TCP

• Transmission Control Protocol:
  • Sequence and Acknowledgement Numbers
  • Reliable Transmission
  • Windows
  • TCP header
  • TCP connection establishment
  • TCP connection shutdown
  • TCP congestion control
  • TCP checksum and pseudo-header
• UDP

Transmission Control Protocol

• connection oriented
• reliable
• full duplex
• streaming
• point-to-point
• communication over IP

• reliable startup
• reliable shutdown (all data received by each receiver)
• end-to-end (not on routers)

Retransmission

• each packet sent carries a sequence number
• sequence numbers vary independently in the two directions
• each packet received causes the receiver to send an acknowledgement (ack) packet
• the sender starts a retransmission timer whenever it sends a data packet (not an ack)
• if the packet or ack was lost, the timer expires and the sender retransmits the data packet
• the receiver uses the sequence number to detect duplicate, out of order, or old packets
• in TCP, (almost) every packet carries an ACK (piggyback acks)

Adaptive Retransmission

• TCP monitors the time from sending to receiving an ack
• the average of this time is a good estimate for the retransmission timer
• also add in a factor (e.g. 4) times the deviation, so we won't retransmit unnecessarily
• if a packet was retransmitted, don't use that packet's "RTT":
  • if the packet or ack was lost, the RTT is to be measured from the retransmit times
  • otherwise the ack could simply be the delayed ack from the original packet
  • more complicated if we retransmitted the packet more than once

Computation of Running Average

• use a factor alpha, e.g. alpha = 1/8 (it is nice if alpha is a power of two)
• if our estimate is E and we get a new sample s, compute new estimate E' = alpha s + (1 - alpha) E
• example for round-trip time, E'_RTT = 1/8 RTT + 7/8 E_RTT
• example for deviation, E'_dev = 1/8 dev + 7/8 E_dev, where dev = |RTT - E_RTT|
• TCP sets its retransmission timer to E_RTT + 4 E_dev
• if E_RTT = 8 ms, E_dev = 4 ms, and our latest RTT is 16 ms, the new values are: E'_RTT = 9 ms, E'_dev = 4.5 ms

TCP windows

• TCP sequence (and ack) numbers and window sizes are all in bytes (allowing us to segment differently during retransmission, if so desired)
• ack number n means we have received all packets up to and including sequence number n-1
• each ack also carries a window w: if a packet has ack n and window w, the sender is allowed to send up to byte number n+w before stopping
• the sequence number carried in a packet is the sequence number of its first byte

TCP acks

• in case of bidirectional data flow, we want to delay acks so they can be piggybacked
• in case of unidirectional data flow, we want to send acks as soon as possible
• TCP usually doesn't know whether the data flow will be uni- or bi-directional, so TCP delays all acks by a small amount
• if additional buffer space becomes available (the application has read some of the data), an ack might be sent just to update the window
• an ack by itself is unreliable and might get lost, so a sender with more data to send must probe the window by sending 1-byte packets until it gets a response

TCP header

    0                   1                   2                   3   
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |          Source Port          |       Destination Port        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                        Sequence Number                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                    Acknowledgment Number                      |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |  Data |           |U|A|P|R|S|F|                               |
   | Offset| Reserved  |R|C|S|S|Y|I|            Window             |
   |       |           |G|K|H|T|N|N|                               |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |           Checksum            |         Urgent Pointer        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                    Options                    |    Padding    |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                             data                              |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
                            TCP Header Format

TCP Header fields

• Source and Destination port: demultiplexing
• Sequence and acknowledgement: reliable delivery
• Data Offset: header size, options
• Window: flow control
• Checksum: correctness
• Urgent Pointer: "special place" in the data stream

TCP Header bits

• SYN: I want to establish a connection
• FIN: I will never again send data on this connection
• RST: kill this connection
• PSH: immediate delivery of this data is probably a good idea
• URG: the urgent pointer is valid
• ACK: the acknowledgement field is valid (set in all but the first SYN packet)

TCP connections

• sequence numbers, windows, etc. must be remembered and applied to incoming packets
• remembering these numbers is a form of state
• since TCP has state, designers decided to have the peers explicitly manage this state (called a connection)
• the peers agree on when to establish ( open) a connection, when to tear it down ( close), and when the connection must be thrown away ( reset)
• the state on each system reflects an understanding about the state on the peer

TCP connection establishment

• when I receive a request to establish a connection, I must check:
  • that I don't already have this socket: one or more of the port numbers or IP numbers must differ from existing connections
  • that an application on my end desires to be connected
  • that I have sufficient resources to handle this connection
• the purpose of the connection establishment phase is to set up consistent connection state on the two peers

TCP State Machine -- Open

                              +---------+ ---------\      active OPEN  
                              |  CLOSED |            \    -----------  
                              +---------+<---------\   \   create TCB  
                                |                   \   \  snd SYN    
                   passive OPEN |     |   CLOSE        \   \           
                   ------------ |     | ----------       \   \         
                    create TCB  |     | delete TCB         \   \       
                                V     |                      \   \     
                              +---------+            CLOSE    |    \   
                              |  LISTEN |          ---------- |     |  
                              +---------+          delete TCB |     |  
                   rcv SYN      |     |     SEND              |     |  
                  -----------   |     |    -------            |     V  
 +---------+      snd SYN,ACK  /       \   snd SYN          +---------+
 |         |<-----------------           ------------------>|         |
 |   SYN   |                    rcv SYN                     |   SYN   |
 |   RCVD  |<-----------------------------------------------|   SENT  |
 |         |                    snd ACK                     |         |
 |         |------------------           -------------------|         |
 +---------+   rcv ACK of SYN  \       /  rcv SYN,ACK       +---------+
   |           --------------   |     |   -----------                  
   |                  x         |     |     snd ACK                    
   |                            V     V                                
   |  CLOSE                   +---------+                              
   | -------                  |  ESTAB  |                              
   | snd FIN                  +---------+                              

TCP 3-way handshake

• from state CLOSED:
  • send SYN, enter state SYN SENT
  • receive SYN and ACK, send ACK, enter state ESTAB, or
  • receive SYN, send ACK, enter state SYN RCVD, then proceed as below
• from state LISTEN:
  • receive SYN, send SYN and ACK, enter state SYN RCVD
  • receive ACK, enter state ESTAB
• retransmissions in case any of these are dropped
• last ack issue

TCP State Machine -- Close

   |  CLOSE                   +---------+                              
   | -------                  |  ESTAB  |                              
   | snd FIN                  +---------+                              
   |                   CLOSE    |     |    rcv FIN                     
   V                  -------   |     |    -------                     
 +---------+          snd FIN  /       \   snd ACK          +---------+
 |  FIN    |<-----------------           ------------------>|  CLOSE  |
 | WAIT-1  |------------------                              |   WAIT  |
 +---------+          rcv FIN  \                            +---------+
   | rcv ACK of FIN   -------   |                            CLOSE  |  
   | --------------   snd ACK   |                           ------- |  
   V        x                   V                           snd FIN V  
 +---------+                  +---------+                   +---------+
 |FINWAIT-2|                  | CLOSING |                   | LAST-ACK|
 +---------+                  +---------+                   +---------+
   |                rcv ACK of FIN |                 rcv ACK of FIN |  
   |  rcv FIN       -------------- |    Timeout=2MSL -------------- |  
   |  -------              x       V    ------------        x       V  
    \ snd ACK                 +---------+delete TCB         +---------+
     ------------------------>|TIME WAIT|------------------>| CLOSED  |
                              +---------+                   +---------+

TCP close, part 1

• from state ESTAB:
  • receive FIN, send ACK, enter state CLOSE WAIT
  • application closes connection, send FIN, enter state LAST ACK
  • receive ACK, enter state CLOSED
• from state ESTAB:
  • application closes connection, send FIN, enter state FINWAIT-1
  • receive FIN, send ACK, enter state CLOSING
  • receive ACK, enter state TIME WAIT

TCP close, part 2

• from state FINWAIT-1, if we get an ACK:
  • receive ACK, enter state FINWAIT-2
  • receive FIN, send ACK, enter state TIME WAIT
• from state TIME WAIT, enter state CLOSED after 4 minutes (2 Maximum Segment Lifetimes)

TCP congestion control

• TCP Reno:
  • if a packet is lost, assume it's because of congestion
  • slow down significantly (cut window by a factor of two)
• TCP Vegas:
  • measure the round-trip time of packets
  • if significantly greater than the minimum, assume due to congestion, slow down slightly (linear)
  • use Reno if packets are lost

Checksum

• In TCP and UDP, the checksum covers the header, data, and a pseudo-header
• Unlike TCP, the UDP checksum is optional, and sent as zero (0x0000) if not computed
• the TCP and UDP pseudo-header has fields derived from the IP header:
• this header is used in computing the checksum, but never sent

    0                   1                   2                   3   
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                        Source address                         |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                     Destination address                       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     zero      |   Protocol    |       TCP/UDP length          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
                          Pseudo Header Format

UDP header format

    0                   1                   2                   3   
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |          Source Port          |       Destination Port        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |            Length             |           Checksum            |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
                            UDP Header Format

UDP Checksum

• As in TCP, the UDP checksum covers the header, data, and a pseudo-header
• Unlike TCP, the UDP checksum is optional, and sent as zero (0x0000) if not computed
• the TCP and UDP pseudo-header has fields derived from the IP header:

    0                   1                   2                   3   
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                        Source address                         |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                     Destination address                       |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |     zero      |   Protocol    |       TCP/UDP length          |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
 
                          Pseudo Header Format