Adaptive Explicit Congestion Notification AECN

1
Adaptive Explicit Congestion Notification (AECN)

• Zici Zheng and Robert Kinicki
• Worcester Polytechnic Institute
• Computer Science Department
• Worcester, MA 01609
• USA

2
Outline

• Motivation for AECN
• Performance Metrics
• Random Early Detection (RED) and ECN Routers
• Topology and Experimental Procedures
• RED and ECN Results
• AECN Results
• Conclusions

3
Motivation for Adaptive ECN

• Congestion is still an Internet problem.
• Researchers advocate Active Queue Management
  (AQM) techniques such as RED and ECN for
  congestion control.
• RED is difficult to tune and unfair.
• ECN is better when it marks.

4
Motivation for Adaptive ECN

• Is ECN also unfair to heterogeneous flows?
• What happens when there are many flows?
• Previously shown that ECN performs better with a
  higher mark probability when there are many
  flows.
• Adaptive ECN can improve goodput and fairness.

5
Performance Metrics

• throughput (Mbps) - the aggregate rate of
  packets generated by all sources.
• goodput (Mbps) - the rate at which packets arrive
  at the receiver. Goodput differs from throughput
  in that retransmissions are excluded from
  goodput.
• delay (sec) - the time required to transmit a
  packet from source node to receiver node.

6
Performance Metrics

• Jains fairness
• For any given set of user throughputs (x1, x2, ,
  xn), the fairness index to the set is defined
• f (x1, x2, , xn)
• max-min fairness
• A flow rate x is max-min fair if any rate x
  cannot be increased without decreasing some y
  which is smaller than or equal to x. To satisfy
  the min-max fairness criteria, the smallest
  throughput rate must be as large as possible.
• visual max-min fairness
• the visual gap between the smallest and the
  largest goodput

7
RED Routers

• Random Early Detection (RED) detects congestion
  early by maintaining an exponentially-weighted
  average queue size.
• RED probabilistically drops packets before the
  queue overflows to signal congestion to TCP
  sources.
• RED attempts to avoid global synchronization and
  bursty packet drops.

8
ECN Routers

• Explicit Congestion Notification (ECN) , a RED
  variant, marks packets to signal congestion.
• ECN must be supported by both TCP senders and
  receivers.
• ECN-compliant TCP senders initiate their
  congestion avoidance algorithm after receiving
  marked ACK packets from the TCP receiver.
• Packets from non-ECN compliant flows are treated
  by the RED mechanism in the ECN router.

9
RED and ECN Router Parameters

• avgq average queue size
• avgq (1-wq) avgq wq instantaneous
  queue size
• wq weighting factor 0.001 lt
  wq lt 0.004
• minth average queue length threshold for
  triggering probabilistic
  drops/marks.
• maxth average queue length threshold for
  triggering forced drops
• maxp maximum dropping/marking probability
• pb maxp (avgq minth) / (maxth
  minth)
• pa pb / (1 count pb)
• buffer_size the size of the router queue in
  packets

10
RED/ECN Router Mechanism
1
Dropping/Marking Probability
maxp
0
Min-threshold
Queue Size
Max-threshold
Average Queue Length (avgq)
11
Simulation Topology
12
Experimental Procedures and Parameter Settings

• 100 second ns-2 simulations
• n flows divided equally among three flow types
  (fragile, average, robust) (n 3m)
• aggregate flow capacity fixed at 90 Mbps
• staggered start of half the flows (0 sec, 2 sec)
• fixed RED/ECN/AECN and TCP parameters for all
  runs
• wq 0.001
• minth 10 maxth 30
• buffer_size 50 packets
• TCP max_send_window_size 64 packets

13
Figure 2 RED and ECN Goodput
14

Figure 3 RED and ECN Fairness
15
Figure 4 RED and ECN Goodput 60 flows, maxp
0.5
16
60 Flows, maxp 0.5
ECN has almost no drops !!
ECN Marks
RED Drops
ECN Drops
17
120 Flows, maxp 0.5
ECN drops are now significant!
ECN Marks
RED Drops
ECN Drops
18
RED/ECN Conclusions

• ECN better than RED especially if ECN maxp set
  higher.
• RED/ECN unfair to fragile and average flows gt
  adaptive maxp needed.
• ECN needs to avoid drops when there are many
  flows.

19
Adaptive ECN flow queues
                           
20
AECN Algorithm

• If avgq gt maxth , drop incoming packet
  same as ECN
• If avgq is below maxth ,
• Add incoming packet to the router queue
  
• Determine whether flow is robust, fragile or
  average
• and add to the appropriate flow
  queue
• If avgq is between minth and maxth ,
• Determine mark probability (maxp)
• and probabilistically mark the first
  unmarked packet
• at the front of the appropriate flow
  queue

21
Determine Mark Probability (maxp)

Robust Flow maxp min (base-maxp ?) ,
1 Average Flow maxp base-maxp Fragile
Flow maxp base-maxp / ?
22
How to choose ? and ? ?

• For this research, assume ? ?
• Goal achieve fairness for fragile and average
  flows
• Pay attention to number of flows

23
Figure 5 AECN Goodput60 flows, base_maxp 0.5
Alpha 2.5 is fairest !!
24
Figure 6 AECN Jains Fairness60 flows,
base_maxp 0.5
Alpha 2.5 is fairest !!
25
Figure 7 AECN Goodput 120 flows, base_maxp 0.8
26
Figure 8 AECN Jains Fairness120 flows,
base_maxp 0.8
Alpha 2.5 is fairest !!
27
Figure 9 AECN Goodputbase_maxp 0.5, ? ?
2.5
28
Figure 10 AECN Goodputbase_maxp 0.8, ? ?
2.5
29
Figure 11 Jains Fairnessbase_maxp 0.5, ? ?
2.5
30
Figure 12 Jains Fairnessbase_maxp 0.8, ? ?
2.5
31
AECN Conclusions

• AECN provides higher goodput when there are a
  larger number of flows.
• Both visual max-min fairness and Jains
  fairness are better for AECN.
• Adapting to both RTT and number of flows is shown
  to be important.
• ? ? 2.5 good settings for these experiments.

32
Future Work

• Find method to adjust maxp as function of RTT
  source hint to eliminate flow classes gt see
  Chablis paper (Choong-Soo Lee, Mark Claypool, and
  Robert Kinicki. Chablis - Achieving Fair
  Bandwidth Allocation with Priority Dropping Based
  on Round Trip Time, WPI-CS-TR-02-19, May 2002,
  ftp//ftp.cs.wpi.edu/pub/techreports/02-19.ps.gz
  )
• Include flow count at router in determining drop
  probability.
• Avoid ECN drops when avgq gets close to maxth .