Скачать презентацию Setting up the Communication Network Problem Wade Trappe Скачать презентацию Setting up the Communication Network Problem Wade Trappe

9a497c0073d33b4b3ebdd11c111d5733.ppt

  • Количество слайдов: 22

Setting up the Communication Network Problem Wade Trappe Setting up the Communication Network Problem Wade Trappe

Lecture Overview l What is a communication network? – Core Questions l Telephone Networks Lecture Overview l What is a communication network? – Core Questions l Telephone Networks – PSTN/GMSC/IGE/LE/PBX and all that stuff – Circuits and Routing, aka. the phone number – End Systems l l Transmission Systems Switching: Overview

Through the Looking Glass: Communication Networks l This class is not about specific protocols, Through the Looking Glass: Communication Networks l This class is not about specific protocols, but rather about the fundamentals underlying networks. – When you use the hypertext transfer protocol (http) or send an email on the Internet, there are many operations (“the fundamentals”) which are hidden from the protocol itself l A web page might be slow, but what goes on “underneath” that makes it slow? – Perhaps you are on a shared medium Ethernet and the slowness is due to backoffs and collision resolution… – Perhaps you are performing satellite communication and a sunflare has increased local radiation… resulting in a higher bit error rate for the underlying signaling… necessitating frequent retransmissions l These are the “underlying fundamentals” of communication networks

A Core Dump of Core Questions l 1. There are several questions that will A Core Dump of Core Questions l 1. There are several questions that will arise in our study of communication network fundamentals: How does one model the application? What are the salient properties of the application that affect operation of a communication network? e. g. Bit rate, traffic pattern, packet size, delay sensitivity, interarrival times, reliability requirements 2. What models are the most appropriate for studying the network performance in different scenarios? e. g. Use random process models (good for basic understanding of the fundamentals involved in network protocols) Flow models (e. g. on/off): good for capturing and studying real-time application behaviors 3. How does one manage simultaneous communications over a shared resource between different users or pairs of users? e. g. This arises in many different ways: switching (to be circuit-oriented or not to be? ), multiple access protocols (TDMA, FDMA, Aloha, etc) 4. How does one build a network system from the ground up? e. g. The idea of modular construction, aka. layering. Shannon loved it and enjoyed it as the Law of Digital Communications. The Law is basically the same in networking. Interestingly, both Laws are changing now!!! (research hint!)

Network Types l There are two basic classes of communication networks: circuitswitched and packet-switched Network Types l There are two basic classes of communication networks: circuitswitched and packet-switched l For the most part, now days, we think of packet-switched networks – This is because the concept behind packet switching (which we shall discuss later) has led to more “engineering efficiency” – In particular, circuit-switching seeks to reserve “dedicated” resources for a communication, whereas packet-switching is more “opportunistic” l We shall primarily discuss issues related to packet networks for the most of this class (some techniques will apply to circuitswitched networks) l However, we will start with circuit-switched networks

Grandfather of Networks: Telephone Network l Public switched telephone networks have been around for Grandfather of Networks: Telephone Network l Public switched telephone networks have been around for a long time l Goal: Provide voice service between two users, regardless of their (global) location l The service is known as POTS (Plain Old Telephone Service) l The term “switching” refers to the fact that we want to connect any users without requiring a separate wire for each possible pair l Example: In this class there are roughly 20 students. If each one of you wanted to connect to every other person with a dedicated line, we would need 20*19= 190 total connections!!! l The idea behind switching is to avoid this naïve approach to communication: – We have 1 connection line going into each house, and these lines will connect to a switching/signaling backbone that will route your call to the appropriate destination l Let us look at a generic “phone network”

Telephone Network Generica Telephones LE Cellular GMSC Network LE LE National IGE PSTN LE Telephone Network Generica Telephones LE Cellular GMSC Network LE LE National IGE PSTN LE PBX LE LE IGE Digital Interconnection Circuits PSTN LE LE PSTN = Public Switched Telephone Network GMSC= Gateway Mobile Switching Center IGE= International Gateway Exchange LE= Local Exchange PBX= Private Branch Exchange

Telephone Network Explained l Telephones at home (or a small office) connect directly to Telephone Network Explained l Telephones at home (or a small office) connect directly to the nearest Local Exchange l Phones located in a corporate office typically connect to a private switching office (Private Branch Exchange) l Think of the PBX as administering a micro-phone universe where any two phones directly connected to the PBX can have an easy connection to each other via the PBX l The PBX are connected to an LE so that calls may be routed outside of the PBX l Cell Phone networks are a small universe and phone calls made within the cell network are administered by the MSC, while phone calls leaving the cellular universe pass through the Gateway Mobile Switching Center l Finally, international calls are routed through International Gateway Exchanges, which are connected by digital connections

The Life Cycle of a Phone Call Backbone Network D E A B C The Life Cycle of a Phone Call Backbone Network D E A B C Local Exchange @ Central Office l End systems (phones) connect to the LEs, which connect to backbone switches l #LEs >> # Backbone Switches l The backbone network is nearly fully connected (dedicated lines between almost all switches)… making a one-hop network

Life Cycle of a Call, pt 2 l When an End User makes a Life Cycle of a Call, pt 2 l When an End User makes a call, it connects to its LE, which seeks to set up a “circuit” between two end systems – To do so, if the call is not local, it connects to the nearest backbone switch, which connects to the switch nearest the target end user’s LE – The target LE then connects its target End user to the circuit that has been set up l Question: So how does the system know which LEs and switches to connect to? l Answer: Its all in the phone number!

867 -5309: What’s that number? l A call going from 732 -445 -0611 to 867 -5309: What’s that number? l A call going from 732 -445 -0611 to 873 -867 -5309 creates a circuit by: 1. Identifying the end systems area code, so the LE at 9732) notices that the area code (873) is different from its own, so it must connect out 2. It establishes a connection with the nearest backbone swtich 3. The backbone switch establishes a (short) connection to the switch servicing the (873) area 4. The (873) switch establishes a connection with the -867 - local exchange 5. The final connection to the end system 5309 is made l That is, the telephone number serves as a means to route through the electomechanical switches of the telephone network l The telephone numbers form a natural hierarchy that is easily extendable to include new numbers: some central agency simply creates new area code numbers l Components: End System, Transmission, Switching, Signaling

Transmission System l l A transmission link is characterized by its information-capacity, the propagation Transmission System l l A transmission link is characterized by its information-capacity, the propagation delay, and its link attenuation Information capacity: Bandwidth is the width of the data pipe, or more specifically, the average number of bits/second. Link Delay: The time taken for a signal to propagate over the medium and is particularly important for long links with delay sensitive applications – Example: Speed of light in fiber is 70% speed of light in a vacuum. In fiber, light travels at 8 msecs/mile – Voice application requirements < 100 ms for non-frustrating conversations – New. York San. Francisco is 20 msec (2500 miles). Not as much of the delay is propagation, so switching and control architectures are important – Satellite: speed of light is higher, but the propagation delay is around 250 msec (36000 kilometers!) Link Attenuation: As a signal travels, it attenuates and it is important to introduce regeneration/amplification on the links. Fiber optics are good as they have minimal attenuation

Switching l Switching governs how a user is connected with every other user l Switching l Switching governs how a user is connected with every other user l Two components: Switch Hardware (Data Plane), and the Switch Controller (Signaling/Control Plane) l A switch transfers information from an input line to an output line. l There are two basic ways to do switching: Space division switching and time division switching l Signaling: Is the decision plane that controls the switches and which establishes how the switches will operate and forward their calls (setting up and tearing down the calls)

Space Division Switching Example Cross-Bar: – Inputs arrive along rows and outputs are connected Space Division Switching Example Cross-Bar: – Inputs arrive along rows and outputs are connected to columns – To perform the connection, the switch establishes the circuit connection at the intersection – To visualize, recall that this is electro-mechanical. Input l A B C D E

Time-Division Switching l l l N inputs are stored in a temporary buffer The Time-Division Switching l l l N inputs are stored in a temporary buffer The switch reads from the buffers N times faster according to a schedule Writes to the outputs before next input buffer is read A Read 1 B C Write 2 3 D

Packet Switching: A brief overview, pg. 1 l Circuit Switching provides a continuous, constant Packet Switching: A brief overview, pg. 1 l Circuit Switching provides a continuous, constant bit rate connection between two points l By doing so, circuit switching implicitly provides quality of service guarantees: (1) A guaranteed bandwidth; (2) a bound on delay once a circuit is established l Problem with circuit switching from a resource allocation point of view: – Once a circuit is formed, those resources are dedicated, regardless of whether they are being used! – Example: (Phone call) There are many instants during a conversation when silence occurs and no “data” is being created. In a circuit-switched network, where the connection is reserved, resources are wasted

Packet Switching: A brief overview, pg. 2 l Packet switching (i. e. store-and-forward switching) Packet Switching: A brief overview, pg. 2 l Packet switching (i. e. store-and-forward switching) addresses these issues – Note: The difference between packet switching and message switching is where the packetization is done l There are two types of packet switching: l Connection-oriented (Virtual-Circuit Based): Session causes the creation of a path (virtual circuit) much like circuit switching, but the capacity of each link is shared dynamically (e. g. with some scheduling policy) with other sessions that use the same link l Connectionless (Datagram Based): Here, each packet contains its source and destination address, as well as payload. The packet and the network are responsible for finding the packet’s way to the destination. Here, intermediate nodes participate in “dynamic routing”, possibly taking advantage of local information to decide the best next step in the delivery l We will look at each of these a little more.

Connection-Oriented VCI-2 A VCI-1 PSE-2 1 3 2 1 PSE-3 2 1 PSE-1 3 Connection-Oriented VCI-2 A VCI-1 PSE-2 1 3 2 1 PSE-3 2 1 PSE-1 3 2 1 3 PSE-4 l CO = Connection oriented l VCI = Virtual Circuit Identifier l PSE = Packet Switching Exchange VCI-3 VCI-4 B

Connection-Oriented l To set up a virtual circuit, the source sends a call request Connection-Oriented l To set up a virtual circuit, the source sends a call request control (signal) packet to its PSE. Signal contains source and destination address as well as a label for this component of the virtual circuit (called a VCI) l Each PSE contains a table that specifies the outgoing link that should be used to reach network address l The PSE uses this destination address to lookup which outgoing link should be used and assigns a new VCI for this link l The routing table is updated l The call request packet is then forwarded to the next PSE and the process continues

A Connection-Oriented Routing Table In Out l VIC 1 – Link 1 PSE-2 Routing: A Connection-Oriented Routing Table In Out l VIC 1 – Link 1 PSE-2 Routing: VCI 2 – Link 1 VCI 3 – Link 3 VCI 2 – Link 1 PSE-3 Routing: VCI 3 – Link 1 VCI 4 – Link 2 l VCI 2 – Link 2 VCI 3 – Link 3 l PSE-1 Routing: VCI 1 – Link 1 VCI 2 – Link 2 l VCI 3 – Link 1 Call clear packets are forwarded to tear down connection.

Connectionless (datagram) l Here, the establishment of an explicit connection is not required. l Connectionless (datagram) l Here, the establishment of an explicit connection is not required. l Rather, a datagram is routed to an appropriate outgoing link based on the local routing table. R 2 A R 1 R 3 R 4 Packet: B A Payload B

Wrap-up l Connection-oriented Examples: – X. 25: Old style file transfer network – ATM: Wrap-up l Connection-oriented Examples: – X. 25: Old style file transfer network – ATM: high bit rate “backbone” style network l Connectionless: the Internet l Packet switching is the more popular style of network l Regardless of which style of network, the process of communication involves protocols, which we will discuss next time. – i. e. OSI and the PHY-layer