COMPUTER NETWORKING:TECHNICAL BACKGROUND

TECHNICAL BACKGROUND

To understand the possibilities of the TCP / IP-based networks fully, it is important to know how they work and what kinds of technical solutions lie behind their services.

Architecture of the Internet

As mentioned in the previous section, the Internet can be described as a network of local area networks (see Figure 4). Therefore, perhaps the most important component of the Internet is LANs. LANs connect computers or, in other words, hosts, and are connected to other LANs by gateways and communication lines. Thus, the four basic logical components of the Internet (or of intranets) are:

1. Hosts

2. Local area networks

3. Gateways

4. Communications lines

The main role of gateways is to provide connections between the communication lines and the local area networks and to route the data packets toward their destinations. This is why they are often called routers. Gateways also play an important role in security by protecting the local area network against external attacks and other illegal or unauthorized access. Quite often, a security zone, referred to as the DMZ, is set up where network traffic is passed through a gateway into a firewall (see Section 15.2). The firewall then provides application-level security to network traffic before sending the information through an internal router to pass to end systems in the intranet.

In most cases, communication lines establish steady, 24-hour connection between their endpoints. In order to increase reliability, and sometimes also for traffic load balancing reasons, alternative routes can be built between the local area networks. If any of the lines is broken, the gateways can auto- matically adapt to the new, changing topology. Thus, the packets are continuously delivered by the alternative routes. Adaptive routing is possible at gateways, particularly for supporting 24 X 7 op- eration. It commonly occurs in the backbone of the Internet.

Packet Switching

Packet switching is one of the key attributes of the TCP / IP technology. The data stream belonging to a specific communication session is split into small data pieces, called packets. The packets are delivered independently at the target host. The separated packets of the same communication session may follow different routes to their destination. In contrast to line-switching communication tech- nologies, in packet switched networks there is no need to set up connections between the commu- nicating units before the start of the requested data transmission. Each packet contains all of the necessary information to route it to its destination. This means that packets are complete from a network perspective.

A good example of line-switched technology is the traditional public phone service. In contrast to packet switching, line switching assumes a preliminary connection setup procedure being per- formed before a conversation starts. After the connection is set up, an individual communication channel (called a circuit) is provided for the data transmission of that communication session. When the data transmission is over, the connection should be closed.

Most Important Protocols

A network protocol is a set of rules that determines the way of communication on the network. All the attached hosts must obey these rules. Otherwise they won’t be able to communicate with the other hosts or might even disturb the communication of the others. For proper high-level commu- nication, many protocols are needed.

The system of protocols has a layered structure. High-level protocols are placed in the upper layers and low-level protocols in the lower layers. Each protocol layer has a well-defined standard

interface by which it can communicate with the other layers, up and down. TCP / IP itself is a protocol group consisting of several protocols on four different layers (see Figure 5). TCP and IP are the two most important protocols of this protocol group.

IP is placed in the internetworking layer. It controls the host-to-host communication on the net- work. The main attributes of IP are that it is connectionless, unreliable, robust, and fast. The most astounding of these is the second attribute, unreliability. This means that packets may be lost, dam- aged, and / or multiplied and may also arrive in mixed order. IP doesn’t guarantee anything about the safe transmission of the packets, but it is robust and fast. Because of its unreliability, IP cannot satisfy the needs of those applications requiring high reliability and guaranteed QoS.

Built upon the services of the unreliable IP, the transport layer protocol, TCP, provides reliable communication for the applications. Reliability is guaranteed by positive acknowledgement and au- tomatic retransmission. TCP performs process-to-process communication. It also checks the integrity of the content. TCP is connection oriented, although the TCP connections are virtual, which means that there is no real circuit setup, only a virtual one. A TCP connection is a reliable, stream-oriented, full-duplex, bidirectional communication channel with built-in flow control and synchronization mechanisms.

There are also routing and discovery protocols that play an important role in affecting network reliability, availability, accessibility, and cost of service.

Client–Server Mechanism

Any communication on a TCP / IP network is performed under the client–server scheme (see Figure 6). When two hosts are communicating with each other, one of them is the client and the other is the server. The same host can be both client and server, even at the same time, for different com- munication sessions, depending on the role it plays in a particular communication session. The client sends requests to the server and the server replies to these requests. Such servers include file servers, WWW servers, DNS servers, telnet servers, and database servers. Clients include ftp (file transfer) programs, navigator programs (web browsers), and telnet programs (for remote access). Note that the term server does not imply special hardware requirements (e.g. high speed, or large capacity, or continuous operation).

The typical mode of operation is as follows: The server is up and waits for requests. The client sends a message to the server. The requests arrive at particular ports, depending on the type of the expected service. Ports simply address information that is acted upon by the serving computer. Clients provide port information so that server responses return to the correct application. Well-known ports are assigned to well-known services. Communication is always initiated by the client with its first message.

Because multiple requests may arrive at the server at the same time from different clients, the server should be prepared to serve multiple communication sessions. Typically, one communication

session is served by one process or task. Therefore, servers are most often implemented on multi- tasking operating systems. There is no such requirement at the client side, where only one process at a time may be acting as a client program.

One of the main features in the new concept of computing grids is the substitution of the widely used client–server mechanism by realizations of a more general distributed metacomputing principle in future computer networking applications (see Section 12.3).

Addressing and Naming

In order to be unambiguously identifiable, every host in a TCP / IP network must have at least one unique address, its IP address. IP addresses play an essential role in routing traffic in the network. The IP address contains information about the location of the host in the network: the associated number (address) of the particular LAN and the associated number (address) of the host in the LAN. Currently, addresses in the version 4 Internet protocol are 32 bits long (IPv4 addresses) and are classified into five groups: A, B, C, D, and E classes. Class A has been created for very large networks. Class A networks are rare. They may contain up to 224 hosts. Class B numbers (addresses) are given to medium-sized networks (up to 65,534 hosts), while class C network numbers are assignedto small (up to 254 hosts) LANs.

The 32-bit-long IP addressing allows about 4 billion different combinations. This means that in principle, about 4 billion hosts can be attached to the worldwide Internet, by using IPv4 addressing.

However, intranets, because they are not connected to the Internet or connected through firewalls,may use IP addresses being used by other networks too.

Because the number of Internet hosts at the start of the third millennium is still less than 100 million, it seems that the available IPv4 address range is wide enough to satisfy all the needs.

However, due to the present address classification scheme and other reasons, there are very serious limitations in some address classes (especially in class B). The low efficiency of the applied address distribution scheme has led to difficult problems. Because of the high number of medium-sized networks, there are more claims for class B network numbers than the available free numbers in this class.

Although many different suggestions have been made to solve this situation, the final solution will be brought simply by implementing the new version of the Internet protocol, IPv6, intended to be introduced early in the 2000s. It will use 128-bit-long IP addresses. This space will be large enough to satisfy any future address claims even if the growth rate of the Internet remains exponential.

There are three different addressing modes in the current IPv4 protocol:

• Unicast (one to one)

• Multicast (one to many)

• Broadcast (one to all)

The most important is unicast. Here, each host (gateway, etc.) must have at least one class A, class B, or class C address, depending on the network it is connected to. These classes provide unicast addresses.

Class D is used for multicasting. Multicast applications, such as radio broadcasting or video conferencing, assign additional D class addresses to the participating hosts.

Class E addresses have been reserved for future use.

Class A, B, and C addresses consist of three pieces of information: the class prefix, the network

number, and the host number. The network number, embedded into the IP address, is the basic information for routing decisions. If the host number part of the address contains only nonzero bits (1s), the address is a broadcast address to that specific network. If all the bits in the host number part are zeroes (0s), the address refers to the network itself.

In order to make them easier to manage, IP addresses are usually described by the four bytes they contain, all specified in decimal notation, and separated by single dots. Examples of IP addresses are:

• Class A: 16.1.0.250

• Class B: 154.66.240.5

• Class C: 192.84.225.2

The human interface with the network would be very inconvenient and unfriendly if users had to use IP addresses when referring to computers they would like access. IP addresses are long numbers, making them inconvenient and difficult to remember and describe, and there is always a considerable risk of misspelling them. They don’t even express the type, name, location, and so on of the related computer or the organization operating that computer. It is much more convenient and reliable to associate descriptive names with the computers, the organizations operating the computers, and / or the related LANs. The naming system of the TCP / IP networks is called the domain name system (DNS). In this system there are host names and domains.

Domains are multilevel and hierarchical (see Figure 7). They mirror the organizational / adminis- trative hierarchy of the global network.

At present, top-level domains (TLDs) fall into two categories:

1. Geographical (two-letter country codes by ISO)

2. Organizational (three-letter abbreviations reflecting the types of the organizations: edu, gov, mil, net, int, org, com)

Second-level domains (SLDs) usually express the name or the category of the organization, in a particular country. Lower-level subdomains reflect the hierarchy within the organization itself.

The tags of the domain names are separated by dots. The first tag in a domain name is the hostname. The tags are ordered from most specific to least specific addressing.

The domain name service is provided by a network of connected, independently operated domain name servers. A domain name server is a database containing information (IP address, name, etc.) about hosts, domains, and networks under the authority of the server. This is called the domain zone.

Domain name servers can communicate with each other by passing information retrieved from their databases. If a host wishes to resolve a name, it sends a query to the nearest domain name server, which will provide the answer by using its local database, called a cache. However, if nec- essary, it forwards the query to other servers. The Internet domain name service system is the largest distributed database in the world.