into groups is defined by type of sold
algorithm of definition of an optimum route):
- Protocols of a distance vector;
- Protocols of the channel condition;
- Protocols of routing policy;
Protocols of a vector of distance - the most simple and the most widespread. Protocols of the given group include RIP IP,
RIP IPX, AppleTalk and Cisco IGRP. This type of the protocol has received the name from a way of information interchange.
The router with the certain periodicity takes addresses of addressees of the information and the metrics from the table
of routing and places these data in messages dispatched to neighbours on updating. The next routers are verified by
obtained data with own tables of routing and make necessary changes. After that they dispatch messages on updating. Thus,
each router receives the information on routes of all network. At obvious simplicity of algorithm to speak about its full
reliability it is impossible. It can work effectively only in small networks. It is connected by that the stream of
messages between routers sharply increases in large networks.
Protocols of the channel condition for the first time have been offered in 1970. These protocols it is much more complex,
than protocols of a vector of distance. Instead of dispatch to neighbours of contents of the tables of routing, each router
carries out a multicasting of the list of routers with which it has direct communication, and list directly the local
networks connected to them. This information is a part of the information on a condition of the channel. It is dispatched
in special messages. Besides the router dispatches messages on a condition of the channel only in case of its change or
after the set interval of time. Protocols of a condition of the channel are difficult in realization and require a
significant memory size for storage of the information on a condition of channels. As examples of these protocol serve
OSPF, IS-IS, Nowell NLSP and Cisco EIGRP.
Protocols of policy of routing concern to the third group of protocols. These protocols most effectively solve a problem of
delivery to the addressee of the information. This category of protocols is used at routing in Internet and allows operators
to receive the information on routing from the next operators on the basis of special criteria. That is during an exchange
the list of the resolved routes (ways) is developed. Algorithms of policy of routing lean on algorithms of a vector of
distance, but the information on routes is based on the list of operators of network Internet. As examples of that
category protocols of can serve BGP and EGP.
All aforesaid concerned to a level of routing. The level of transfer of packages is realized on algorithms of switching
and, as a rule, is identical to the majority of protocols of routing. The intermediate router, having the address of a
following router, sends it a package addressed specially on the physical address (MAC-LEVEL) of this router, but with
the address (a network level) the addressee. To the address of the addressee the router defines, whether it knows how to
transfer a package to a following router in a way. If knows, the package is sent a following router by replacement of the
physical address of the addressee by the physical address of a following router. If the router does not know, the package
is ignored. On the following router all repeats. In process of passage of a package through a network, its physical
address varies, but the address of a network level remains constant.
The primary goal of packages level transfer is a switching of packages from different users. The general scheme
of transfer of packages is those: one of possible transit units (this information acts from a level of routing on which
it is calculated to the address of the addressee) gets out, the target heading of a channel level is formed and carried
out send a package. Besides at this stage the fragmentation of packages, check of the control sum can be made.
Routers (is more exact - a level of routing) work at a network level of reference model OSI. The level of promotion of
packages functions at a channel level.
Work at a network level allows to make intellectual processing of packages. As routers basically work with report IP,
they should keep in contact without creation of logic connection between subscribers. Thus each package is processed and
sends to the addressee independently.
Manufacturers at creation of routers use three basic types of architecture:
- Uniprocessor;
- Strengthened uniprocessor;
- Symmetric multiprocessing.
|