4 Network Fundamentals

Prof. Bhushan Trivedi

Introduction

Now we start our journey on networks and begin with fundamentals. Our approach is little different. We have seen how networks based on TCP/IP works and now we will be looking at nuts and bolts. This is different than most other approaches but not uncommon and found superior to conventional methods1. It is found that people end up learning more when they clearly perceive the need to learn what they are learning. You will be able to understand these fundamentals from clear perspective of where are they be useful to us.

The ingredients

The network is an interconnection of networking devices if you ask a layman. The word networking device (and not computer) is, used with a purpose. We are entering into an era where many devices other than computers are becoming part of the network. We call them internet of things and we will look at them in the modules 34, 35 and 36. Many non-computer devices like smart phones, access control devices located in many offices, peripherals like printers and scanners are now able to act as an independent node in networks. Thus networks are no longer confined to computers.

However, we need little better definition and thus we will have to dig little deeper. Let us start.

The first ingredient that we need is some interconnecting media. That could be wired or wireless. Another point, the computing devices which communicate with each other are working independently and have no control (usually) over others. Thus they take a decision on their own. We call these devices nodes. These nodes, work in autonomous fashion, even though are part of the network and work also as per the networking discipline. Thus a node, which is independent in deciding many things like when and what to send, also follow a typical structure of the packet for the receiver to understand it unambiguously. The interconnection that we have just meant to be used for sharing information and not for controlling devices. Thus if we say that there is an interconnection between a device A and device B, that interconnection is used for communication between A and B. This interconnection is not used for A or B to control the other.

A critical component of the network is the protocol. When the devices are independent of each other, and if they want to communicate, they must agree on some common ways to

1 One of the interested areas of the author is “Effective Teaching”, apart from writing and publishing few papers on pedagogy, he has conducted 25 workshops on the said subject across India.do so. That demands a kind of protocol where everyone is brought to the same platform. The protocol does not only help the nodes to communicate but bring the discipline in the entire process and tries to bring solutions to common problems commonly related to multiple communicating parties like arbitration. We have looked at SMTP as an application layer protocol earlier. The Network communication should only take place under the control of some protocol. The protocol has to clearly identify the commands used in communication and possible responses to a specific commands issued by other communicating parties. Sometimes the commands are also called requests. Most of the Internet protocols are request-response protocols, where both involved parties alternatively send requests and respond back.

If you have a query why a protocol is needed, what if two different devices interpret your command differently? For example, the requester might ask them to compile files, and one device collects files while other device tries to run a compiler for it. Both are valid interpretations and thus we need to have consensus to what ‘compile’ mean. That is the reason we need protocols.

Definition

Author’s definition of a network is as follow. It is more than sufficient for us to begin with. “A collection of computing devices, autonomous in nature, interconnected by wired or wireless means and governed by a set of standards in order to share data and resources”.

Network Categories

Networks can be categorized in a few ways. Here is the list

1.      Scope, based on the question “what is the span of the network”

2.      Connection, how nodes are connected to each other, wired or wireless

3.      The utility, sensor networks, IOT, home networks etc.

4.      Communication, broadcast or point to point

5.      Structure, Adhoc, and fixed topology

Scope

The networks can be of a limited scope, for example, a PAN or a Personal Area Network is confined to a single laptop or desktop and a few devices like phones, printers, headphones and other small devices, typically in a range of 10 meters. Bluetooth is a common standard used in practice. The wireless version, WPAN (wireless PAN) is the one which is used. Other protocols like ZigBee for mesh networks and Body Area Network which is a collection of wearable computing devices are also quite popular. Normally one of the nodes is the master and rest are slaves. Master (which is usually the laptop) is also connected to other networks (like the Internet) usually.

LAN or Local Area Network is the network which is usually confined to 2 to 3 kilometers and owned by a single party. LANs are the most common type of networks as there are quite a few advantages of using such networks.

1. All nodes and connectors are of same quality and capacity as there is a single owner. Thus there is no likelihood of incompatible devices getting connected.

2. A single and consistent network-wide policy can be designed and implemented.

3. Filtering can be done at the periphery. For example, a LAN might have a device which filters all spam emails at the entry point only. The packets which are unwanted or does not follow security policy designed by the administrator can be filtered out. Similarly, user requests for websites not allowed by organization policy are also filtered out. This process is known as gateway level filtering.

4. When the devices and connectors are homogeneous, there is less likelihood of congestion and little need for flow control.

LAN nodes are usually connected using a STAR topology, that means there is a central point, a server, with a device which helps it connect it to many other nodes, thus forming a star-like structure. Wireless networks, logically, also forms the start structure. In past, there were few other topologies were used but none of them are used currently.

Metropolitan Area Networks (MAN)

There was a proposal to connect all nodes of the city together and get many benefits like, help citizens book movie tickets, or book seats in the restaurants, video on demand services and so on. IEEE designed a standard (802.6.1). At the same point in time when this was being proposed, the Internet started services of a similar kind with no restriction geographically and thus MANs died in infancy. As they were confined to cities, they are little bigger in scope as compared to LANs. One of the MAN protocols which are gaining popularity but no longer confined to a city is called Wi-Max which is basically a Wireless MAN standard for communication by IEEE.

Wide Area Network (WAN)

Wide Area Networks or WANs were a little bigger version of MAN and had a little bit of success. For example, India had a WAN known as ERNet (Education and Research Network) and Inflibnet (Information and Library Network). ERNet used to connect academic and research institutes while Inflibnet used to connect libraries. With the advent of Internet, some networks join Internet and thus making their services available to everybody and not only confined to their registered users while some other died of silent death. Looking at the scope, WANs were very large and in some cases country wide while in some cases even spread to places continents apart.

The Internet

The Internet, the Internet that we all are used to, is of the widest scope. Some critical points citing the difference of the Internet vs. the LAN is as follows.

1. the network is provided as a service, one pays and gets the service. The user does not own the network infrastructure; he only owns his own machine which connects to the Internet.

2. There is an issue of speed mismatch which plagues the Internet due to heterogeneous devices and connectors which are not the case in LAN.

3. More likelihood of delays, bandwidth issues, no quality of service with the Internet. With LAN managed by a single owner, it is possible.

There are some attempts to provide LAN-like services to the internet users but there is no significant headway yet.

Let us also clarify one more critical point. WWW (word wide web) is not the Internet. The internet is just the connection infrastructure which connects the members. WWW is a service running on top of it which provides information in the form of linked documents.

Connection

We can connect members using either wired or wireless way. In either case, the transmission is done using Electro-Magnetic Waves (EMW) only. The EM waves do not require a media to propagate and thus wired or wireless communication does not alter the nature of transmission from that angle. Having mentioned that, media plays a big role in data transmission rate. We will study that later in module 7.

Wired connections use Unshielded Twisted Pair (UTP) cables or FO (Fiber Optic) cables. Wireless connections are broadly divided into two categories, one type of connections based on Mobile Service Providers, for example, 3G and 4G networks provided by mobile service providers like AirTel and Reliance. Another is based on wireless LAN and MAN protocols like Wi-Fi and Wi-Max. Both of them are quite different but the difference is diminishing and would soon disappear by the time 5G arrives.

One of the fundamental differences between wired and wireless devices that a wired device usually communicates with exactly one other device while a wireless device inherently broadcasts its signals. This is, obviously, due to their physical structure as wires connect two devices while wireless is basically sending signals in the vicinity so whoever in the vicinity will be in a position to receive. This demands clear strategies for arbitration in wireless networks. In the wireless network like GSM or Wi-Fi, if multiple senders start sending at the same point of time and using the same frequency, the signals collide, produce garbage and it will only result in chaos. Some wireless networks allow collisions and have a strategy of “retransmit after random time if there is a collision”. On the contrary, most other wireless networks provide different frequencies and fix timeslots for users to make sure no two users collide at any given point of time. While one typical solution based on CDMA (Code Division Multiple Access) is smart enough to manage multiple signals without collision. Let us look at some more details of wired and wireless networks.

Wired networks

Some wired networks, most notably the first version of Ethernet, still being used at places, used to broadcast but most of the others send point to point messages.

Wired networks used two types of cabling, the copper-based UTP-5 and UTP-6 are most commonly used cables for low-end solutions. High-end solutions prefer fiber optic cables. The way the fiber optic cables are becoming cheaper and newer applications demand (online gaming for example) greater bandwidth, FO may be a preferred choice for low-end applications soon. Ethernet is the de-facto standard for wired LAN today. It comes in many varieties, fast Ethernet (100 Mb) is so far the most popular, but Gigabit (1 Gb) and 10G (10 Gigabit) are the choices of users today. 40/100 Gigabit version is on the horizon and thus it is the choice of the users who would like to be future proof.

FO cables have a much better bandwidth (in terms of terabytes) so they are better. There are other advantages of FO cables which we will discuss when we study physical layer. The world bandwidth, for the time being, can be assumed as the maximum amount of data which can be carried through that media. The text demands much lesser bandwidth, usually in KBs, while the video demands much higher bandwidth, usually in 10s of MBs to 100s of MBs (depending on the length and type of compression used).

Apart from cables, there are many other parameters which differentiate wired networks. The first version of Ethernet allowed any user to transmit as and when he deems fit. This simple method leads to problems when the probability of multiple senders sending is high. When multiple senders send together, multiple signals result into another signal which is garbage and nobody gets anything out of that. That phenomenon is known as collision and it was a very difficult problem. Later versions of Ethernet did away from that and used methods where there is no broadcast and all connections are the point to point.

Wireless networks

Wireless networks help user escape the wire and remain on the go. That itself is the motivation for such networks. The first such network was ALOHA built by computer center at Hawaii islands by the University of Hawaii. Ethernet first version was designed by drawing inspiration from it. We have 802.11 which people call Wi-Fi and 802.16 which is more popularly known as Wi-Max. However, though those words are used as synonyms and we might freely do that in future, it is important to note that 802.11 and Wi-Fi or 802.16 and Wi-Max are not truly equal. Wi-Fi and Wi-Max are standards from alliances of the vendors while 802.11 and 802.16 are standards from IEEE. When the vendors started implementing the standards by IEEE and build products like access points and wireless cards, the critical issue was to have interoperability. For example, an access point from vendor A must work with a wireless card from vendor B. So vendor alliances got together and formed their own standards which were known as Wi-Fi and Wi-Max standards respectively. However, there is no harm in using the slang to address these networks. Wi-Fi is an acronym of Wireless Fidelity and Wi-Max is an acronym of Worldwide Interoperability for Microwave Access. Wi-Fi is basically a wireless LAN while Wi-Max is a wireless MAN. We will look at both of them in due course.

Another important player in the wireless fray is known as MANet or Mobile Ad hoc Network. As the name suggests, it is a type of ad hoc network, which in turn is a subset of wireless networks. When the users form networks without any fixed topology, for example, when multiple laptops connect together, or other mobile devices (including current smartphones) connect with such laptops, they are ad hoc in nature, that means they do not follow any topology. Additionally, they are dynamic in nature, nodes come and join, go and leave, thus a network has a dynamic membership. Such networks are called MANets. Not only the membership but the way communication happen in MANet is also dynamic, as there are no special purpose routers available, the intermediate nodes shoulder responsibility of routing the packets. As there is a chance of them moving as well, there is no fixed route for any communication. The nodes which are close proximity, called neighbors, may change any moment and our assumptions about routing which we elaborated in first three modules, are no longer acceptable. Thus MANets throw many challenges and thus demand special solutions. The key attributes of MANets are nodes connected in peer to peer fashion, no central authority which demands distributed solutions, and dynamic nature. Two major problems with such networks, apart from dynamism are, power consumption and security as they are using batteries and may come into contact with some malicious node.

Whatever challenges they pose, MANets are high in demand as they offer quite unique services, for example, accessing a fellow student’s (or co-fliers, or a roommates) files is impossible otherwise.

Sensor Networks and IOT devices based networks

When tiny devices like car sensors form a network, or Implantable Medical Devices (IMD) form their own networks (for example, Glucose Sensors, Insulin Pumps, Pace Makers etc.), they are called sensor networks. Sensors used for wildlife monitoring, or sensing fire in forests, landslide detection, measuring the quality of water, measuring the condition of devices, and many other things. The latest entrant is the one where these nodes are additionally accessible through the web and thus addressing them becomes easy. That type is popularly known as IOT or internet of things and is a buzzword today.

Such devices throw much more challenges than the MANets as they have weaker batteries, deployed in hostile conditions and much larger in number. Handling them is equally challenging. Obviously, they also help in getting information not possible otherwise. For example, a lab supervisor does not need to go and check the health of every instrument of the lab, the IOT device or a sensor can report that to his laptop. A farmer does not need to find where his cattle is, the sensor attached to a body of each of the animal is able to send information and the farmer can locate each one of them just sitting in his hut.

Communication

How the network communicates differentiate them in two categories, broadcast and point to point. Broadcast networks broadcast their messages and all the connected nodes listen. Whichever node is the actual receiver, usually mentioned in the broadcast message, responds back while others ignore that message. On the contrary, a point to point network directly communicates with the receiver and usually through some intermediary.

While broadcasting, the sender sends messages to everybody in the network, popularly known as a broadcast. The receiver looks at the destination address, which is part of the message, and decides whether to accept or rejects a message based on whether he is the recipient of the message or not. On the contrary, point to point is another method deployed by the physical layer, in that case, it only sends the information to the actual recipient and not everybody in the vicinity. Most wireless networks follow broadcasting while most wired networks follow point to point communication but it is not always so.

Broadcasting is usually preferred with small networks. A node does not need to know exactly where the recipient is. It will just broadcast the message. Out of all who received, the intended recipient will accept and others ignore the message. Simple. The point to point method demands exact address of the recipient so the sender can send and intermediaries can route the message. This is little more complicated but better for large networks which cannot afford to broadcast.

Broadcasting wastes lots of bandwidth and the legitimate users’ time who are not actually involved in transmission. Point to point operation is little more sophisticated in nature and does not disturb others so other nodes can continue communicating at the same point in time.

In some cases, the message is for everybody, for example, a new router may like to announce its presence to all nodes of the network or an incoming node to this network is looking for some help. In such cases, broadcasting is better. In most other cases, point to point communication is better.

In the case of wireless networks, the communication is inherently broadcast and thus broadcast communication is natural for them. As the point to point networks save huge bandwidth and only feasible solution for most but a few small networks, they are used despite being more complicated. Look at references [1] and [2] for more details.

Utility

By design, networks are generic and used for many purposes. Remote database access, software and hardware maintenance, fire sharing, social networking, file upload-download, remote access to accounts, e-commerce, and e-banking all are possible to be done using conventional networks. However, two types of networks are quite distinct. One we have already discussed, the sensor networks and second, home networks for smart homes and now also for conventional homes. Interestingly, with the advent of IOT devices, many other types of networks are possible, for example, vehicular networks which connects sensors of different vehicles and communicates with central servers for traffic related solutions, IMD (Implantable medical devices) and other measuring devices in the clinic for getting accurate and periodic measurements of the patients in an automatic manner, sensors which manage health of machines in a factory and so on. Parking sensors and car sensors to communicate and advice how to park or calculate fine for wrong parking and sending it to the owner automatically, measuring different parameters for the entire city like water consumption and electricity consumption etc. and help citizen make decisions on that data are a few examples.

Why we consider these networks special, because they demand things which normal, conventional networks do not require. These networks are more resource constrained, have less memory and less computing power. These networks are also having many physical constraints; for example, the body sensors cannot be designed for a short term use. They must use their power in most optimistic ways. Their design also requires passing human safety measures.

There are many challenges and questions that are yet to be answered on these special networks. Consult the references [1] and [2] for further discussions and some examples.

Components of a network

Let us briefly look at the components of the network. When we connect our computing devices to a network, these components help run the network and data transfer. We will look at network cards, communication lines, intermediary devices like hubs, switches and routers in this part.

Network Interface Card

The first component is the Network Interface Card or Network Interface Controller. Earlier it used to come as an expansion card which we need to install on the motherboard slot. Nowadays the NIC come built into the motherboard. NIC is also known as the LAN adapter or network adapter. The ubiquitous Ethernet card or wireless card are examples of LAN adapters.

Cable

An IO cable is one which connects a NIC to a plug that is housed on the wall. The plug usually contains another cable which carries it to the switch of some sort. Both cables are of the same type. As we have seen before, most used cables are UTP and FO.

Frequency band

While wired networks need cables, wireless networks need frequency bands to communicate. The frequency band is the range of frequencies for communication. As a thumb rule, more the number of frequencies in the band, more the data rate. Most wireless networks including GSM and Wi-Fi, allocate specific frequency band of specific width to every sender. The sender uses that frequency band for its communication exactly like a wired network uses a cable to communicate.

Frequency allocation is more critical than attaching a cable as the wireless transmission is open to all and thus must be regulated. One cannot use a frequency assigned to another communicator. Elaborate schemes for frequency allocation and usage is provided by government regulating authorities like DoT (Department of Telecommunication) in India and FCC (Federal Communications commission) in the US.

Interconnecting devices

All servers and nodes are interconnected by some intermediaries. These intermediaries range from seemingly dumb devices (called repeaters and amplifiers) which only improve the signal and intelligent devices (called routers or intelligent switches) which can choose, on the fly, the outgoing line to send the incoming packet, based on dynamic criteria designed by network administrators. Let us talk about each one of them.

Repeaters and amplifiers

An amplifier just increases the strength of the signal while the repeater reshapes the signal. Amplifier works with the analog signal while repeater works with the digital or the square signal. They are usually sitting on a path which is long enough to weaken the signal. When the signal is very weak, the receiver will not be able to interpret exactly what is sent by the sender. Figure 4.3 shows the function of a repeater. The transmission media distorts the square signal. When such distorted signal is fed into a repeater, it reshapes it. Both, the amplifier and the repeater works at the physical layer.

Hubs and layer2 switches

Repeater and amplifier are one to one connectors. Hub connects one input to multiple outputs. It is no longer used by network administrators today but was quite popular some time back. It just broadcasts an incoming signal to all outgoing connections, known as ports. Usually, the hub has a specified number of ports, in multiple of 2, which allows that many devices connect through the hub. For example, we can have an 8-port hub or 16-port hub where there are 8 or 16 total connections. When we have a hub, for example, an 8-port hub, any one port can transmit and all other (7 in this case) ports it is received by devices connected. That is basically broadcasting. Figures 4.4 and 4.5 depicts the difference between a hub and a switch.

Another device, quite similar to it, called switch, has a simple improvement over the hub, it can send a message to a specific port rather than broadcasting it. Unlike hubs, this device does a complicated job of extracting a destination address of each packet, find out the port number where that device is connected to, and transfer the packet over that port. If that destination port is busy, it might even store that packet meanwhile. It is possible that two different ports are sending at the same point of time to different destinations. This way, the switch can support multiple connections in parallel and it can provide improvement about an order of magnitude over hubs. More ports, better the improvement. Switches used to be more expensive earlier than hubs due to this built-in intelligence but not so now and thus nobody is using hubs any longer.

These switches are called data link layer switches or layer-2 switches as they segregate traffic based on layer-2 or data link layer (or physical) addresses. When a switch segregate traffic using layer-3 (IP addresses) or layer-4 (TCP in one direction, UDP in another), or layer-5 (SMTP in one direction, FTP in another, Web traffic in the third direction and so on), they are called switches of those specific layers.

Router and layer-3 switch

Router works at the IP layer (or network layer or layer-3). It looks at the IP address of the incoming packet and decides where to send it. A switch can also be used in place of a router. The difference is, for every connection, the switch is going to route the packets the same, and thus do the job much faster than a router. Unlike that, routers deliberate on every packet and make an individual decision for each packet based on current network conditions. Thus routers act autonomous and help to thwart congestions. Both routers and switches are used in practice, routers are used when traffic is unpredictable, for example, between two networks of different companies, while switches are used when not so. For example, between company branches. A typical type of layer-3 switch which is routing traffic based on a typical tag is known as MPLS (multi-protocol label switch). Here the traffic is switched based on those tags which are fixed and thus for fixed tags the traffic is routed consistently.

Summary

The networks can be categorized in many ways, based on scope, transmission mechanism, wired or wireless way to connect, utility etc. are major ways. Different network components, like cables, Network adapters and intermediaries are important for us to learn.

References

1. Computer Networks by Bhushan Trivedi, Oxford University Press
2. Data Communication and Networking, Bhushan Trivedi, Oxford University Press