Professional Documents
Culture Documents
1
IPTV and P2PTV architectures
user
DSLAM
Edge
router
IP network
operator
Program
sources
Internet
IPTV P2PTV
2
Which is “better”?
3
Our Model of IPTV
Considered
Backbone User 1
routing edge router One link with capacity C
one domain
Data center
DSLAM
U 2
Router k
…
Us Blocking :
M
Request for channel i
LA
…
DS
is blocked.
…
Causes:
M
LA
1. DSLAM
DS
…
(DSL Access Multiplexer)
One Channel Another Channel 2. Link capacity C
(more popular)
4
IPTV Case study
• Amount of users:
¾ 500,000 subscribers
¾ 41% active in rush hours
¾ 1200 DSLAMs
5
Blocking Results of IPTV case study
-1
10
-2
10
-3
blocking probability B(i)
10
-4
10
-5
10
-6
10
K=170,Q=400,s=171
-7 K=82,Q=400,s=171 K: # of channels
10
K=170,Q=600,s=171 s: # of users/DSLAM
K=170,Q=400,s=205
-8 Q: traffic intensity
10
1 10 100
channel index i
6
Capacity Results of IPTV case study
180
B=10-7
160
B=10-5
needed capacity C (Mb/s)
140
B=10-3
120
100
80
B: E2E blocking
60 Probability in IPTV
1 10 100
channel index i
7
Our Model of P2PTV
Channel i
Ni
Parent u
|P | M
1 2
b wup Blocking :
Next 1-s not available
Other children
Other children when the current 1-s ends.
bw[1, U ]
user U Causes:
Current Next 1 second 1. Chunks unavailability
1 second Next R chunks
2. Parent upload band
r
3. Parent leaves
downloading
displaying at the same time
8
P2PTV Case study
• SopCast
• Measurements on PlanetLab
• What measurements?
¾ Bandwidth usage
¾ Parent distribution and children distribution
¾ QoE
9
Results of P2PTV case study
0.35
K=170, Q=60, N=205200
K=50, Q=60, N=205200
K=170, Q=600, N=205200
0.3
K=170, Q=60, N=2200000
blocking probability b(i)
0.25
0.2
0.15
K: # channels
N: # of users
Q: traffic intensity
0.1
0 10 100
channel index i
10
IPTV vs. P2PTV
11
Summary of our Contribution
12
13