Unformatted text preview: additional algorithms. Two of these, FIFO and RED,
represent baseline cases where routers do not attempt to
achieve fair bandwidth allocations. The other two algorithms, FRED and DRR, represent di erent approaches to
FIFO First In First Out - Packets are served in a
rst-in rst-out order, and the bu ers are managed
using a simple drop-tail strategy; i.e., incoming packets are dropped when the bu er is full.
RED Random Early Detection - Packets are served
in a rst-in rst-out order, but the bu er management is signi cantly more sophisticated than drop-tail.
RED 9 starts to probabilistically drop packets long 1 1.8
FIFO 0.8 DRR
FIFO 1.6 1.4 Bandwidth (Mbps) Bandwidth (Mbps) 1.2
0.6 0.4 1 0.8 0.6 0.4 0.2 0.2 0 0
5 10 15
Flow Number 20 25 30 a 5 10 15
Flow Number 20 25 30 b Figure 3: Simulation results for a 10 Mbps link shared by N ows. a The average throughput over 10 sec when N = 32,
and all ows are UDPs. The arrival rate for ow i is i + 1 times larger than its fair share. The ows are indexed from 0.
b The throughputs of one UDP ow indexed 0 sending at 10 Mbps, and of 31 TCP ows sharing a 10 Mbps link.
before the bu er is full, providing early congestion
indication to ows which can then gracefully backo before the bu er over ows. RED maintains two
bu er thresholds. When the exponentially averaged
bu er occupancy is smaller than the rst threshold, no
packet is dropped, and when the exponentially averaged bu er occupancy is larger than the second threshold all packets are dropped. When the exponentially
averaged bu er occupancy is between the two thresholds, the packet dropping probability increases linearly
with bu er occupancy.
FRED Flow Random Early Drop - This algorithm
extends RED to provide some degree of fair bandwidth allocation 14 . To achieve fairness, FRED maintains state for all ows that have at least one packet
in the bu er. Unlike RED where the dropping decision is based only on the bu er...
View Full Document
- Fall '10
- Eugene Ng
- Scheduling algorithm, Round-robin scheduling, Scheduling algorithms, Ow, Fair queuing, ows