380 likes | 475 Views
CSE 331: Introduction to Networks and Security. Slide Set 2. Fall 2001 Instructor: Carl A. Gunter. Issues. Information movement implies: An information source An information destination A path from source to destination Addresses are used to locate source and destination
E N D
CSE 331: Introduction to Networks and Security Slide Set 2 Fall 2001 Instructor: Carl A. Gunter
Issues • Information movement implies: • An information source • An information destination • A path from source to destination • Addresses are used to locate source and destination • Path can be static or dynamic
Paths Are Made of Links • Links are interconnected by zero or more network elements, e.g., switches, routers, hubs, bridges, etc. • Links have delay and throughput • Path delay is sum of link delays plus switching delays • Path throughput = bottleneck link t’put
Some Link Types • Wired Ethernet LAN (10, 100 Mbps) • SONET fiber (up to 9.6 Gbps) • CATV (usually 1-6 mbps, asymmetric) • ISDN (64 Kbps*n with bonding) • POTS (56 Kbps) • Wireless Ethernet (2, 11 Mbps) • Infrared IrDA (115 Kbps to 4 Mbps) • CDPD (19.2 Kbps)
Making a network • Rules for interconnecting links • Rules for interpreting addresses • Link properties versus end-to-end properties • e.g., ATM guarantees in-order arrival • e.g., TCP makes a reliable stream from datagrams
Together, rules make a protocol • Protocols are logically layered • Peer-to-peer, corresponding layers • Higher layers constructed using lower layer protocols • Examples: OSI and IP
Requirements • Connectivity • Cost-effective resource sharing • Functionality • Performance
Direct Links Point to Point Multiple Access
Hosts Network Elements Network “Cloud” Links Switched Networks
Internetworks Router (Gateway)
Addressing Unicast Broadcast Multicast Anycast Size Local Wide Metropolitan Desk Further Distinctions
Sharing Resources • How can hosts share the network if they want to use it at the same time? • How can they share links? • How can they share switches?
Multiplexing Strategies • Two approaches • Synchronous time-division multiplexing (STDM). • Statistical multiplexing. • Tradeoffs • STDM can ensure service for a complete communication, but may waste resources. • Statistical multiplexing ensures less, but can make better use of resources.
B A F C Functionality The aim of the network is to support the communication needs of applications. E D Reliable unicast channel Unreliable multicast channel with in-order delivery
Dealing with failure • Congestion • Bit or burst errors • Link or node outages
Performance • Bandwidth (throughput) • The number of bits that can be transmitted over the network in a certain period of time. • Latency (delay) • How long it takes a single bit to propagate from one end of the network to the other. • Round Trip Time (RTT): how long it takes for a bit to get from one end of the network to the other and back.
Key equations (Perceived) Latency = Propagation + Transmit + Queue Propagation = Distance / SpeedOfLight Transmit = Size / Bandwidth
Some Units • Mbps = 10**6 bits/sec • byte = 8 bits • KB = 2**10 bytes (= 8,192 bits) • MB = 2**20 bytes (= 8,388,608 bits) • ms = 10**-3 seconds • s = 10**-6 seconds
Bandwidth vs. Latency • Which is the better deal: • Improve your bandwidth from 1 Mbps to 100 Mbps, or • Improve your RTT from 100 ms to 1 ms? • The answer depends on what you need to send.
Transmit Time 1 Mbps 8 s 100 Mbps .08 s Perceived Latency100 ms 1 ms 1 Mbps 100.008 ms 1.008 ms 100 Mbps 100.00008 ms 1.00008 ms 99% 99% .008% .8% Latency Bound • Send a 1 byte message
Perceived Latency 100 ms 1 ms 1 Mbps 210.1 sec 210.001 sec 100 Mbps 21.1 sec 21.001 sec .05% .5% 90% 90% Transmit Time 1 Mbps 3.5 min 100 Mbps 21 sec Bandwidth Bound • Send a 25 MB message
Other measures • Bit width (seconds): 1 bit / bandwidth. • Delay x Bandwidth (bits). • Instructions per mile: number of instructions a machine can execute in the time it takes to send a bit for a mile.
Cycles per mile example • RTT from Penn to Stanford: 120ms • Best in principle is 48ms • On 400 MHz workstation, 48 million cycles elapse in that time. • 6000 mile round trip means 8000 cycles per mile.
Architecture • Many requirements introduce complexity. • Complexity can be controlled by abstraction. • Fundamental idea is to create a separation of concerns, so that each module can focus primarily on its own objectives.
Protocol “Stack” idea • Realization of layered model • Optimizations designed to improve performance • e.g., reducing copying • e.g., removing functional redundancy • Applications at top, cable at bottom
Applications on hosts Application Application Chicago Application Application Madison Saul Application Application Eniac
Application Programs Process-to-Process Channels Host-to-Host Connectivity Hardware Request / Reply Channel Message Stream Channel Basic stack
Protocol Stacks illustrated: • Service and peer interfaces High-level Object High-level Object Service interface Service Interface Protocol Protocol Peer-to-peer interface Host #2 Host #1
ETH Src ETH Dst Example: TCP/IP over Ethernet IP Header TCP Packet (Sequence #, Checksum & Data)
Encapsulation Challenges • Copying Data between layers • Size differences • ATM: 53 bytes, 48 payload • Ethernet: 1536 bytes, 1500 payload • IP: variable to 65,536 • Solved with fragmentation and reassembly
Internet protocol graph FTP HTTP NV TFTP UDF TCP IP Ethernet ATM FDDI