1. Sharing the Network It slices, it dices, it sequences ….. All of this and error checking too!

2. The World Before the Packet  All Connections needed to be point to point –Terminal to Mainframe. –Point-to-Point: typically a leased line.  Message Switched systems –Messages routed one hop at a time –Messages are not divided (by the network) –Forwarding stations need enough memory to hold message. –Prone to single point of failure.

3. How to Share a Wire?  Frequency Slicing –Like Radio Stations –Used in Coaxial Ethernet (freq for Tx and Rx)  Time Slicing –Each station gets a portion of time  Spread Spectrum –Transmission is coded, decoded at Receiver –Bandwidth is larger than needed –Like being at a party. Lots of people talk at once, but you can make out individual conversations.

4. Packet Switching  Theory developed in the 60s –Store and forward (still) –Messages are divided into Packets –Routing decisions are made by packet  Redundant Links allow fault-tolerance  All packets in a message need not travel the same path

5. OSI Transport Layer  Sequencing: Assure proper packet order  Network Failure Recovery: reestablish a broken connection  Flow Control  Error Recovery  Connections: Establish and Termination  Higher Protocol Addressing

6. TCP in the Stack of Things

7. TCP  TCP provides a “shim” between applications and the Network Layer. Applications issue a single call to TCP. It takes care of the rest.  TCP is connection-oriented –Provided reliable end-to-end communication –Accuracy over timeliness.  TCP Packets have large overhead –32-bits: Source/Destination ports –64-bits Seq/Ack numbers –16-bits Flags –48-bits Window size/Checksum/Urgent Pointer –That’s 160 bits (minimum) of overhead

8. Establishing a TCP Connection  Three Way Handshake  SYN: Client->Server (on well known port) –Random sequence number –Client Port number  SYN->ACK: Server-Client –Ack=Seq+1, –Server sets aside buffer space  ACK: Client to Server –Seq=received ACK,  Connection (Socket) created

9. Reliability  Sequence numbers id each Byte –Allows for fragmentation, errors, loss  ACK set to next expected Byte

10. Error Detection  Sequence/Acknowledgement cover loss/duplication  Checksum verifies (weakly) data

11. Flow Control  Sliding Window –Receiver indicates size of receive buffer –Sender does not exceed size until Ack. –ACK can change size of receive buffer  Congestion Control –Acks (or lack thereof) infer congestion –Retransmission Timers

12. TCP Uses  Used where accuracy is important –My paycheck =1.5*10 3 (3 is 0011 in binary) –A 1-bit error could change my paycheck from $1500 to $1.50 or (better yet) $150,000,000,000 –A 1-bit error not import in say…. VOIP  Used where speed is not important

13. User Datagram Protocol  UDP: Best Effort (like Post Office) –No handshakes, reliability, ordering, data integrity –Very useful for small Query/Replies –DNS, Streaming, VOIP, TFTP,RIP,  Header: –16-bits Source Port –16-bits Destination Port –16-bits Length –16-bits Checksum

14. Services  Services are designated by “Ports” –Its just a number –0-1023: well-known (assigned by IANA) –1024-49151: registered (loosely) –49152-65535: dynamic/private ports used by clients when connecting to servers

15. Port 53: Domain Name System  DNS –Converts IP name to Number –Typically used by other services –Can be invoked by command nslookup –UDP used for host lookups –TCP used for server-server updates (reliable)

16. Port 80 HTTP  System of Interlinked (web) hypertext Docs  Created by Tim Berners-Lee for CERN in 1989 to share scientific documents  Before HTTP users used applications with arcane commands (I miss those days)  Became popular in 1993 with release of the Mosaic Web Browser

17. Web Traffic