Networking Introduction
Computer Networking is the process of interconnecting computer systems via telecommunication methods to exchange data and share resources. Networks are everywhere!
The internet is a network of networks that connects billions of devices. It consists of hosts, conencted to LANs, WANs, and the Internet. They exchange messages through packet switches (a link-layer switch or a router), via communication links (cables, wireless local area links, satellite). Eventually the message reaches its destination. A route is the sequence of switches that a packet goes through. A series of protocols controls the sending and receiving of information to and from end systems and packet switching.
1. Packet Switching
The internet is a packet switched network - information is transmitted in packets (a formatted unit of data), and switches / routers operate on individual packets by forwarding them to other switches or end systems. Every forwarding decision is taken on the basis of the information contained in the packet.
Circuit Switching uses an established dedicated communication path between two devices. It requires an expensive set-up phase, and requires a dedicated link between any two devices, hence it cannot be used on the internet. However, unlike packet switching, it doesn't suffer from processing cost of forwarding and space overhead on packets.
2. Communication Protocols
End systems and packet switches run protocols. A communication protocol is an agreement between communicating parties on how communication is to proceed. It consists of:
- A handshake: establishes the identities and / or the context.
- A conversation: free-form exchange.
- A closing: terminates the conversation.
It is a set of rules, a complete and unambiguous executable specification to account for all possible situations. It must also define the necessary message formats.
3. 5-Layer Hybrid Model
Protocols can rely on other protocols (e.g. HTTP relies on Internet Connection), forming a protocol stack. We have 5 generic layers:
- Application Layer: supports network applications.
- Transport Layer: supports end-to-end communication.
- Network Layer: supports packet forwarding.
- Link Layer: supports link-level communication.
- Physical Layer: supports physical transmission.
Why was it deisgned this way?
When designing the OSI model, the designers paid attention to:
- Addressing - how to denote the intended recipient.
- Error Control - how to detect (and fix) transmission errors.
- Flow Control - how to manage the rate of transmission.
- Multiplexing - how to support multiple communications in parallel.
- Routing - how to route packets to their destination.
Most network layers offer one or both of the following services:
- Connection Oriented: first establish a connection, then exchange data and finally release the connection.
- Connectionless: the data is put into an envelope, which contains the destination address. The envelope is then sent to the destination.
A service is a set of primitives that a layer provides to the layer above it. A protocol is a set of rules that prescribe the layout and meaning of packets, and the order in which specific packets can be sent.
3.1 Application Layer
Defines applications functionality and message formats:
- Traditional protocols like DNS, SMTP, FTP and HTTP/HTTPS.
- Modern protocols such as middleware protocolsto support destributed systems (e.g. Java RMI, Google Protocol Buffers).
- High Level protocols to support applications (e.g. WebRTC, WebSockets).
- Peer to Peer protocols (e.g. BitTorrent, Skype).
3.2 Transport Layer
The transport layer is about observation, offering connection oriented and connectionless services. It provides the actual network interface to applications:
- May use sockets (UNIX, Windows).
- Allows to set up a connection to reliably deliver data in the order it was sent.
- Provides flow control to prevent the sender from overwhelming the receiver.
- Often supports secure connections.
- May support datagrams (unreliable, unordered packets).
3.3 Network Layer
The network layer describes how routing and congestion is to be done:
- How do we find which computers / routers are in the network?
- How do we find the best route from A to B?
- What happens if a router goes down?
- Should multicasting / broadcasting be supported?
- What happens if a router becomes congested?
- Can we detect and avoid hot spots?
3.4 Link Layer
The link layer is about getting the data into the LAN. To do this, it must:
- Detect bit transmission errors (e.g. by adding redundant bits for error detection), like a parity bit or a checksum.
- Specifies how a number of computers can share a common channel by using MAC addresses.
4. Quantifying Data Transfer
Bandwidth is the amount of information that can get into or out of the connection in a time unit.
Throughput is the amount information that does get into or out of the connection in a time unit. At steady state (zero traffic acculumation), the input throughput is equal to the output throughput.
Goodput is the amount of useful information that gets into or out of the connection in a time unit. It is the throughput minus the overhead, such as headers or reattempts.
Latency - the time it takes for 1 bit to go through the connection.
Or, in mathematical terms:
- Latency (propagation delay):
, or distance over wave propagation speed. - Throughput (link bandwidth):
or transferred bits over duration. - Packetization (transmission delay):
. - Transfer Time:
.
Every time a packet hops, another it adds router delay. This consists of processing delay (check bit error, determine output link), which is typically negligible, and queueing delay (time waiting at ouptut link for transmission). This depends on the congestion level of the router.
Traffic intensity depends on