程序代写代做代考 dns Computer Systems Transport Layer

Computer Systems Transport Layer
Dr. Mian M. Hamayun
m.m.hamayun@bham.ac.uk
Based on material and slides from
Computer Networking: A Top Down
Approach, 7th Edition – Chapter 3 Jim Kurose, Keith Ross Pearson/Addison Wesley

Lecture Objective
The objective of this lecture is to understand the conceptual and implementation aspects of transport layer protocols
Slide# 2 of 63

Lecture Outline
 Transport Layer Services & Protocols  Multiplexing / Demultiplexing
 User Datagram Protocol (UDP)
 Principles of Reliable Data Transfer
 Transmission Control Protocol (TCP)  Flow Control
 Connection Management
 Summary
Slide# 3 of 63

Recap – Network Layers
Slide# 4 of 63

Transport Layer Services
 Provide logical communication between application processes running on different hosts.
 Transport protocols run in end systems (not routers)
 send side: breaks app messages into segments, passes to network layer
 receive side: reassembles segments into messages, passes to application layer
 More than one transport protocol available to apps
 Internet: TCP and UDP
Slide# 5 of 63

Transport vs. Network Layer
 Network layer: logical communication between hosts.
 Transport layer: logical communication between processes
relies on, enhances, network layer services
Household Analogy:
12 kids in Ann’s house sending letters to 12 kids in Bill’s house:
 hosts = houses
 processes = kids
 app messages = letters in envelopes
 transport protocol = Ann and Bill who demux to in-house siblings
 network-layer protocol = postal service
Slide# 6 of 63

Internet Transport Layer Protocols
 Reliable, in-order delivery (TCP)
congestion control flow control connection setup
 Unreliable, unordered delivery: UDP
no-frills extension of “best- effort” IP
 Services not available: delay guarantees bandwidth guarantees
Slide# 7 of 63

Multiplexing / Demultiplexing
multiplexing at sender:
handle data from multiple sockets, add transport header (later used for demultiplexing)
demultiplexing at receiver:
use header info to deliver received segments to correct socket
Slide# 8 of 63

How Demultiplexing Works?
 Host receives IP datagrams  each datagram has source IP
address, destination IP address
 each datagram carries one transport-layer segment
 each segment has source, destination port number
 port numbers 0 – 65535
 well-known ports 0 – 1023
 Host uses IP addresses & port numbers to direct segment to appropriate socket
TCP/UDP segment format
Slide# 9 of 63

Connectionless Demultiplexing
 When a UDP Socket is created, a host-local port# is assigned to it.
 When creating a datagram to send into a UDP socket, we must specify
 Destination IP Addr  Destination Port #
 When a host receives UDP segment:
checks destination port# in segment
directs UDP segment to socket with that port #
 IP datagrams with same dest. port #, but different source IP addresses and/ or source port numbers will be directed to same socket at destination
Slide# 10 of 63

Connectionless Demultiplexing – Example
Slide# 11 of 63

Connection-oriented Demultiplexing
 TCP socket identified by 4-tuple:
 source IP address  source port number  dest IP address
 dest port number
 Demux: receiver uses all four values to direct segment to appropriate socket
 Server host may support many simultaneous TCP sockets:
 each socket identified by its own 4-tuple
 Web servers have different sockets for each connecting client
 non-persistent HTTP will have different socket for each request
Slide# 12 of 63

Connection-oriented Demultiplexing – Example
Slide# 13 of 63

UDP: User Datagram Protocol [RFC 768]
 “No Frills”, “Bare Bones” Internet transport protocol
 “Best Effort” service, UDP segments may be:
 Lost
 Delivered out-of-order
 Connectionless:
 no handshaking between UDP
sender, receiver
 each UDP segment handled independently of others
 UDP used in:
 streaming multimedia apps (loss
tolerant, rate sensitive)
 DNS (Domain Name System)
 SNMP (Simple Network Management Protocol)
 Reliable transfer over UDP:
 add reliability at application layer e.g. QUIC (Quick UDP Internet Connections) protocol
 application-specific error recovery!
Slide# 14 of 63

UDP: Popular Internet Applications
Slide# 15 of 63

UDP: Segment Header
UDP segment format
length, in bytes of UDP segment, including header
Why is there a UDP?
 no connection establishment (which can add delay)
 simple: no connection state at sender, receiver
 small header size (8 bytes)
 no congestion control: UDP can blast away as fast as desired
Slide# 16 of 63

UDP Checksum
Goal: detect “errors” (e.g., flipped bits) in transmitted segment
Sender:
 Treat segment contents, including header fields, as sequence of 16-bit integers
 Checksum: addition (one’s complement sum) of segment contents
 Sender puts checksum value into UDP checksum field
Receiver:
 Compute checksum of received segment
 Check if computed checksum equals checksum field value:
 NO – error detected
 YES – no error detected.
 But maybe errors nonetheless? More later …
Slide# 17 of 63

UDP Checksum – Example
wraparound
sum checksum
11110011001100110 11101010101010101
11011101110111011 11011101110111100
10100010001000011
Note: when adding numbers, a carryout from the most significant bit needs to be added to the result
Slide# 18 of 63

Principles of Reliable Data Transfer
Slide# 19 of 63

Reliable Data Transfer over a Perfectly Reliable Channel: rdt1.0
Underlying channel perfectly reliable  no bit errors
 no loss of packets
 Separate FSMs for sender, receiver:  sender sends data into underlying channel  receiver reads data from underlying channel
Slide# 20 of 63

RDT2.0: Channel with Bit Errors
Underlying channel may flip bits in packet checksum to detect bit errors
The question: how to recover from errors:
Acknowledgements (ACKs): receiver explicitly tells sender that packet
received OK
Negative Acknowledgements (NAKs): receiver explicitly tells sender that packet had errors
sender retransmits packet on receipt of NAK New mechanisms in rdt2.0 (beyond rdt1.0):
error detection
receiver feedback: control messages (ACK,NAK) receiver->sender
Slide# 21 of 63

RDT2.0: FSM Specifications
Slide# 22 of 63

RDT2.0: Operation with No Errors
Slide# 23 of 63

RDT2.0: Operation with Errors
Slide# 24 of 63

RDT2.0 has a fatal flaw!
What happens if ACK / NAK corrupted?
 Sender doesn’t know what happened at receiver!
 Can’t just retransmit: possible duplicate
Handling Duplicates:  Sender retransmits current
packet if ACK/NAK corrupted  Sender adds sequence
number to each packet
 Receiver discards (doesn’t
deliver up) duplicate packet
Stop and Wait
sender sends one packet, then waits for receiver response
Slide# 25 of 63

RDT2.1: Sender, handles garbled ACK/NAKS
Slide# 26 of 63

RDT2.1: Receiver, handles garbled ACK/NAKS
Slide# 27 of 63

RDT2.1: Discussion
Sender:
 Seq # added to pkt  Two seq. #’s (0,1) will
suffice. Why?
 Must check if received ACK/NAK corrupted
 Twice as many states
 state must “remember” whether “expected” packet should have seq # of 0 or 1
Receiver:
 Must check if received packet is duplicate
state indicates whether 0 or 1 is expected packet seq #
 Note: receiver can not know if its last ACK/ NAK received OK at sender
Slide# 28 of 63

RDT2.2: A NAK-free Protocol
 Same functionality as rdt2.1, using ACKs only  Instead of NAK, receiver sends ACK for last pkt
received OK
 receiver must explicitly include seq # of packet being ACKed
 Duplicate ACK at sender results in same action as NAK: retransmit current packet
Slide# 29 of 63

RDT2.2: Sender, Receiver FSM Fragments
Slide# 30 of 63

RDT3.0: Channels with Errors and Loss
New Assumption:
underlying channel can also lose packets (data, ACKs)
 Checksum, seq. #, ACKs, retransmissions will be of help … but not enough
Approach: sender waits “reasonable” amount of time for ACK
 Retransmits if no ACK received in this time
 If packet (or ACK) just delayed (not lost):
retransmission will be duplicate, but seq. #’s already handles this
receiver must specify seq # of packet being ACKed
Requires countdown timer
Slide# 31 of 63

RDT3.0: Channels with Errors and Loss
Slide# 32 of 63

RDT3.0 in Action!
Slide# 33 of 63

RDT3.0 in Action!
Slide# 34 of 63

Performance of RDT 3.0
 rdt3.0 is correct, but performance stinks
 e.g. 1 Gbps link, 15ms propagation delay, 8000 bit packet:
 Usender: utilization – fraction of time sender busy sending  If RTT = 30ms, 1KB packet every 30ms: 33kB/sec
throughput over 1 Gbps link!
 Network protocols limits use of physical resources!
Slide# 35 of 63

Pipelined Protocols
Pipelining: sender allows multiple, “in-flight”, yet-to-be- acknowledged packets
range of sequence numbers must be increased buffering at sender and/or receiver
Two generic forms of pipelined protocols:
 Go-Back-N (GBN) and Selective Repeat (SR)
Slide# 36 of 63

TCP Overview
[RFCs 793, 1122, 1323, 2018, 2581]
 Point-to-Point:
one sender, one receiver
 Reliable, in-order byte steam:
no “message boundaries”
 Pipelined:
TCP congestion and flow control set window size
Full duplex data:
bi-directional data flow in same
connection
MSS: maximum segment size (app data size, usually 1460 bytes)
Connection-oriented:
handshaking (exchange of control messages) initializes sender, receiver state before data exchange
Flow controlled:
sender will not overwhelm the receiver
Slide# 37 of 63

TCP: Segment Structure
URG: urgent data (generally not used)
ACK: ACK # valid
PSH: push data now (generally not used)
RST, SYN, FIN: connection estab (setup, teardown commands)
Internet checksum (as in UDP)
counting
by bytes
of data
(not segments!)
# bytes receiver willing to accept
Slide# 38 of 63

TCP Sequence Numbers & ACKs
 Implicit numbering for each byte of data  File Size = 500,000 bytes
 MSS = 1000 bytes
 Total Segments = 500
 First Segment Seq. # = 0
 Second Segment Seq. # 1000
 Third Segment Seq. # 2000 & so on.
Slide# 39 of 63

TCP Sequence Numbers & ACKs
Sequence Numbers:
byte stream “number” of first byte in segment’s data
Acknowledgments:
seq # of next byte expected from other side
cumulative ACK
Q: How receiver handles out-
of-order segments?
A: TCP spec doesn’t say, up to the implementer (usually kept)
Slide# 40 of 63

TCP Sequence Numbers & ACKs
Slide# 41 of 63

TCP Round Trip Time (RTT), Timeout
Q: How to set TCP timeout value?
 longer than RTT  but RTT varies
 too short: premature timeout, unnecessary retransmissions
 too long: slow reaction to segment loss
Q: How to estimate RTT?
SampleRTT: measured time from segment transmission until ACK receipt
ignore retransmissions
 SampleRTT will vary, want
estimated RTT “smoother”
average several recent measurements, not just current SampleRTT
Slide# 42 of 63

TCP Round Trip Time (RTT), Timeout
EstimatedRTT = (1-)*EstimatedRTT + *SampleRTT
 Exponential Weighted Moving Average (EWMA)
 Influence of past sample decreases exponentially fast
 typical value:  = 0.125
Slide# 43 of 63

TCP Round Trip Time (RTT), Timeout
 Timeout Interval: EstimatedRTT plus “safety margin”  large variation in EstimatedRTT -> larger safety margin
 Estimate SampleRTT deviation from EstimatedRTT:
DevRTT = (1-)*DevRTT + *|SampleRTT-EstimatedRTT|
(typically,  = 0.25) TimeoutInterval = EstimatedRTT + 4*DevRTT
estimated RTT
“safety margin”
Slide# 44 of 63

TCP Reliable Data Transfer
 TCP creates reliable data transfer service on top of IP’s unreliable best-effort service
pipelined segments cumulative acks
single retransmission timer
 Retransmissions are triggered by:
timeout events duplicate acks
Let’s initially consider simplified TCP sender:
 ignore duplicate acks
 ignore flow control, congestion control
Slide# 45 of 63

TCP Sender Events
Data Received from App:
 Create segment with seq #
 Seq # is byte-stream number of first data byte in segment
 Start timer if not already running
think of timer as for oldest unacked segment
expiration interval: TimeOutInterval
Timeout:
 Retransmit segment that
caused timeout  Restart Timer
Ack Received:
 If ack acknowledges previously unacked segments
update what is known to be ACKed
start timer if there are still unacked segments
Slide# 46 of 63

TCP Sender (Simplified)
Slide# 47 of 63

TCP Retransmission Scenarios
Lost ACK Scenario Time-out Scenario; Seg#100 not retransmitted
Slide# 48 of 63

TCP Retransmission Scenarios
Cumulative ACK
Slide# 49 of 63

TCP ACK Generation
[RFC 1122, RFC 2581]
Event at Receiver
Arrival of in-order segment with expected seq #. All data up to expected seq # already ACKed
Arrival of in-order segment with expected seq #. One other segment has ACK pending
Arrival of out-of-order segment higher-than-expect seq. # . Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver Action
Delayed ACK. Wait up to 500ms
for next segment. If no next segment, send ACK
Immediately send single cumulative ACK, ACKing both in-order segments
Immediately send duplicate ACK, indicating seq. # of next expected byte
Immediate send ACK, provided that segment starts at lower end of gap
Slide# 50 of 63

TCP Fast Retransmit
 Time-out period often relatively long:
Long delay before resending lost packet
 Detect lost segments via duplicate ACKs.
Sender often sends many segments back-to-back
If segment is lost, there will likely be many duplicate ACKs.
TCP Fast Retransmit
If sender receives 3 ACKs for same data
(“triple duplicate ACKs”), resend unacked segment with smallest sequence #.
Likely that unacked segment lost, so don’t wait for timeout.
Slide# 51 of 63

TCP Fast Retransmit
Slide# 52 of 63

TCP Flow Control
[Speed-matching b/w Sender & Receiver]
Flow Control
Receiver controls sender, so sender won’t overflow receiver’s buffer by transmitting too much data, too fast.
Slide# 53 of 63

TCP Flow Control
[Speed-matching b/w Sender & Receiver]
Receiver “advertises” free buffer space by including rwnd value in TCP header of receiver-to-sender segments
 RcvBuffer size set via socket options (typical default is 4096 bytes)
 many operating systems autoadjust RcvBuffer
Receiver-side Buffering
Receiver:
rwnd = RcvBuffer –
[LastByteRcvd – LastByteRead]
Sender:
LastByteSent – LastByteAcked <= rwnd Sender limits amount of unacked (“in-flight”) data to receiver’s rwnd value  Guarantees receive buffer will not overflow Slide# 54 of 63 TCP Connection Management Before exchanging data, sender/receiver “handshake”:  agree to establish connection (each knowing the other willing to establish connection)  agree on connection parameters Slide# 55 of 63 TCP 3-Way Handshake Client State LISTEN choose init seq num, x server state LISTEN SYN RCVD SYNSENT SYNbit=1, Seq=x SYNbit=1, Seq=y ACKbit=1; ACKnum=x+1 ACKbit=1, ACKnum=y+1 choose init seq num, y send TCP SYNACK msg, acking SYN send TCP SYN msg ESTAB received SYNACK(x) indicates server is live; send ACK for SYNACK; this segment may contain client-to-server data received ACK(y) indicates client is live ESTAB Slide# 56 of 63 TCP 3-Way Handshake closed Socket connectionSocket = welcomeSocket.accept(); SYN(x) SYNACK(seq=y,ACKnum=x+1) create new socket for communication back to client listen Socket clientSocket = newSocket("hostname","port number"); SYN(seq=x) SYN sent SYNACK(seq=y,ACKnum=x+1) ACK(ACKnum=y+1) SYN rcvd ACK(ACKnum=y+1) ESTAB Slide# 57 of 63 Closing a TCP Connection  Client, server 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  Simultaneous FIN exchanges can be handled Slide# 58 of 63 Closing a TCP Connection Client state ESTAB FIN_WAIT_1 FIN_WAIT_2 TIMED_WAIT segment lifetime CLOSED Server state ESTAB CLOSE_WAIT send data LAST_ACK CLOSED clientSocket.close() can no longer send but can receive data FINbit=1, seq=x ACKbit=1; ACKnum=x+1 close wait for server can still FINbit=1, seq=y ACKbit=1; ACKnum=y+1 can no longer send data timed wait for 2*max Slide# 59 of 63 TCP Client & Server States Slide# 60 of 63 TCP Client & Server States Slide# 61 of 63 Summary We have studied:  Principles behind transport layer services:  multiplexing, demultiplexing  reliable data transfer  flow control  Implementation of the Internet Protocols  UDP  TCP Slide# 62 of 63 References / Links  Chapter #3: Transport Layer, Computer Networking: A Top-Down Approach (7th edition) by Kurose & Ross Slide# 63 of 63