1. Transmission Control Protocol (TCP) BSAD 146 Dave Novak Sources: Network+ Guide to Networks, Dean 2013

2. Last time… Routing Static and Dynamic routing Routing Protocols EGPs BGP IGPs Focused on OSPF CIDR Do you understand the notation?

3. Outline Discuss role of layer 4 transport protocols from perspective of OSI model services Process-level communication TCP How different from IP Connection-based service Reliability aspects UDP

4. Transport protocols From OSI model perspective IP is a network layer protocol IP has its own layer (internetworking layer) in the TCP/IP model IP provides a universal routable address packet format Connectionless, best-effort service (unreliable service)

5. Transport protocols Layer 4 transport layer protocols work closely with layer 3 network layer protocols to provide additional services: You typically see protocols stacks (from various vendors) referred to by the primary layer 3 and 4 protocols in the stack Process-to-process communication End-to-end error control

6. Transport protocols IP focuses on host-to-host communication, not process-to-process communication What is process-to-process communication? Individual hosts run multiple processes at one time Multiple applications, utilities, etc. Must be some protocol to manage multiple processes between hosts

7. Transport protocols Layer 4 transport layer protocols are responsible for more refined or detailed management Process-to-process communication 2 transport layer protocols in TCP/IP protocol suite 1) 2)

8. Internet Protocol (IP) IP is connectionless Provides a generic, universal address format that is understood by all devices on the internet Independent of specific data link technologies Routes all IP datagrams independently No guarantee that packets will travel on same route to destination

9. Internet Protocol (IP) Allows for fragmentation and reassembly of datagrams Can subdivide datagrams to fit in any frame payload

10. Transmission Control Protocol (TCP) Connection-oriented service Establishes virtual connection between 2 hosts Provides reliable service Guaranteed delivery of data with acknowledgement from the receiver

11. Transmission Control Protocol (TCP) Point-to-point Establishes connections between two specific endpoints – single process running on sending host to single process running on receiving host

12. Transmission Control Protocol (TCP) Manage process-to-process communication Address out of order delivery Address packet loss Adjust routing based on network conditions (congestion control) Adjust how many packets are sent in a given time period (flow control) Establish an explicit “connection” between sender and receiver Address corrupt data

13. TCP Segments TCP uses same format for all messages (the segment “looks” the same regardless of the type of message) Data transmission ACK (acknowledgement) 3-way handshake (connection establishment and tear-down) All TCP messages called segments Segments encapsulated in IP datagram payload

14. TCP Encapsulation Layer 4 (TCP segment) Layer 3 (IP datagram) Layer 2 (Frame) IP datagram payload – TCP segment header becomes part of IP payload Frame payload – IP header placed in frame payload IP datagram header Frame header

15. TCP – Ports Transport layer protocol (TCP or UDP) identify the protocol or process that generated the data that will be sent over the network Often map this to a specific port number 1) On a computer, a place to be physically connected to another device (example, serial port, parallel port) 2) Using TCP/IP, a logical connection (a client application specifies an application on a remote computer (SOFTWARE) 3) In programming, to move an application for the development environment to be run in another environment

16. TCP – Ports Common internet applications are assigned specific, well-known port numbers TCP and UDP maintain their own separate lists of well-known ports

17. TCP – Examples of Common Ports You can visit http://en.wikipedia.org/wiki/List_of_TCP_and_UDP_port_numbers http://en.wikipedia.org/wiki/List_of_TCP_and_UDP_port_numbers To view a list of well-known TCP and UDP port numbers

18. Sockets TCP/IP system addresses traffic to another system using combination of: 1) IP address 2) Port number This combination of IP address and port number is called a socket Sending socket – 137.99.101.55:4532 Receiving socket – 185.100.32.56:80

19. TCP – Sockets Don’t specify port number when typing URL because applications assume you are connecting to well known port numbers Your Web browser generally addresses all URLs to TCP port 80 – default HTTP port You could configure a Web server to operate on an alternate port – any users connecting to your Web server would need to know the port number and specify it http://www.privateserver.com:81

20. TCP Mapping the OSI and TCP/IP models Physical layer Transport layer Network layer Data Link layer Session layer Presentation layer Application layer 7 5 4 2 3 1 6 Network Interface layer Internet layer Transport layer Application layer FTP, HTTP, DHCP Binary, ASCII TCP IP LAN technology, cabling

21. TCP Services Major features of TCP Connection oriented Point-to-point communication Complete reliability Full duplex communication Connection startup and shutdown

22. Reliable service – what does this mean? Many types of problems can arise on a communications network Duplicate messages Out-of-order delivery Delayed messages Lost messages TCP employs a number of techniques to address these issues and provide “reliable” service

23. Achieving Reliability 4 components to TCP reliable service 1) 2) 3) 4)

24. 1) Connection Establishment Before two systems exchange application layer data, a virtual connection between 2 processes on 2 hosts explicitly established Ensures both sender and receiver are: 1) Present (actually exist and are accessible) 2) Operating properly 3) Ready to receive data

25. 1) Connection Establishment Connection implemented through software and remains alive during data exchange Then the connection terminated in an orderly, systematic fashion Usually a single file exchange

26. 1 ) Connection Establishment Three-way handshake algorithm Systematic connection establishment and tear down requires an exchange of three messages SYN segment – synchronization segment for establishment FIN segment – for termination

27. 1 ) Connection Establishment Three-way handshake algorithm Each system (sender on one end, receiver on the other) establishes separate connections running in opposite directions Full duplex protocol Each terminates separately

28. 3-Way Handshake (initialization) Client initiatesServer 1) SYN 2) ACK + SYN 3) ACK 1) Client sends session Initialization request to server. Includes “plan” for transmission 2) Server sends back its starting sequence number and an acknowledgement 3) Client acknowledges receipt of server’s sequence number

29. Terminating Both Connections (tear down) Client initiatesServer 1) Client FIN 2) ACK 3) Server FIN 1) Client sends termination request to server 2) Server acknowledges that it has received client’s request 3) Server generates its own FIN message 4) Client acknowledges receipt of server’s FIN 4) ACK FIN and ACK cannot piggy back Server cannot include its FIN in the ACK sent to client

30. End-to-End Service Direct connection from process on one host to process on another host Set up connection Socket Exchange data Receiver knows exactly how much data to expect Close connection

31. 2) Retransmission If data are lost, or there are problems, TCP can retransmit messages TCP allows multiple applications to communicate with multiple destinations at same time TCP must handle variety of delay scenarios and times that can change A timer is set to decide when to retransmit The retransmission timer can differ by application

32. 2) Retransmission

33. Timer for retransmission depends on: Network congestion Distance to destination Bandwidth of link TCP can adapt to network conditions and to the needs of individual applications

34. Adaptive Retransmission TCP uses adaptive retransmission Monitors delay on each retransmission and adapts to accommodate changing conditions TCP estimates round trip time (RTT) for each active connection

35. Adaptive Retransmission Adapts by subtracting time message is sent from time ack is received Ack received 8:03:45:07 Message sent 8:03:45:02 8:03:45:07 - 8:03:45:02 = 0:0:0:05 Uses weighted moving average of RTT Weights most recent activity most heavily Adjusts to conditions on network

36. Adaptive Retransmission Fixed timeouts specifying retransmission times do not work well on a large internetwork Why would this be the case?

37. Adaptive Retransmission

38. 3) Flow Control Managing the flow of data between two processes Why would this be important? What would happen if the sending computer is much faster and has a larger buffer than the receiving computer?

39. 3) Flow Control TCP utilizes the sliding window scheme for flow control Connection established Each host allocates buffer to hold incoming data until it can be processed Size of buffer sent to other computer Remaining buffer size specified in ack when each transmission is received Sending computer adjusts amount of data it sends based on size of remaining buffer size If zero buffer remains, sending computer must stop transmitting until space is available

40. 3) Flow Control

41. TCP window – amount of unacknowledged data the sender can send to a particular connection without receiving acknowledgement from the receiver When the window capacity is achieved, the sender MUST wait for acknowledgement before sending more data

42. 3 ) Flow Control Offered window = TCP window # TCP segments receiving system is authorized to send before it must wait for ACK As receiver acknowledges bytes, left side of window moves to the right As receiver passes acknowledged bits up to application layer, right side of window moves to the right The offered window never can exceed 6 TCP segments

43. 4) Congestion Control Managing amount of data being injected onto the network by specific host Why would this be important? What happens if switches and/or links become overloaded?

44. 4 ) Congestion Control Packet loss and delay most often caused by congestion, not hardware failure TCP can exacerbate congestion problem through retransmission How (what happens if you add cars to an already congested freeway)?

45. 4) Congestion Control Congestion collapse TCP uses packet loss as a measure of congestion measurement TCP determines how busy the network is by the number and frequency of lost packets

46. 4 ) Congestion Control Simplified algorithm for congestion control When packet loss occurs, TCP immediately reduces number of packets sent to one (regardless of the maximum number it can send) If ack for single packet received, send 2 packets Receiver acknowledges successful receipt of data If ack for both packets received, send 4 packets Exponential increase, until half Send Window Size (max amount that can be sent) achieved Than cut back to a less aggressive increase

47. User Datagram Protocol (UDP) While TCP captures most of our attention, UDP is also widely used transport protocol in TCP/IP stack Process-to-process communication Connectionless datagram service Provides NO reliability Very little overhead

48. User Datagram Protocol (UDP) Communication consists of 2 messages Request and reply – there is no connection establishment

49. UDP If TCP provides reliability and UDP does not, why would UDP be used at all?

50. Summary Process-to-process communication TCP How different from IP Four reliability components UDP