1. Advanced Computer Networks
CS716
Advanced Computer Networks
By Dr. Amir Qayyum 1

2. Lecture No. 32

3. TCP Flow Control Issues
Problem: app. delivers tiny pieces of data to TCP
e.g. telnet in character mode
Each piece sent as segment, returned as ACK
Very inefficient
Solutions
Delay transmission to accumulate more data
Nagle’s algorithm
Send first piece
Accumulate data until first piece ACK’d
Send accumulated data and restart accumulation
Not ideal for some traffic, e.g. mouse motion

4. TCP Flow Control Issues
Problem: slow application reads data in tiny pieces
Receiver advertises tiny window
Sender fills tiny window
Known as silly window syndrome
Solution: due to Clark
Advertise window opening only when MSS or ½ of buffer is available
Sender delays sending until window is MSS or ½ of receiver’s buffer (estimated)
Overridden by using PUSH

5. TCP Flow Control Math Send buffer size: MaxSendBuffer
Receive buffer size: MaxRcvBuffer
Receiving side
LastByteRcvd - LastByteRead < = MaxRcvBuffer
AdvertisedWindow = MaxRcvBuffer - (NextByteExpected NextByteRead)
Sending side
LastByteSent - LastByteAcked < = AdvertisedWindow
EffectiveWindow = AdvertisedWindow - (LastByteSent LastByteAcked)
LastByteWritten - LastByteAcked < = MaxSendBuffer
block sender if (LastByteWritten - LastByteAcked) + y > MaxSenderBuffer
Always send ACK in response to arriving data segment
Persist when AdvertisedWindow = 0

6. TCP Bit Allocation Limitations
Sequence numbers vs packet lifetime
Assumed that IP packets live less than 60s
Can we send 232 (4G) bytes in 60 seconds ?
Only need a data rate of 573 Mbps!
Less than an STS-12 line... (less than Gigabit Ethernet)
Advertised window vs delay-bandwidth
Only 16 bits (64kB) for advertised window
For cross-country RTT of 100 milliseconds, adequate for a mere 5.24 Mbps!

7. Protection Against Wrap Around
32-bit SequenceNum
Bandwidth Time Until Wrap Around
T1 (1.5 Mbps) 6.4 hours
Ethernet (10 Mbps) 57 minutes
T3 (45 Mbps) 13 minutes
FDDI (100 Mbps) 6 minutes
STS-3 (155 Mbps) 4 minutes
STS-12 (622 Mbps) 55 seconds
STS-24 (1.2 Gbps) 28 seconds

8. Keeping the Pipe Full 16-bit AdvertisedWindow
Bandwidth Delay x Bandwidth Product
T1 (1.5 Mbps) 18KB
Ethernet (10 Mbps) 122KB
T3 (45 Mbps) 549KB
FDDI (100 Mbps) 1.2MB
STS-3 (155 Mbps) 1.8MB
STS-12 (622 Mbps) 7.4MB
STS-24 (1.2 Gbps) 14.8MB

9. Adaptive Retransmission Algorithm
Original algorithm used only RTT estimate
Theory: measure RTT for each segment + its ACK
Estimate RTT
Timeout is 2 × RTT to allow for variations

10. Adaptive Retransmission Algorithm
Practice
Use exponential moving average ( = 0.8 to 0.9)
Estimate =  × estimate + (1 - ) measurement
Measured RTT
depends on 
time

11. Adaptive Retransmission Algorithm
Problem: it did not handle variations well
Ambiguity for retransmitted packets: was ACK in response to first, second, etc. transmission ?
Measured RTT
time
transmission
retransmission
RTT ? ? ?

12. Adaptive Retransmission (Original Algorithm)
Measure SampleRTT for each segment/ACK pair
Compute weighted average of RTT
EstRTT = a × EstRTT + b × 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 × EstRTT

13. Karn/Partridge Algorithm
Sender
Receiver
Sender
Receiver
Original transmission
Original transmission TT TT
ACK
Retransmission
SampleR
SampleR
Retransmission
ACK
Do not sample RTT when retransmitting
Double timeout (RTT estimate) after each retransmission
Still did not handle variations well
Did not solve network congestion problems as desired

14. Jacobson/Karels Algorithm
Estimate variance of RTT
Calculate mean interpacket RTT deviation as an approximation of variance
(Diff = RTT_estimate – measurement)
RTT_estimate =
a x RTT_estimate + (1 - a ) × measurement
Using another exponential moving average
Deviation =
β x |RTT_estimate - measurement| + (1 - β) deviation
β = 0.25, α = in RTT_estimate

15. Jacobson/Karels Algorithm
Use variance estimate as component of RTT estimate
next_RTT (timeout) =
RTT_estimate + 4 × deviation
Protects against high jitter
Algorithm only as good as clock granularity(500ms on Unix)
Accurate timeout mechanism important to congestion control