9Abtin - Stanford University

EE384Y: Packet Switch Architectures II
Cell Switching vs. Packet Switching
Abtin Keshavarzian
Yashar Ganjali
Department of Electrical Engineering
Stanford University
June 5, 2002
Split
Combine
Motivation
2x2 Switch
June 5, 2002
Cell Switching vs. Packet Switching
2
Outline
 Background: Cells vs. Packets
 Basic extensions of cell switching algorithms
 Stability of packet switching algorithms
 Waiting Algorithms
 Non-waiting Algorithms
 Stability under i.i.d. traffic
 Simulation results
June 5, 2002
Cell Switching vs. Packet Switching
3
Background
 Cell Switching:
 Fixed length cells
 100% throughput using MWM for any
admissible traffic pattern
 Several “fast” algorithms for i.i.d. traffic
 Packet Switching:
 Packets of different length
 Scheduling algorithms?
June 5, 2002
Cell Switching vs. Packet Switching
4
From Cells to Packets
 Algorithm 1: Consider each packet as a
cell with length Lmax and use any cellbased algorithm.
Maximum Packet Length
Current packet
 Algorithm 2: Do the same as 1, except
renew the input-output matching when all
lines are free.
Packet 1
Packet 2
Packet 3
June 5, 2002
Cell Switching vs. Packet Switching
5
Cell-Based -> Packet-Based
 Packet-Based X (PBX):
 Start with any cell-based
algorithm X
 At each time slot, keep
all the lines which are in
the middle of sending a
packet
 For all free lines, recompute a (sub-)matching
using algorithm X
June 5, 2002
a
b
c
d
e
f
g
h
Cell Switching vs. Packet Switching
6
Under any admissible input traffic
IS Packet-Based X
Always Stable?
June 5, 2002
Cell Switching vs. Packet Switching
7
A Counter-example
Time
A 1,1
A 1,2
A 2,1
A 2,2
5
3
8
2
10
6
1
4
7
9
53
2
6
14
June 5, 2002
Cell Switching vs. Packet Switching
8
Waiting vs. Non-Waiting
Algorithms
 Non-Waiting Algorithms:
 Renew the matching amongst
free input-output ports at every
possible time slot.
 Previous example shows that no
non-waiting algorithm is stable in
general.
3
1
 Waiting Algorithms:
 In some time slots, do not start sending
packets even if the corresponding input-output
ports are free.
June 5, 2002
Cell Switching vs. Packet Switching
9
Stability of
Non-Waiting Algorithms
under
i.i.d. Traffic
June 5, 2002
Cell Switching vs. Packet Switching
10
PB-MWM: i.i.d. traffic
1. At time slot n,
find MWM
2. Use the same
matching for
the next k time
slots
a
d
c
b
Lemma: The weight of the matching used by 2
>=
weight{MWM at time (n+k)} - 2Nk
June 5, 2002
Cell Switching vs. Packet Switching
11
PB-MWM: i.i.d. traffic
 Start with MWM at state zero
 Go back to state 0 with probability at
least p
1-p
1-p
p
0
p
1
1-p
3
2
p
p
June 5, 2002
Cell Switching vs. Packet Switching
12
Stability Theorem
 Using previous Lemma for PB-MWM &
 Using the fact that we return to the first
state in a finite number of steps on
average,
 we can show that
E{weight(PB_MWM)} >= weight(MWM) – const
Theorem: PB-MWM is stable for
i.i.d. traffic
June 5, 2002
Cell Switching vs. Packet Switching
13
Simulation Results
June 5, 2002
Cell Switching vs. Packet Switching
14
Simulation Results
June 5, 2002
Cell Switching vs. Packet Switching
15
Conclusion
1. Non-Waiting PB-X algorithms
unstable in general
2. PB-MWM stable for i.i.d. traffic
3. PB-MWM performs slightly better
than CB-MWM for low traffic
June 5, 2002
Cell Switching vs. Packet Switching
16
Questions?