Presentation is loading. Please wait.

Presentation is loading. Please wait.

Overview of Networking

Similar presentations


Presentation on theme: "Overview of Networking"— Presentation transcript:

1 Overview of Networking
Assembled by Charles Erwin 5/7/2019 CS 521 Computational Science

2 Agenda/Topics to Be Covered
Core Networking (Review) Topologies (Review) Distributed Systems Mobile Computing Summary 5/7/2019 CS 521 Computational Science

3 Core Networking The basic idea in networking is data transmission. This can be done via circuit switching or packet switching. For circuits: think of the old phone system. Connections are determined before transmission occurs. 5/7/2019 CS 521 Computational Science

4 Core Networking (cont)
Another way is packet switching. Think of this as the postal system: mail gets routed to the destination in different ways, but sometimes it gets lost. Most modern telephone networks now use packet switching (VOIP etc). 5/7/2019 CS 521 Computational Science

5 Core Networking (cont)
Packets are pieces of data that are transferred over networks. They hold both destination and usually point of origin. Packet Switched networks rely on the intermediate nodes to transfer the message. When a packet arrives to a node, if it is not the correct destination it is forwarded to the node’s best guess where it should go. Sometimes this isn’t a guess but often it is. 5/7/2019 CS 521 Computational Science

6 Core Networking (cont)
Now that we know what packets are, the question is how to transfer them? The answer is that protocols are developed that allow two computers how to understand each other and communicate effectively. Protocols must be well-defined in order to work. Most protocol definitions are widely available for implementation. 5/7/2019 CS 521 Computational Science

7 Core Networking (cont)
About the missing packets…packets can be lost in transit. IP (Internet Protocol) is only a best-effort protocol. That means delivery is attempted but not guaranteed. Naturally this sounds bad but there are several protocols, namely TCP, which try to provide reliable delivery, so not all hope is lost. 5/7/2019 CS 521 Computational Science

8 Core Networking (cont)
With so many protocols available, it would be easy for someone to become bogged down in trying to implement them correctly. In order to effectively implement protocols, the idea of layering proves to be a useful abstraction tool. 5/7/2019 CS 521 Computational Science

9 Core Networking (cont)
5/7/2019 CS 521 Computational Science

10 Core Networking Summary
The crux of understanding networking relies on abstraction and protocols. Hardware involved in Networking changes, so do protocols, so Networking is always a changing field. You can’t change the laws of physics! 5/7/2019 CS 521 Computational Science

11 Topology Basics Network Topology defines the shape of the graph.
For example: a Star topology focuses on a central hub 5/7/2019 CS 521 Computational Science

12 Topology Basics Summary
A topology only represents in the logical design of a network, not necessarily a physical setup. Nothing in life is free, TANSTAAFL so to speak, so every topology has its pros and cons (connectivity, points of failure etc). 5/7/2019 CS 521 Computational Science

13 An Example Topology (EFnet)
5/7/2019 CS 521 Computational Science

14 Distributed Systems A long time ago computer scientists determined that by using high speed networks, they could connect multiple computers to make one more powerful system. Hardware also progressed and designs for parallel systems, or systems that have more than one CPU, were manufactured as another way to increase computing power. 5/7/2019 CS 521 Computational Science

15 Distributed Systems (cont)
The result was that now multiple processor systems were classified as either loosely coupled or tightly coupled system. Tightly coupled systems share physical memory (ie parallel systems). Loosely coupled systems do not, meaning that memory sharing has to be achieved using other methods. 5/7/2019 CS 521 Computational Science

16 Distributed Systems (cont)
Why then use loosely coupled systems instead of tightly coupled systems? There are millions of reasons, all preceded by a dollar sign. While powerful multiprocessor systems offer great computing power, they are very expensive. So instead of paying millions of dollars for a powerful supercomputer you can buy 100 PCs for about $1,000 each and get a very powerful distributed system for $100,000 + a high speed network. 5/7/2019 CS 521 Computational Science

17 Distributed Systems (cont)
Parallel versus Distributed systems example: 5/7/2019 CS 521 Computational Science

18 Distributed Systems Summary
Why use Distributed systems? It is easier on the wallet. You can replace individual nodes to upgrade the system without having to take down the whole system. A broken mainframe = no processing. A down node on a distributed system doesn’t. Why not use Distributed systems? They have many consistency and timing issues (global clock problem) as well as being complex to administrate. Again TANSTAAFL rears its ugly head. 5/7/2019 CS 521 Computational Science

19 Mobile Computing Sometimes clients on a network do not want to be forced to remain in one location. Today many methods exist to allow nodes which do not stay in one location to remain connected. One method is probably well known to you: Cellular Networks. 5/7/2019 CS 521 Computational Science

20 Mobile Computing (cont)
A Cellular network relies on allocating channels to users within a grid system. Each part of the grid system (usually hex) is called a cell. While within a cell, users are allocated channels from a base station that is connected to some other type of network (satellite, fiber-optic). 5/7/2019 CS 521 Computational Science

21 Mobile Computing (cont)
There are many channel allocation algorithms available. It is important to effectively allocate channels because there are only so many channels available and many users who may want to connect. 5/7/2019 CS 521 Computational Science

22 Mobile Computing (cont)
Cellular is not the only type of Mobile network, there are also so called Mobile ad-hoc networks (MANET). These consist of self-configuring groups of nodes connected by wireless links. If you take the union of all of these connections you have an arbitrary topology. Ad-hoc networks have many challenges (low-bandwidth, high latency), but there are designs on how to implement common protocols on them. 5/7/2019 CS 521 Computational Science

23 References/Future Reading
Wikipedia: A lot of useful information on networking, good if you don’t have networking papers available. RFCs: Source for networking protocol designs: End-to-End argument, or why IP is best effort. Network Architecture for Challenged Internets 5/7/2019 CS 521 Computational Science


Download ppt "Overview of Networking"

Similar presentations


Ads by Google