1. Transport Layer: Sliding Window Reliability
Outline
RTT estimation using EWMA
Stop and Wait
Sliding Window Algorithms
CS 640

2. Transport layer functions
Transport protocol defines the pattern/sequence for end hosts sending packets
Transport has to deal with a number of things
Multiplexing/demultiplexing – ports and message queues
Error detection within packets - checksums
Reliable, in order delivery of packets - Today
Flow control – Today
Connection management - TBD
Congestion control – TBD
We’re moving toward understanding TCP
CS 640

3. Methods of Reliability
Packets can be lost and/or corrupted during transmission
Bit level errors due to noise
Loss due to congestion
Use checksums to detect bit level errors
Internet Checksum is optionally used to detect errors in UDP
Uses 16 bits to encode one’s complement sum of data + headers
When bit level errors are detected, packets are dropped
Build reliability into the transmission protocol
Using acknowledgements and timeouts to signal lost or corrupt frame
CS 640

4. Acknowledgements & Timeouts
An acknowledgement (ACK) is a packet sent by one host in response to a packet it has received
Making a packet an ACK is simply a matter of changing a field in the transport header
Data can be piggybacked in ACKs
A timeout is a signal that an ACK to a packet that was sent has not yet been received within a specified timeframe
A timeout triggers a retransmission of the original packet from the sender
How are timers set?
CS 640

5. Acknowledgements & Timeouts
CS 640

6. Propagation Delay
Propagation delay is defined as the delay between transmission and receipt of packets between hosts
Propagation delay can be used to estimate timeout period
How can propagation delay be measured?
What else must be considered in the measurement?
CS 640

7. Exponentially weighted moving average RTT estimation
EWMA was original algorithm for TCP
Measure SampleRTT for each packet/ACK pair
Compute weighted average of RTT
EstRTT = a x EstimatedRTT + b x SampleRTT
where a + b = 1
a between 0.8 and 0.9
b between 0.1 and 0.2
Set timeout based on EstRTT
TimeOut = 2 x EstRTT
CS 640

8. Stop-and-Wait Process
Sender
Receiver
Sender doesn’t send next packet until he’s sure receiver has last packet
The packet/Ack sequence enables reliability
Sequence numbers help avoid problem of duplicate packets
Problem: keeping the pipe full
Example
1.5Mbps link x 45ms RTT = 67.5Kb (8KB)
1KB frames imples 1/8th link utilization
CS 640

9. Solution: Pipelining via Sliding Window
Allow multiple outstanding (un-ACKed) frames
Upper bound on un-ACKed frames, called window
Sender
Receiver T
ime …
CS 640

10. Buffering on Sender and Receiver
Sender needs to buffer data so that if data is lost, it can be resent
Receiver needs to buffer data so that if data is received out of order, it can be held until all packets are received
Flow control
How can we prevent sender overflowing receiver’s buffer?
Receiver tells sender its buffer size during connection setup
How can we insure reliability in pipelined transmissions?
Go-Back-N
Send all N unACKed packets when a loss is signaled
Inefficient
Selective repeat
Only send specifically unACKed packets
A bit trickier to implement
CS 640

11. Sliding Window: Sender
Assign sequence number to each frame (SeqNum)
Maintain three state variables:
send window size (SWS)
last acknowledgment received (LAR)
last frame sent (LFS)
Maintain invariant: LFS - LAR <= SWS
Advance LAR when ACK arrives
Buffer up to SWS frames
SWS
LAR
LFS …
CS 640

12. Sliding Window: Receiver
Maintain three state variables
receive window size (RWS)
largest frame acceptable (LFA)
last frame received (LFR)
Maintain invariant: LFA - LFR <= RWS
Frame SeqNum arrives:
if LFR < SeqNum < = LFA accept
if SeqNum < = LFR or SeqNum > LFA discarded
Send cumulative ACKs – send ACK for largest frame such that all frames less than this have been received
RWS
LFR
LFA …
CS 640

13. Sequence Number Space
SeqNum field is finite; sequence numbers wrap around
Sequence number space must be larger then number of outstanding frames
SWS <= MaxSeqNum-1 is not sufficient
suppose 3-bit SeqNum field (0..7)
SWS=RWS=7
sender transmit frames 0..6
arrive successfully, but ACKs lost
sender retransmits 0..6
receiver expecting 7, 0..5, but receives the original incarnation of 0..5
SWS < (MaxSeqNum+1)/2 is correct rule
Intuitively, SeqNum “slides” between two halves of sequence number space
CS 640

14. Another Pipelining Possibility: Concurrent Logical Channels
Multiplex 8 logical channels over a single link
Run stop-and-wait on each logical channel
Maintain three state bits per channel
channel busy
current sequence number out
next sequence number in
Header: 3-bit channel num, 1-bit sequence num
4-bits total
same as sliding window protocol
Separates reliability from order
CS 640

15. Stop & wait sequence numbers
Sender
Receiver
Sender
Receiver
Sender
Receiver
Frame 0
Frame 0
Frame 0
imeout
imeout
ACK 0 T T
ACK 0
ACK 0
Frame 0
Frame 1
Frame 0
imeout
imeout T
ACK 0
ACK 1
ACK 0 T
Frame 0
(c)
(d)
ACK 0
(e)
Simple sequence numbers enable the client to discard duplicate copies of the same frame
Stop & wait allows one outstanding frame, requires two distinct sequence numbers
CS 640

16. Sliding Window Example
Sender
Receiver 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 A3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 3 4 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 6 A4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14
CS 640

17. Sliding Window Summary
Sliding window is best known algorithm in networking
First role is to enable reliable delivery of packets
Timeouts and acknowledgements
Second role is to enable in order delivery of packets
Receiver doesn’t pass data up to app until it has packets in order
Third role is to enable flow control
Prevents server from overflowing receiver’s buffer
CS 640

18. Chapter 6
The Transport Layer

19. The Transport Service Services Provided to the Upper Layers
Transport Service Primitives
Berkeley Sockets

20. Services Provided to the Upper Layers
The network, transport, and application layers.

21. Why the transport layer ?
1. The network layer exists on end hosts and routers in the network. The end-user cannot control what is in the network. So the end-user establishes another layer, only at end hosts, to provide a transport service that is more reliable than the underlying network service.
2. While the network layer deals with only a few transport entities, the transport layer allows several concurrent applications to use the transport service.
3. It provides a common interface to application writers, regardless of the underlying network layer. In essence, an application writer can write code once using the transport layer primitive and use it on different networks (but with the same transport layer).

22. Transport Service Primitives
The primitives for a simple transport service.

23. Transport Service Primitives (2)
The nesting of TPDUs, packets, and frames.

24. Transport Service Primitives (3)
A state diagram for a simple connection management scheme. Transitions labelled in italics are caused by packet arrivals. The solid lines show the client's state sequence. The dashed lines show the server's state sequence.

25. The socket primitives for TCP.
Berkeley Sockets
The socket primitives for TCP.

26. Elements of Transport Protocols
Addressing
Connection Establishment
Connection Release
Flow Control and Buffering
Multiplexing
Crash Recovery

27. Transport Protocol
(a) Environment of the data link layer.
(b) Environment of the transport layer.
Both data link layer and transport layer do error control, flow control, sequencing. The differences are:
1. Storage capacity in subnet. Frames must arrive sequentially, TPDUs can arrive in any sequence.
2. Frames are delivered to hosts, TPDUs need to be delivered to users, so per user addressing and flow control within the hosts is necessary.

28. Addressing TCP calls TSAP s ... ports ATM calls TSAPs ... AAL-SAP
TSAPs (Transport Service Access Point) , NSAPs (Network SAP).
TCP calls TSAP s ... ports
ATM calls TSAPs ... AAL-SAP

29. Connection Establishment (1)
How a user process in host 1 establishes a connection with a time-of-day server in host 2.

30. Connection Establishment (2)
Three protocol scenarios for establishing a connection using a three-way handshake. CR denotes CONNECTION REQUEST. (a) Normal operation, (b) Old CONNECTION REQUEST appearing out of nowhere. (c) Duplicate CONNECTION REQUEST and duplicate ACK.

31. Connection Establishment (3)
(a) TPDUs may not enter the forbidden region.
(b) The resynchronization problem.

32. Abrupt disconnection with loss of data.
Connection Release
Abrupt disconnection with loss of data.

33. Connection Release (2)
The two-army problem.

34. Connection Release (3)
6-14, a, b
Four protocol scenarios for releasing a connection. (a) Normal case of a three-way handshake. (b) final ACK lost.

35. Connection Release (4)
6-14, c,d
(c) Response lost. (d) Response lost and subsequent DRs lost.

36. Flow Control and Buffering
Dynamic buffer allocation. Buffer allocation info travels in separate TPDUs.
The arrows show the direction of transmission. ‘…’ indicates a lost TPDU.
Potential deadlock if control TPDUs are not sequenced or timed out

37. Multiplexing Upward multiplexing.
Downward multiplexing. Used to increase the bandwidth, e.g., two ISDN connections of 64 kbps each yield 128 kbps bandwidth.

38. The Internet Transport Protocols: UDP
Introduction to UDP
Remote Procedure Call
The Real-Time Transport Protocol

39. Introduction to UDP The UDP header.
UDP only provides TSAPs (ports) for applications to bind to. UDP does not provide reliable or ordered service. The checksum is optional.

40. Steps in making a remote procedure call. The stubs are shaded.

41. The Real-Time Transport Protocol
(a) The position of RTP in the protocol stack. (b) Packet nesting.

42. The Real-Time Transport Protocol (2)
The RTP header. X indicated the presence of an extension header.
CC says how many contributing sources are present (0 to 15).
Syn. Source Id. tells which stream the packet belongs to.
For feedback information is used an associated protocol called
RTCP (Real Time Control Protocol)

43. The Internet Transport Protocols: TCP
Introduction to TCP
The TCP Service Model
The TCP Protocol
The TCP Segment Header
TCP Connection Establishment
TCP Connection Release
TCP Connection Management Modeling
TCP Transmission Policy
TCP Congestion Control
TCP Timer Management
Wireless TCP and UDP
Transactional TCP

44. The TCP Service Model Some assigned ports. Port Protocol Use 21 FTP
File transfer 23
Telnet
Remote login 25
SMTP 69
TFTP
Trivial File Transfer Protocol 79
Finger
Lookup info about a user 80
HTTP
World Wide Web
110
POP-3
Remote access
119
NNTP
USENET news

45. The TCP Service Model (2)
(a) Four 512-byte segments sent as separate IP datagrams.
(b) The 2048 bytes of data delivered to the application in a single READ CALL.

46. TCP Service Model (3)
All TCP connections are full-duplex and point-to-point.
TCP provides a byte stream. i.e it does not preserve message boundaries
At sender TCP may immediately send or buffer data at its discretion.
Sender can use a PUSH flag to instruct TCP not to buffer the send.
Sender can use URGENT flag to have TCP send data immediately and have the receiver TCP signal the receiver application that there is data to be read.

47. Some TCP features Every byte has its own 32 bit sequence number.
Sending and receiving entities exchange data in segments
Each segment is the 20 byte header and data (total up to 64K)
TCP may aggregate multiple writes into one segment or split one write into several segments.
A segment size if the smaller of either 64K or the MTU of the network layer (MTU of Ethernet is about bytes)
A segment must fit in a single IP payload.

48. Some TCP features TCP uses the sliding window protocol as its base.
Sender sends segment, starts timer waits for ack. It no ack then retransmit. Receiver acks in separate segment or “piggyback” on data segment.
TCP must deal with reordred segments.
A lot of algorithms have been developed to make TCP efficient under diverse network conditions. We will look at a few of them.