[Ch3] UDP, TCP

UDP: connectionless transport

UDP

UDP: User Datagram Protocol

• connectionless: no call setup -> no delay before data delivery
• doesn't care errors at receiver -> just error-check
  • UDP datagram can be lost or delivered out-of-order to app
• UDP doesn't control over what data is send & when -> App can send its data as fast as it wants
• used in streaming multimedia apps (loss tolerant, rate sensitive), DNS, SNMP
• reliable transfer over UDP: add reliability at application layer, application-specific error recovery

 

UDP checksum

UDP Checksum

• sender: header field를 포함한 segment contents를 sequence of 16-bit integers로 -> sum 계산 -> one's complement sum (receiver에서의 계산을 쉽게 하기 위해) -> sender puts checksum value into UDP checksum field
• receiver compute checksum and check if equals checksum field value (sum === 0 ? YES=no error : NO=error)

 

👉 UDP checksum과 TCP checksum의 알고리즘은 같다.

---

Reliable Data Transfer (RDT)

Reliable Data Transfer Mechanism

• Checksum
  • used to detect bit errors in a transmitted packet.
• Acknowledgement
  • used by receiver to tell the sender that packets has been received correctly.
  • Ack will typically carry the sequence number of the packets being acknowledged.
  • Ack may be individual or cumulative, depending on the protocol
• Negative Acknowledgement
  • used by receiver to tell the sender that packet has not been received correctly.
  • Negative ack will typically carry the sequence number of the packet that was not received correctly.
• Sequence Number
  • used for a sequential numbering of packets of data flowing from sender to receiver
  • Gaps in the sequence numbers of received packets allow the receiver to detect a loss packet.
  • Packets with duplicate sequence numbers allow the receiver to detect duplicate copies of a packet.
• Timer
  • used to timeout/retransmit a packet, possibly because the packet or its ACK was lost within the channel.
  • premature timeout: when a packet was delayed but not lost
    • if too short timeout value => unnecessary retransmit: 네트워크 자원 낭비
  • slow reaction: when a packet was recieved by the receiver but the receiver-to-sender ACK was lost, duplicate copies of a packet may be received by a receiver
    • if too long timeout value => duplicate or long delay: 패킷 loss 감지 및 필수적인 재전송이 느려짐 = end-to-end latency가 나빠진다.
• Pipelining & Window
  • sender may be restricted to sending only packets with sequence numbers that fall within a given range.
  • by allowing multiple packets to be transmitted but not yet acknowledged, sender utilization can be increased over a stop-and-wait mode of ooperation.
    • stop & wait: 하나씩 보내고 Ack을 기다리고 다시 보냄
  • window size W may be set on the basis of the receiver's ability to receive and buffer messages, or the level of congestion in the network
    • pipelining: inflight bytes (unacked) -> maximum # of inflight bytes = window size W
  • W = min(congestion window, receiver window)

---

TCP

TCP overview

• point-to-point: one sender - one receiver
• reliable, in-order byte stream (sequence number per byte)
• connection-oriented: handshaking before data exchange
• full duplex data: bi-directional data flow in same connection
  • MSS: maximum segment size
• flow controlled: sender will not overwhelm receiver -> receiver가 알려줌 (rwnd)
• congestion controlled: sender will not overwhelm network -> sender가 유추 (cwnd)
• pipelined: TCP congestion and flow control set window size

- rwnd: TCP receiver의 receive buffer 잔여량을 TCP receiver가 TCP header에 넣어 TCP sender에게 알려줌으로써 sending rate 조절하게 하는 flow control을 위한 파라미터

- cwnd: TCP sender의 local parameter로 network 상황을 간접적으로 예측하여 조정하는 값

 

TCP Segment Structure

• seuence #
  • 초기값 랜덤, 항상 유효
  • byte stream number of first byte in segment's data
• ack #
  • ACK bit === 1: 유효
  • sequence number of next byte that a receiver expect to receive for in-order delivery
  • cumulative ACK

- receiver 는 TCP segment header의 ACK flag bit가 0인 경우에도 Ack number 필드를 처리한다? (X)

- TCP가 connection을 맺을 때마다 sequence number는 항상 초기값 0으로 시작한다? (X)

- receiver가 받은 TCP segment header의 sequence number 필드 값은 body에 실려있는 데이터의 첫번째 바이트에 붙여진 sequence number이다.

 

 

 

TCP Round Trip Time (RTT)

TCP Timeout Value

• longer than RTT -> 실제 네트워크의 RTT값 이용해서 계산
  • if too short -> premature timeout, unnecessary retransmissions
  • if too long -> slow reaction to segment loss
  • 👉 Goal: reducing unneeded retransmission & avoiding unnecessary delay
• How to estimate RTT?
  • Sample RTT: measured time from segment transmission until ACK receipt
    • Ignore retransmission

EstimatedRTT = (1 - a) * EstimatedRTT + a * SampleRTT

• influence of past sample decreases exponentially fast
• typical value: a = 0.125
• 오래된 값이 weight를 적게 받는다

TCP sender가 timeout 값을 조정하지 않고 network router의 buffer 크기를 증가시키면?

• TCP sender가 delayed pkt을 loss로 오인 -> 불필요한 retransmission 증가
• RTT 증가 -> throughput 감소
• 중간 라우터의 buffer에 TCP receiver가 이미 받은 pkt들이 차지하는 비율 증가

 

 

Timeout Interval = EstimatedRTT + 4 * DevRTT

• 4 * DevRTT = safety margin

 

TCP Reliable Data Transfer (RDT)

• pipelined segments, cumulative acks => 3 duplicate acks (fast retransmission before timeout)
• single retransmission timer per connection -> timer for oldest unacked segment == SendBase

3 Duplicate Acks

• if sender receives 3 duplicate Acks for same data, then t he sender resend unacked segment with smallest sequence # without wating for timeout
• 네트워크에서 drop 되지는 않았지만 늦게 도착한 패킷 (delayed not lost pkt)을 loss로 오판하여 불필요한 retx을 보내는 것을 막기 위해

👉 fast retransmit

👉 TCP는 timeout 값은 넉넉히 설정하고 pipelining과 cumulative Ack 을 이용해서 timeout 전에 loss를 유추하는 방법 사용

 

TCP Flow Control

Flow Control

• receiver controls sender (sending rate), so sender won't overflow receiver's buffer by transmitting too much, too fast.
• rwnd value in TCP header of receiver-to-sender segments
  • rwnd == free buffer space
• sender limits amount of unacked (in-flight) data to receiver's rwnd value

👉 guarantees receive buffer will not overflow

 

Connection Management

TCP 3-way handshake

 

TCP: Closing a connection

• client, sender each close their side of connection
• send TCP segment with FIN bit = 1
• respond to received FIN with ACK
  • on receiving FIN, ACK can be combined with own FIN (FINACK)