TCP/IP data transfer protocol

The Internet, which is a network of networks and unites a huge number of different local, regional and corporate networks, operates and develops through the use of a single TCP/IP data transfer protocol. The term TCP/IP includes the name of two protocols:

• Transmission Control Protocol (TCP) - transport protocol;
• Internet Protocol (IP) is a routing protocol.

Routing protocol. The IP protocol ensures the transfer of information between computers on a network. Let's consider the operation of this protocol by analogy with the transfer of information using regular mail. In order for the letter to reach its intended destination, the address of the recipient (who the letter is to) and the address of the sender (from whom the letter is from) are indicated on the envelope.

Similarly, information transmitted over the network is “packed in an envelope” on which the IP addresses of the recipient’s and sender’s computers are “written”, for example “To: 198.78.213.185”, “From: 193.124.5.33”. Contents of the envelope on computer language called IP packet and is a set of bytes.

In the process of forwarding regular letters, they are first delivered to the nearest address to the sender. Postal office, and then are transmitted along the chain of post offices to the post office closest to the recipient. At intermediate post offices, letters are sorted, that is, it is determined to which next post office a particular letter should be sent.

IP packets on the way to the recipient computer also pass through numerous intermediate Internet servers on which the operation is performed routing. As a result of routing, IP packets are sent from one Internet server to another, gradually approaching the recipient computer.

Internet Protocol (IP) provides routing of IP packets, that is, delivery of information from the sending computer to the receiving computer.

Determining the route for information to pass through. The “geography” of the Internet differs significantly from the geography we are accustomed to. The speed of obtaining information does not depend on the distance of the Web server, but on the number of intermediate servers and the quality of communication lines (their capacity) through which information is transmitted from node to node.

You can get acquainted with the route of information on the Internet quite simply. Special program tracert.exe, which is included with Windows, allows you to track through which servers and with what delay information is transferred from the selected Internet server to your computer.

Let's see how access to information is implemented in the "Moscow" part of the Internet to one of the most popular search servers Russian Internet www.rambler.ru.

Determining the route of information flow

2. In the window MS-DOS session in response to the system prompt to enter the command.

3. After some time, a trace of information transfer will appear, that is, a list of nodes through which information is transmitted to your computer, and the time of transmission between nodes.

Tracing the route of information transmission shows that the server www.rambler.ru is located at a “distance” of 7 transitions from us, i.e. information is transmitted through six intermediate Internet servers (through the servers of the Moscow providers MTU-Inform and Demos). The speed of information transfer between nodes is quite high; one “transition” takes from 126 to 138 ms.

Transport protocol. Now let’s imagine that we need to send a multi-page manuscript by mail, but the post office does not accept parcels or parcels. The idea is simple: if the manuscript does not fit into a regular postal envelope, it must be disassembled into sheets and sent in several envelopes. In this case, the sheets of the manuscript must be numbered so that the recipient knows in what sequence these sheets will be combined later.

A similar situation often occurs on the Internet when computers exchange large files. If you send such a file as a whole, it can “clog” the communication channel for a long time, making it inaccessible for sending other messages.

To prevent this from happening, the sending computer must be set to big file into small parts, number them and transport them in separate IP packets to the recipient computer. On the recipient computer you need to collect original file from individual parts in the correct sequence.

Transmission Control Protocol (TCP), that is, the transport protocol, ensures that files are split into IP packets during transmission and files are assembled during reception.

Interestingly, for the IP protocol responsible for routing, these packets are completely unrelated to each other. Therefore, the last IP packet may well overtake the first IP packet along the way. It may turn out that even the delivery routes for these packages will be completely different. However TCP protocol will wait for the first IP packet and assemble the source file in the correct sequence.

Determining the time of IP packet exchange. Time of exchange of IP packets between local computer and the Internet server can be determined using the ping utility, which is included in the operating system Windows systems. The utility sends four IP packets to the specified address and shows the total transmission and reception time for each packet.

Determining the time of IP packet exchange

1. Connect to the Internet, enter the command [Programs-MS-DOS Session].

2. In the window MS-DOS session in response to the system prompt to enter the command.

3. In the window MS-DOS session The result of testing the signal in four attempts will be displayed. The response time characterizes the speed parameters of the entire chain of communication lines from the server to the local computer.

Questions to Consider

1. What ensures the holistic functioning of the global computer network Internet?

Practical tasks

4.5. Trace the route of information from one of the most popular Internet search servers www.yahoo.com, located in the “American” segment of the Internet.

4.6. Determine the time of exchange of IP packets with the www.yahoo.com server.

Internal routing protocol RIP This routing protocol is designed for relatively small and relatively homogeneous networks (Belman-Ford algorithm). The protocol was developed at the University of California (Berkeley), is based on developments by Xerox and implements the same principles as the routed routing program used in OS Unix (4BSD)

The RIP protocol must be able to handle three types of errors: Round-robin routes. Since the protocol does not have mechanisms for identifying closed routes, it is necessary to either blindly trust partners or take measures to block this possibility. To suppress instabilities, RIP should use a small value for the maximum possible number of steps (

The discrepancy between the routing table and the real situation is typical not only for RIP, but is typical for all protocols based on the distance vector, where information messages actualizations carry only pairs of codes: the address of the destination and the distance to it. Rice. An illustration explaining the occurrence of cyclic routes when using a distance vector.

Command field code values: Reserved for internal sun microsystem purposes. 5-6 Disable trace mode (obsolete);4 Enable trace mode (obsolete);3 Response containing information about distances from the sender’s routing table; 2 Request for partial or complete routing information; 1 Meaning of Command

DISADVANTAGES OF RIP: RIP does not work with subnet addresses. If the normal 16-bit Class B host ID is not 0, RIP cannot determine whether the non-zero portion is a subnet ID or a full IP address. RIP takes a long time to restore communication after a router failure (minutes). In the process of establishing a regime, cycles are possible. The number of steps is important, but not the only route parameter, and 15 steps is not the limit for modern networks.

OSPF Protocol (Dykstra's Algorithm) The OSPF (Open Shortest Pass First, RFC, RFC, algorithms proposed by Dykstra) protocol is an alternative to RIP as an internal routing protocol. OSPF is a route state protocol (the metric used is Quality of Service Factor). Each router has complete information about the state of all interfaces of all routers (switches) of the autonomous system. OSPF is implemented in the gated routing daemon, which also supports RIP and the external BGP routing protocol.

The OSPF routing table contains: destination IP address and mask; type of destination (network, edge router, etc.); function type (a set of routers for each TOS function is possible); area (describes the area whose connection leads to the goal, several entries are possible of this type, if the scope of the border routers overlaps); path type (characterizes the path as internal, interregional or external, leading to AS); price of the route to the destination; the next router to which the datagram should be sent; advertising router (used for inter-area communications and for autonomous systems to communicate with each other).

Advantages of OSPF: Each address can have multiple routing tables, one for each type of IP operation (TOS). Each interface is assigned a dimensionless price that takes into account throughput, message transport time. Each IP operation can be assigned its own price (quality factor). If equivalent routes exist, OSFP distributes the flow evenly across those routes. Subnet addressing is supported (different masks for different routes). With point-to-point communication, an IP address is not required for each end. (Saving addresses!) Using multicasting instead broadcast messages reduces the load on uninvolved segments. Disadvantages: It is difficult to obtain channel preference information for nodes supporting other protocols or with static routing. OSPF is only an internal protocol.

External BGP Protocol The BGP protocol (RFC-1267, BGP-3; RFC-1268; RFC-1467, BGP-4; , 1655) was developed by IBM and CISCO. The main goal of BGP is to reduce transit traffic. But not every host that uses BGP is a router, even if it exchanges routing information with a neighboring autonomous system's border router. BGP routers exchange route change messages

Format of BGP Route Change Messages The token field contains 16 octets and its contents can be easily interpreted by the recipient. The length field has two octets and specifies the total length of the message in octets, including the header. The type field is a message type code and can take the following values: (still alive) KEEPALIVE4 (attention) NOTIFICATIO N 3 (change) UPDATE2 (open) OPEN1

The following types of attribute type codes are available: ORIGIN (type code = 1) is a standard mandatory attribute that defines the origin of travel information. Generated autonomous system, which is the source of routing information. The attribute value in this case can take the following values: Incomplete - reachability information network layer obtained in some other way. 2 EGP - network layer reachability information was obtained using an external routing protocol; 1 IGP - network layer reachability information is internal to the originating autonomous system; 0 Description Attribute code

AS_sequence: yn" title="AS_PATH (type code = 2) is also a standard required attribute, which is composed of a collection of path segments. Each AS_PATH segment consists of three parts. AS_sequence: yn" class="link_thumb"> 22 !} AS_PATH (type code = 2) is also a standard required attribute that is made up of a collection of path segments. Each AS_PATH segment consists of three parts. AS_sequence: An ordered set of autonomous system routes in an UPDATE message. 2 AS_set: unordered set of routes in the update message; 1 Description NEXT_HOP segment type code (type code = 3) is a standard required attribute that specifies the IP address of the edge router that should be considered the next hop target on the path to the destination. MULTI_EXIT_DISC (type code = 4) is an optional intransitive attribute that occupies 4 octets and is a positive integer. The value of this attribute can be used to select one of several paths to a neighboring autonomous system. LOCAL_PREF (type code = 5) is an optional attribute occupying 4 octets. It is used by a BGP router to communicate to its BGP peers in its own autonomous system the degree of preference for an advertised route. ATOMIC_AGGREGATE (type code = 6) is a standard attribute that is used to inform peers to select a route that provides access to a broader list of addresses. . AS_sequence: up"> . AS_sequence: an ordered set of autonomous system routes in an UPDATE message. 2 AS_set: an unordered set of routes in an update message; 1 Description NEXT_HOP segment type code (type code = 3) - a standard required attribute that defines the IP address of the edge router, which should be considered the target of the next hop on the path to the destination. MULTI_EXIT_DISC (type code = 4) is an optional non-transitive attribute that occupies 4 octets and is a positive integer. The value of this attribute can be used when selecting one of several paths to neighboring autonomous system. LOCAL_PREF (type code = 5) is an optional 4-octet attribute. It is used by a BGP router to communicate to its BGP peers in its own autonomous system the degree of preference for an advertised route. ATOMIC_AGGREGATE (type code = 6) is a standard attribute that is used to inform partners to select a route that provides access to a wider list of addresses."> . AS_sequence: yn" title="AS_PATH (type code = 2) is also a standard required attribute, which is composed of a collection of path segments. Each AS_PATH segment consists of three parts. AS_sequence: yn"> title="AS_PATH (type code = 2) is also a standard required attribute that is made up of a collection of path segments. Each AS_PATH segment consists of three parts. AS_sequence: up"> !}

The BGP routing database consists of three parts: Contains information that the local BGP router has selected to distribute to neighbors using UPDATE messages. ADJ-RIBS-OUT: 3. Contains local routing information that the BGP router selected, based on routing policy, from ADJ-RIBS-IN. LOC-RIB:2. Remembers routing information that is received from update messages. This is a list of routes from which you can choose. (policy information base - PIB). ADJ-RIBS-IN: 1.

Features of BGP: BGP differs from RIP and OSPF in that it uses TCP as its transport protocol. A host that uses BGP is not necessarily a router. Messages are processed only after they have been fully received. BGP is a distance vector protocol. BGP regularly sends TCP messages to neighbors confirming that the peer is alive. If two BGP routers try to communicate with each other at the same time, this is called a collision, and one of the connections must be dropped. When routers communicate, an attempt is made to implement the highest protocol first; if one of them does not support this version, the version number is downgraded.

Dynamic routing protocols are designed to automate the process of building routing tables for routers. The principle of their use is quite simple: routers, using the order established by the protocol, send certain information from their routing table to others and adjust their table based on data received from others.

This method of constructing and maintaining routing tables greatly simplifies the task of administering networks that may undergo changes (for example, expansion) or in situations where any routers and/or subnets fail.

It should be noted that the use of dynamic routing protocols does not eliminate the possibility of “manually” entering data into router tables. Entries made in this way are called static, and entries obtained as a result of information exchange between routers are called dynamic. In any routing table it is always present, according to at least, one static entry is the default route.

Modern routing protocols are divided into two groups: vector-distance protocols and link-state protocols.

In vector-distance protocols, each router sends out a list of addresses of networks available to it (“vectors”), each of which is associated with a “distance” parameter (for example, the number of routers to this network, a value based on link performance, etc. .). The main representative of the protocols in this group is the RIP protocol (Routing Information Protocol).

Link-state protocols are based on a different principle. Routers exchange topological information with each other about connections in the network: which routers are connected to which networks. As a result, each router has a complete picture of the network structure (and this view will be the same for everyone), based on which it calculates its own optimal routing table. The protocol for this group is OSPF (Open Shortest Path First).

RIP protocol.

RIP (Routing Information Protocol) is the simplest dynamic routing protocol. It belongs to the vector-distance protocols.

Under a vector, RIP defines the IP addresses of networks, and the distance is measured in hops (“hops”)—the number of routers a packet must go through to reach a specified network. It should be noted that maximum value distance for the RIP protocol is 15, the value 16 is interpreted in a special way “network unreachable”. This determined the main drawback of the protocol - it turns out to be inapplicable in large networks where routes exceeding 15 hops are possible.

RIP version 1 has a number of significant practical use shortcomings. Important issues include the following:

• Estimatedka distances only taking into account the number of transitions. The RIP protocol does not take into account the actual performance of communication channels, which may be ineffective in heterogeneous networks, i.e. networks combining communication channels different devices,performance that use different network technologies.
• The Problem of Slow Convergence. Routers using the RIP protocol. They send out routing information every 30 seconds, and their work is not synchronized. In a situation where a certain router detects that some network has become unavailable, then in the worst case (if the problem was identified immediately after the next broadcast) it will notify its neighbors about this after 30 seconds. For neighboring routers everything will happen the same way. This means that information about the unavailability of a network can take a long time to spread to routers; obviously, the network will be in an unstable state.
• Broadcasting routing tables. The RIP protocol originally assumed that routers send information in broadcast mode. This means that the sent packet is forced to be received and analyzed at the link, network and transport levels by all computers on the network to which it is sent.

Partially these problems are solved in version 2 (RIP2).

OSPF protocol

OSPF (Routing (Open Shortest Path First)) is a newer dynamic routing protocol and is a link-state protocol.

The operation of the OSPF protocol is based on the use of a single database by all routers, which describes how and with which networks each router is connected. Describing each connection, communication routers
They add a metric to it - a value characterizing the “quality” of the channel. For example, 100 Mbps Ethernet networks use a value of 1, and 56 Kbps dial-up connections use a value of 1785. This allows OSPF routers (as opposed to RIP, where all channels are equal) to take into account real bandwidth and identify efficient routes. An important feature of the OSPF protocol is that it uses multicast rather than broadcast.

These features, such as multicast instead of broadcast, no restrictions on route length, periodic exchange of only short status messages, and consideration of the “quality” of communication channels allow the use of OSPF in large networks. However, such use can create a serious problem - a large number of routing information circulating in the network and increasing routing tables. And since the algorithm for finding efficient routes is quite complex in terms of computational volume, large networks may require high-performance and, therefore, expensive routers. Therefore, the ability to build efficient routing tables can be considered both an advantage and a disadvantage of the OSPF protocol.

Routing protocols.

Routing protocols (eg, RIP, OSPF, NLSP) should be distinguished from the actual network protocols (eg, IP, IPX). Both perform the functions of the network layer of the OSI model - they participate in the delivery of packets to the recipient through a heterogeneous composite network. But while the former collect and transmit purely service information over the network, the latter are intended to transmit user data, as protocols do link layer. Routing protocols use network protocols like vehicle. When exchanging routing information, routing protocol packets are placed in the data field of network layer or even transport layer packets, therefore, from the point of view of packet nesting, routing protocols should formally be classified as more high level than network.

Each router can support multiple networking and routing protocols.

Routing protocols determine the network topology and store information about it in the routing table. If a router does not use a routing protocol, it stores static routes or uses a separate protocol on each interface. Typically, routers operate with a single routing protocol.

Routing information contains a metric, that is, a measure of time or distance, and several timestamps. Forwarding information includes output interface data and address next system along the way. Routers typically store data about multiple possible next hop routers in a single table row.

Routing protocols perform two most important functions .

Firstly, with their help, the optimal path for transmitting a packet over the network is determined. Usually the path chosen is one that provides minimum time delivery with maximum reliability.

The routing protocol involves constantly collecting information about the state of routes and updating routing tables when the network topology changes due to failures or overloads. Thus, routing tables always contain accurate information about the network topology.

Secondly,The function of routing protocols is to transmit packets over a ,network. When receiving the next packet, the router reads the destination address from the packet header and determines in which direction (through which node) the packet should be further transmitted. Information from the routing table is used to make this decision.

The protocols used to create a routing table can be divided into three categories:

Distance vector length protocols;

Channel state protocols;

Routing Policy Protocols.

Vector Length Protocols- the simplest and most common type of routing protocol. Most of the protocols of this type used today originate from the Xerox Routing Information Protocol (sometimes they are even called that). Protocols of this class include RIP (TCP/IP stack), RIP (IPX/SPX stack), AppleTalk routing table management protocol RTMP and Cisco IGRP(Interior Gateway Routing Protocol).

Periodically, each router copies the destination addresses and metrics from its routing table and places this information in update messages sent to its neighbors. Neighboring routers compare the received data with their own routing tables and make the necessary changes.

This algorithm is simple and, as it seems at first glance, reliable. Unfortunately it works the best way in small networks with a (preferably complete) lack of redundancy. Large networks cannot do without periodic messaging to describe the network, but most of them are redundant. For this reason, complex networks have problems when communication links fail, since non-existent routes can remain in the routing table for a period of time. long period time. Traffic directed along this route will not reach its destination.

The second category of environment maintenance protocols consists of link state protocols. Link state protocols are more complex than vector length protocols. Instead, they offer a deterministic solution to the problems typical of their predecessors. Instead of broadcasting the contents of its routing tables to its neighbors, each router broadcasts a list of routers with which it has connections. direct communication, and directly connected to it local networks. This information about the state of the channel is sent in special announcements. With the exception of broadcasting periodic messages about its presence on the network, a router broadcasts link state announcements only when the link information changes or after a specified period of time has passed.

The disadvantage of channel state protocols such as OSPF, IS-IS and NLSP,is their complexity and high memory requirements. They are difficult to implement and require a significant amount of memory to store channel state announcements.

The third category of environment maintenance protocols includes routing rules protocols. If routing protocols based on vector length and channel state algorithms solve the problem of the most efficient delivery of a message to the recipient, then the routing problem is the most efficient delivery of a message to the recipient along allowed paths. Protocols such as BGP(Border Gateway Protocol) or IDRP(Interdomain Routing Protocol), allow Internet operators to obtain routing information from neighboring operators based on contracts or other non-technical criteria. The algorithms used for routing policy rely on vector length algorithms, but the metric and path information is based on a list of trunk operators.

To automatically build routing tables in composite networks, special service protocols are used - the so-called routing protocols. They can be implemented based on different algorithms that differ in the methods of constructing routing tables, methods of selecting the best route, and other features.

In previous editions of the "First Lessons" column on routing principles, it was assumed that the routing tables for each destination address indicate only the next (closest) router, and not the entire chain from the beginning to the end node. In this approach, routing is performed in a distributed manner - each router is responsible for selecting only one leg of the path, and the final route is the result of the work of all the routers through which a given packet passes. Such routing algorithms are called one-hop.

There is also the exact opposite, multi-step approach - Source Routing. In accordance with it, the source node indicates in the packet sent to the network the full route through all intermediate routers. This method does not require the construction and analysis of routing tables. This speeds up the packet's passage through the network and relieves the load on routers, but at the same time huge pressure falls on the end nodes. This scheme is used much less frequently than the distributed one-hop routing scheme.

Static Algorithms and Simple Routing

Depending on the method of generating routing tables, one-step algorithms are divided into three classes:

• fixed (or static) routing algorithms;
• simple routing algorithms;
• adaptive (or dynamic) routing algorithms.

In the first case, all entries in the routing table are static. The network administrator himself decides which routers should send packets with certain addresses, and enters the corresponding entries into the routing table manually (for example, using the route utility in UNIX or Windows NT).

The table is typically created during the loading process and edited as needed. Such fixes may be necessary, in particular, if a router in the network fails and its functions are transferred to another.

Tables are divided into single-route tables, in which one path is specified for each recipient, and multi-route tables, when several alternative paths are offered. In the case of multi-route tables, a rule for selecting one of the routes must be specified. Most often, one path is the main one, and the rest are backup ones.

Obviously, the fixed routing algorithm with its method of manually generating routing tables is only suitable for small networks with a simple topology. However, it can also be effectively used on highways large networks with a simple structure and obvious the best ways following packets in a subnet.

In simple routing algorithms, the routing table is either not used at all, or is built without the participation of routing protocols. There are three types of simple routing:

• random routing, when an arriving packet is sent in the first available direction other than the original one;
• flood routing, when a packet is broadcast in all possible directions except the original one (similar to how bridges process frames with an unknown address);
• routing based on accumulated experience, when the route is selected using a table, but the table is built in the same way as in the case of a bridge by analyzing the address fields of incoming packets.

ADAPTIVE ROUTING

The most widely used algorithms are adaptive (or dynamic) routing. They provide automatic update routing tables after changing the network configuration. Using adaptive algorithm protocols, routers can collect information about the topology of connections in the network and quickly respond to all changes in the configuration of connections. Routing tables usually contain information about the time interval during which a given route will remain valid. This time is called the route's Time To Live (TTL).

Adaptive algorithms are distributed in nature, i.e. there are no specially designated routers in the network for collecting and summarizing topological information: this work is distributed among all routers.

IN Lately There has been a trend to use so-called route servers: they collect routing information and then distribute it to routers upon request. In this case, the latter are either freed from the function of creating a routing table, or create only part of the table. Interaction between routers and route servers is carried out using special protocols, for example Next Hop Resolution Protocol (NHRP).

Adaptive routing algorithms must meet several important requirements. First of all, they are obliged to ensure the choice of, if not the optimal, then at least a rational route. The second condition is their essential simplicity, so that the corresponding implementations do not consume significant network resources: In particular, they should not generate too much computation or heavy overhead traffic. And finally, routing algorithms must have the property of convergence, that is, they must always lead to an unambiguous result in an acceptable time.

Modern adaptive routing information exchange protocols, in turn, are divided into two groups, each of which is associated with one of the following types of algorithms:

• distance vector algorithms (DVA);
• Link State Algorithm (LSA).

In distance vector algorithms, each router periodically and broadcasts a vector over the network, the components of which are the distances from this router to all networks known to it. Distance usually refers to the number of transit nodes. The metric may be different, taking into account not only the number of intermediate routers, but also the transit time of packets between neighboring routers or the reliability of paths.

Having received a vector from a neighbor, the router increases the distance to the networks specified in it by the length of the path to this neighbor and adds to it information about other networks known to it, which it learned about directly (if they are connected to its ports) or from similar advertisements of other routers, and then broadcasts the new vector value throughout the network. In the end, each router learns information about all the networks available in the interconnected network and the distance to them through neighboring routers.

Distance vector algorithms only work well in small networks. In large ones, they load communication lines with intense broadcast traffic. Configuration changes are not always processed using this algorithm correctly, since routers do not have an accurate idea of ​​the topology of connections in the network, but only have generalized information - a vector of distances - also obtained through intermediaries. The operation of a router in accordance with the distance vector protocol resembles the operation of a bridge, since such a router does not have an accurate topological picture of the network. The most common protocol based on the distance vector algorithm is the RIP protocol.

Link state algorithms allow each router to obtain sufficient information to build an accurate graph of network connections. All routers operate on the same graphs, making the routing process more resilient to configuration changes. “Broadcasting” (i.e., transmitting a packet to all the router’s closest neighbors) is performed here only when the state of connections changes, which does not happen so often in reliable networks. The vertices of the graph are both routers and the networks they connect. Information distributed over the network consists of a description of connections of various types: router-router, router-network.

In order to understand the state of the communication lines connected to its ports, the router periodically exchanges short HELLO packets with its closest neighbors. This service traffic also clogs the network, but not to the same extent as, for example, RIP packets, since HELLO packets have a much smaller volume.

Examples of protocols based on the link state algorithm are IS-IS (Intermediate System to Intermediate System) OSI stack, OSPF (Open Shortest Path First) TCP/IP stack and NLSP protocol of the Novell stack.

INTERNET STRUCTURE

Most of the routing protocols used in modern packet-switched networks originate from the Internet and its predecessor, the ARPANET. In order to understand their purpose and features, it is useful to get acquainted with the structure Internet networks, which left its mark on the terminology and types of protocols.

The Internet was originally built as a network connecting a large number of independent systems. From the very beginning, its structure featured a core backbone network, and networks connected to the backbone were considered as autonomous systems. The backbone and each of the autonomous systems had their own administration and routing protocols. It must be emphasized that the division into autonomous systems is not directly related to the division of the Internet into networks and name domains. An autonomous system unites networks where routing is carried out under the general administrative control of one organization, and the name domain is common for computers (possibly belonging to different networks), in which the assignment of unique symbolic names occurs under the same guidance. Naturally, the scope of the autonomous system and the name domain may in a particular case coincide if one organization performs both of these functions.

Routers used to form networks and subnets within an autonomous system are called interior gateways, and those that connect autonomous systems to the network backbone are called external gateways. The network backbone is also an autonomous system. All autonomous systems have a unique 16-bit number, which is assigned centrally by the appropriate Internet administrative authority.

The routing protocols used within autonomous systems are called Interior Gateway Protocol (IGP), and the protocols for exchanging routing information between external gateways and backbone gateways are called Exterior Gateway Protocol (EGP). Any internal IGP protocol can also operate within the backbone network.

Dividing the entire Internet network into autonomous systems is necessary for multi-level modular organization, without which it is impossible to significantly expand any large system. Changing routing protocols within any autonomous system should not affect the operation of other autonomous systems. In addition, dividing the Internet into autonomous systems facilitates the aggregation of information at backbone and external gateways. Internal gateways can use sufficiently detailed interconnection graphs for internal routing to select the most efficient route. However, if information of this level of granularity were stored in all routers on the network, then the topological databases would grow so large that gigantic memory would be required, and the time required to make routing decisions would become unacceptably long.

Therefore, detailed topological information remains within the autonomous system, which external gateways present to the rest of the Internet as a single entity. They communicate minimally about the internal composition of the autonomous system necessary information- the number of IP networks, their addresses and the internal distance to these networks from a given external gateway.

The CIDR classless routing technique can significantly reduce the amount of routing information transmitted between autonomous systems. Thus, if all the networks within an autonomous system begin with a common prefix, say 194.27.0.0/16, then the external gateway of the autonomous system should advertise only this address, without separately advertising the existence of, for example, the network 194.27.32.0/ within the autonomous system. 19 or 194.27.40.0/21, since these addresses are aggregated into the address 194.27.0.0/16.

The Internet structure with a single backbone shown in Figure 1 has been such for a long time, so a protocol for exchanging routing information between ASes, called EGP, was developed specifically for it. However, as networks of service providers have evolved, the structure of the Internet has become much more complex, with arbitrary connections between autonomous systems. Therefore, the EGP protocol gave way to the BGP protocol, which allows you to recognize the presence of loops between autonomous systems and exclude them from intersystem routes. The EGP and BGP protocols are used by Internet service providers only on the external gateways of autonomous systems. Corporate network routers run internal routing protocols such as RIP and OSPF.

RIP and OSPF

The RIP protocol (Routing Information Protocol) is an internal distance vector routing protocol. This is one of the earliest protocols for exchanging routing information, and is still extremely common due to its ease of implementation. In addition to the RIP version for TCP/IP networks, a RIP version is also available for Novell's IPX/SPX networks. The RIP protocol for IP comes in two versions: first and second. RIP v.1 does not support masks, i.e. it distributes between routers only information about network numbers and distances to them, and does not send out information about the masks of these networks, considering that all addresses belong to standard classes A, B or C. The RIP v.2 protocol transmits information about network masks, so it is more responsive to today's requirements.

The OSPF (Open Shortest Path First) protocol was adopted in 1991. As an implementation of the link state algorithm, it was designed for use in large heterogeneous networks. The computational complexity of the OSPF protocol grows rapidly as the network size increases, that is, the number of networks, routers, and connections between them increases. To solve this problem, the OSPF protocol introduces the concept of a network “area” (not to be confused with the Internet Autonomous System). Routers belonging to a certain area build a graph of connections only for it, which reduces the dimension of the network. No link information is passed between areas, but edge routers exchange specific information about the addresses of the networks located in each area and the distance from the edge router to each network. When transmitting packets between areas, one of the area border routers is selected, namely the one with a distance to the desired network less. When transmitting addresses to another area, OSPF routers aggregate multiple addresses using a common prefix into one.

OSPF routers can accept address information from other routing protocols, such as RIP, which is useful for operating in heterogeneous networks. This address information is processed in the same way as external information between different areas.

Natalya Olifer is a LAN columnist. She can be contacted at: