1. Chapter 6 Congestion Control and Resource Allocation
Outline
6.1 Issues in Resource Allocation
6.2 Queuing Discipline
6.3 TCP Congestion Control
6.4 Congestion Avoidance Mechanisms
6.5 Quality of Service

2. 6.1 Issues in Resource Allocation
Two sides of the same coin
pre-allocate resources so that to avoid congestion
control congestion if (and when) is occurs
Two points of implementation
hosts at the edges of the network (transport protocol)
routers inside the network (queuing discipline)
Underlying service model
best-effort (assume for now)
multiple qualities of service (later)

3. Framework Connectionless flows
sequence of packets sent between source/destination pair
maintain soft state (state information) at the routers
Taxonomy for resource allocation mechanisms
router-centric versus host-centric
reservation-based versus feedback-based
window-based versus rate-based

4. Evaluation Criteria Effectiveness and Fairness
Power (ratio of throughput to delay)

5. 6.2 Queuing Discipline First-In-First-Out (FIFO) Fair Queuing (FQ)
does not discriminate between traffic sources
Fair Queuing (FQ)
explicitly segregates traffic based on flows
ensures no flow captures more than its share of capacity
variation: weighted fair queuing (WFQ)
Problem?

6. FQ Algorithm For single flow
Suppose clock ticks each time a bit is transmitted
Let Pi denote the length of packet i
Let Si denote the time when start to transmit packet i
Let Fi denote the time when finish transmitting packet i
Fi = Si + Pi
When does router start transmitting packet i?
if before router finished packet i - 1 from this flow, then immediately after last bit of i - 1 (Fi-1)
if no current packets for this flow, then start transmitting when arrives (call this Ai)
Thus: Fi = MAX (Fi - 1, Ai) + Pi

7. FQ Algorithm (cont) For multiple flows
calculate Fi for each packet that arrives on each flow
treat all Fi’s as timestamps
next packet to transmit is one with lowest timestamp
Not perfect: can’t preempt current packet
Example
The link is never left idle as long as there is at least one packet in the queue (work-conserving)
If the link is fully loaded and there are n flows sending data, then I cannot use more than 1/n of the link bandwidth.
Flow 1
Flow 2
(a)
(b)
Output
F = 8
F = 10
F = 5
F = 2
(arriving)
(transmitting)

8. 6.3 TCP Congestion Control
Idea
TCP assumes best-effort network (FIFO or FQ routers) each source determines network capacity for itself
uses implicit feedback
ACKs pace transmission (self-clocking)
Challenge
determining the available capacity in the first place
adjusting to changes in the available capacity

9. 6.3.1 Additive Increase/Multiplicative Decrease (AIMD)
Objective: adjust to changes in the available capacity
New state variable per connection: CongestionWindow
limits how much data source has in transit
MaxWin = MIN(CongestionWindow, AdvertisedWindow)
EffWin = MaxWin - (LastByteSent LastByteAcked)
Idea:
increase CongestionWindow when congestion goes down
decrease CongestionWindow when congestion goes up

10. AIMD (cont)
Question: how does the source determine whether or not the network is congested?
Answer: a timeout occurs
timeout signals that a packet was lost
packets are seldom lost due to transmission error
lost packet implies congestion

11. AIMD (cont) Algorithm In practice: increment a little for each ACK
Source
Destination
Algorithm
increment CongestionWindow by one packet per RTT (linear increase)
divide CongestionWindow by two whenever a timeout occurs (multiplicative decrease)
In practice: increment a little for each ACK
Increment = (MSS * MSS)/CongestionWindow
CongestionWindow += Increment

12. AIMD (cont)
Trace: sawtooth behavior

13. 6.3.2 Slow Start
Objective: determine the available capacity in the first
Effectively increases the congestion window exponentially, rather than linearly
Idea:
begin with CongestionWindow = 1 packet
double CongestionWindow each RTT (increment by 1 packet for each ACK)
Why called “slow start” ? It is compared to the original behavior of TCP, not to the linear mechanism
Source
Destination

14. Slow Start (cont) Exponential growth, but slower than all at once
Used in two different situations …
when first starting connection
when connection goes dead waiting for timeout
Trace
Colored line = value of CongestionWindow, Solid bullets = timeouts, Hash marks = time when each packet is transmitted, vertical bars = time when a packet that was eventually retransmitted was first transmitted
Problem: lose up to half a CongestionWindow’s worth of data if the source is aggressive at the beginning, as TCP is during exponential growth

15. 6.3.3 Fast Retransmit and Fast Recovery
Problem: coarse-grained implementation of TCP timeouts lead to long periods during which the connection went dead and while waiting for a timer to expire
Fast retransmit: use duplicate ACKs to trigger retransmission
TCP waits for 3 duplicate ACKs

16. Results (Fast Retransmission)
The long periods during which the congestion window stays flat and no packets are sent have been eliminated
Eliminated about half of the coarse-grained timeouts
Fast recovery
skip the slow start phase that happens between when fast retransmit detects a lost packet and additive increase begins
go directly to half the last successful CongestionWindow
Slow start is only used at the beginning of a connection and whenever a coarse-grained timeout occurs. At all other times, the congestion window is following a pure AIMD pattern.

17. 6.4 Congestion Avoidance Mechanisms
TCP’s strategy
control congestion once it happens, not trying to avoid congestion in the first place
repeatedly increase load in an effort to find the point at which congestion occurs, and then back off
TCP needs to increase losses to find the available bandwidth of the connection
Alternative strategy
predict when congestion is about to happen
reduce rate before packets start being discarded
call this congestion avoidance, instead of congestion control
Two different mechanisms
router-centric: DECbit and RED Gateways
host-centric: TCP Vegas

18. 6.4.1 DECbit Add binary congestion bit to each packet header Router
monitors average queue length over last busy+idle cycle
set congestion bit if average queue length > 1
attempts to balance throughout against delay

19. End Hosts Destination echoes bit back to source
Source records how many packets resulted in set bit
If less than 50% of last window’s worth had bit set
increase CongestionWindow by 1 packet
If 50% or more of last window’s worth had bit set
decrease CongestionWindow by times

20. 6.4.2 Random Early Detection (RED)
Invented by Sally Floyd and Van Jacobson in the early 1990s, differs from the DECbit in two major ways
Notification is implicit
just drop the packet (TCP will timeout)
could make explicit by marking the packet
Early random drop
rather than wait for queue to become full, drop each arriving packet with some drop probability whenever the queue length exceeds some drop level

21. RED Details Compute average queue length
AvgLen = (1 - Weight) * AvgLen +
Weight * SampleLen
0 < Weight < 1 (usually 0.002)
SampleLen is queue length each time a packet arrives
The weighted running average calculation tries to detect long-lived congestion, by filtering out short-term changes in the queue length

22. RED Details (cont)
Weighted Running Average Queue Length

23. RED Details (cont) Two queue length thresholds
if AvgLen <= MinThreshold then
enqueue the packet
if MinThreshold < AvgLen < MaxThreshold then
calculate probability P
drop arriving packet with probability P
if ManThreshold <= AvgLen then
drop arriving packet

24. RED Details (cont) Computing probability P Drop Probability Curve
TempP = MaxP * (AvgLen - MinThreshold)/ (MaxThreshold - MinThreshold)
P = TempP/(1 - count * TempP)
Drop Probability Curve

25. Tuning RED
Probability of dropping a particular flow’s packet(s) is roughly proportional to the share of the bandwidth that flow is currently getting
MaxP is typically set to 0.02, meaning that when the average queue size is halfway between the two thresholds, the gateway drops roughly one out of 50 packets.
If traffic is bursty, then MinThreshold should be sufficiently large to allow link utilization to be maintained at an acceptably high level
Difference between two thresholds should be larger than the typical increase in the calculated average queue length in one RTT; setting MaxThreshold to twice MinThreshold is reasonable for traffic on today’s Internet
Penalty Box for Offenders

26. 6.4.3 Source-based Congestion Avoidance -- TCP Vegas
Idea: source watches for some sign that router’s queue is building up and congestion will happen too; e.g.,
RTT grows
Sending rate flattens

27. Algorithms By RTT Grows
Algorithm 1: The congestion window normally increases as in TCP, but every two round-trip delays, the algorithm checks to see if the current RTT is greater than the average of the minimum and maximum RTTs seen so far. If it is, then the algorithm decreases the congestion window by one-eighth.
Algorithm 2: The window is adjusted once every ywo round-trip delays based on the product
w = (CurrentWindow –OldWindow)x(CurrentRTT – OldRTT)
If w > 0, the source decreases the window size by 1/8,
If w <=0, the source increases the window by one maximum packet size.

28. Algorithms by sending rate flattens
Algorithm 3: Every RTT, it increases the window size by one packet and compares the throughput achieved to the throughput when the window was one packet smaller.
If the difference is less than ½ the throughput achieved when only one packet was in transit, the algorithm decreases the window by one packet.
This scheme calculates the throughput by dividing the number of bytes outstanding in he network by the RTT.
Algorithm 4: TCP Vegas. The goal is to maintain the “right” amount of extra data in the network. If a source is sending two much extra data, it will cause long delays and possibly lead to congestion. If a connection is sending too little extra data, it cannot responds rapidly enough to transient increases in the available network bandwidth.

29. TCP Vegas
Congestion window vs observed throughput rate (the three graphs are synchronized).

30. TCP Vegas Algorithm
Let BaseRTT be the minimum of all measured RTTs (commonly the RTT of the first packet)
If not overflowing the connection, then
ExpectRate = CongestionWindow/BaseRTT
Source calculates sending rate (ActualRate) once per RTT
Source compares ActualRate with ExpectRate
Diff = ExpectedRate - ActualRate
if Diff < a (too little extra data)
increase CongestionWindow linearly in next RTT
else if Diff > b (too much extra data)
decrease CongestionWindow linearly in next RTT
else
leave CongestionWindow unchanged

31. TCP Vegas Algorithm (cont)
Parameters
a = 1 packet
b = 3 packets
The goal is to keep the ActualRate between these two thresholds, that is, within the region
TCP Vegas does use multiplicative decrease when a timeout occurs; the linear decrease is an early decrease in the congestion window that, hopefully, happens before congestion occurs.