Professional Documents
Culture Documents
Won S. Yoon
wonsyoonalum.mit.edu
ISIT 2008
Introduction
Discussion
Introduction
Introduction
Introduction
Physical Layer
Network Layer
Discussion
Introduction
Introduction Traditionally,
Physical Layer
• Assume an infinite backlog of data.
Network Layer
• Try to maximize throughput and minimize error.
Discussion
• Solution: encode a large number of bits into a long codeword.
Introduction Traditionally,
Physical Layer
• Assume an infinite backlog of data.
Network Layer
• Try to maximize throughput and minimize error.
Discussion
• Solution: encode a large number of bits into a long codeword.
Network Layer
Discussion
Discussion
τ τ τ τ
d1 d1 d1 d1
τ τ τ τ
d1 d1 d1 d1
τ τ τ τ
d1 d1 d1 d1
τ τ τ τ τ τ τ τ
d3 d3
τ τ τ τ
d1 d1 d1 d1
τ τ τ τ τ τ τ τ
d3 d3
Discussion
Network Layer
Discussion
τ τ τ τ τ τ τ τ
d3 d3
τ τ τ τ τ τ τ τ
d3 d3
τ τ τ τ τ τ τ τ
d3 d3
τ τ τ τ τ τ τ τ
d3 d3
τ τ τ τ τ τ τ τ
d3 d3
Introduction Rate control: adapt the rate of packets encoded into the stream.
Physical Layer
Introduction Rate control: adapt the rate of packets encoded into the stream.
Physical Layer
Introduction Rate control: adapt the rate of packets encoded into the stream.
Physical Layer
End-point solutions:
1. Admit one packet at a time (patient): per-packet block code.
2. Admit all packets at once (greedy): maximally-joint block code.
Introduction Rate control: adapt the rate of packets encoded into the stream.
Physical Layer
End-point solutions:
1. Admit one packet at a time (patient): per-packet block code.
2. Admit all packets at once (greedy): maximally-joint block code.
Introduction
Network Layer
Rate Encoder Channel Decoder
ctrl
Discussion queue
data
packets
Introduction
Network Layer
Rate Encoder Channel Decoder
ctrl
Discussion queue
data
packets
Introduction
Network Layer
Rate Encoder Channel Decoder
ctrl
Discussion queue
data
packets
• Stream codes
• Stream codes
– Information-theoretic view: coding and error exponents.
• Stream codes
– Information-theoretic view: coding and error exponents.
– No known studies of delay for bursty data.
• Stream codes
– Information-theoretic view: coding and error exponents.
– No known studies of delay for bursty data.
Network Layer
Discussion
Network Layer
1. Physical layer: random-coding error exponents.
Discussion
Network Layer
1. Physical layer: random-coding error exponents.
Discussion
2. Network layer: queueing model, end-to-end delay.
Physical Layer
Network Layer
Discussion
Physical Layer
Network Layer
Discussion
Discussion
Discussion
• Slide the next packet into register as message Ut .
– If no packet, encode null symbol (known at receiver).
Discussion
• Slide the next packet into register as message Ut .
– If no packet, encode null symbol (known at receiver).
Discussion
• Slide the next packet into register as message Ut .
– If no packet, encode null symbol (known at receiver).
Network Layer
N(t)
Discussion
Network Layer
N(t)
Discussion
Network Layer
N(t)
Discussion
Network Layer
N(t)
Discussion
Network Layer
N(t)
Discussion
Network Layer
Discussion
subtrees
0 with correct
time-T bit.
1
subtrees
0 with correct
time-T bit.
1
subtrees
0 with correct
time-T bit.
1
Network Layer
Discussion
Discussion
Network Layer
Discussion
Discussion
• Stream Codes:
• Stream Codes:
– For ML [Forney74℄ and sequential [Jelinek74,Anderson92℄
fixed-rate decoding, the block-coding RCE applies.
– What about for per-packet MAP decoding?
Introduction
Physical Layer
For MAP decoding of stage t after delay d:
Network Layer "N #
Xt
Discussion
Pe,t,d ≤ e−dj E0 (ρ)+jmρ e−dE0 (ρ)+(1+Nd )mρ
k=0
Introduction
Physical Layer
For MAP decoding of stage t after delay d:
Network Layer "N #
Xt
Discussion
Pe,t,d ≤ e−dj E0 (ρ)+jmρ e−dE0 (ρ)+(1+Nd )mρ
k=0
Nt Nd
0 t t+d
d ≥ λ
1− µ
E0 1 1
– µ= mρ and σ = E0 log ǫ
PNt
– wpast = log j=0 e−dj E0 +jmρ
Physical Layer
Network Layer
Discussion
Network Layer
Network Layer
Discussion
Introduction
Physical Layer
Network Layer
Discussion
Introduction
Physical Layer
Network Layer
Discussion
Introduction
Physical Layer
Network Layer
Discussion
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
Network Layer
Discussion
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
However, the delay due to the past is more complicated:
Network Layer
Discussion
h P i
1 Nt −dj E0 +jmρ
past mρ E log j=0 e
Dstream =
µ−λ
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
However, the delay due to the past is more complicated:
Network Layer
Discussion
h P i
1 Nt −dj E0 +jmρ
past mρ E log j=0 e
Dstream =
µ−λ
Bounds:
N
X
E max Sj ≤ E log eSj
0≤j≤N
j=0
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
However, the delay due to the past is more complicated:
Network Layer
Discussion
h P i
1 Nt −dj E0 +jmρ
past mρ E log j=0 e
Dstream =
µ−λ
Bounds:
N
X XN
E max Sj ≤ E log eSj ≤ log E eSj
0≤j≤N
j=0 j=0
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
However, the delay due to the past is more complicated:
Network Layer
Discussion
h P i
1 Nt −dj E0 +jmρ
past mρ E log j=0 e
Dstream =
µ−λ
Bounds:
N
X XN
E max Sj ≤ E log eSj ≤ log E eSj
0≤j≤N
j=0 j=0
1
Introduction The future is simple: each packet imposes another µ workload.
Physical Layer
However, the delay due to the past is more complicated:
Network Layer
Discussion
h P i
1 Nt −dj E0 +jmρ
past mρ E log j=0 e
Dstream =
µ−λ
Bounds:
N
X XN
E max Sj ≤ E log eSj ≤ log E eSj
0≤j≤N
j=0 j=0
Network Layer
Discussion
1
Compare with block codes in the regime λ > d1 :
1
Compare with block codes in the regime λ > d1 :
∗ σµ + 1
Dblock =
µ−λ
1
Compare with block codes in the regime λ > d1 :
∗ σµ + 1 1 σµ + 1
Dblock = +
µ−λ 2 µ−λ
1
Compare with block codes in the regime λ > d1 :
∗ σµ + 1 1 σµ + 1
Dblock = +
µ−λ 2 µ−λ
• Comparing the second terms (delay due to the past),
1
Compare with block codes in the regime λ > d1 :
∗ σµ + 1 1 σµ + 1
Dblock = +
µ−λ 2 µ−λ
• Comparing the second terms (delay due to the past),
1 1
mρ log 1−β 1 σµ + 1
<
µ−λ 2 µ−λ
1
Compare with block codes in the regime λ > d1 :
∗ σµ + 1 1 σµ + 1
Dblock = +
µ−λ 2 µ−λ
• Comparing the second terms (delay due to the past),
1 1
mρ log 1−β 1 σµ + 1
<
µ−λ 2 µ−λ
∗ ∗
=⇒ Dstream < Dblock for high throughput.
Network Layer
Discussion
Discussion
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
σµ + 1 ∗
≤ E[Dstream ]
µ−λ
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
σµ + 1 ∗ σµ + 1
≤ E[Dstream ] ≤
µ−λ µ−λ
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
h P i
1 Nt −dj E0 +jmρ
σµ + 1 ∗ σµ + 1 mρ E log j=0 e
≤ E[Dstream ] ≤ +
µ−λ µ−λ µ−λ
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
h P i
1 Nt −dj E0 +jmρ
σµ + 1 ∗ σµ + 1 mρ E log j=0 e
≤ E[Dstream ] ≤ +
µ−λ µ−λ µ−λ
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
h P i
1 Nt −dj E0 +jmρ
σµ + 1 ∗ σµ + 1 mρ E log j=0 e
≤ E[Dstream ] ≤ +
µ−λ µ−λ µ−λ
Discussion
• Use a technique from [MusyTelatar06℄: find bounds and show
that bounds are very close.
h P i
1 Nt −dj E0 +jmρ
σµ + 1 ∗ σµ + 1 mρ E log j=0 e
≤ E[Dstream ] ≤ +
µ−λ µ−λ µ−λ
50000
Introduction
Physical Layer
Network Layer
40000
Discussion
30000
delay (slots)
20000
10000
0
Physical Layer
Network Layer
Discussion
Discussion
Future work:
• Simplify the past workload expression.
• Examine more bursty arrivals.
– Greedy policies not optimal anymore?