Scadoosh: Scaling Down the Footprint of Rate-Adaptive Live Streaming on CDN Infrastructure
-
Upload
gwendal-simon -
Category
Technology
-
view
404 -
download
3
description
Transcript of Scadoosh: Scaling Down the Footprint of Rate-Adaptive Live Streaming on CDN Infrastructure
Scadoosh:Scaling Down CDNInfrastructureGwendal Simon
Motivations
Akamai’s infrastructure will have to expand by a factorof 100 times in the next five years just to keep up withthe demand for real-time video.
Paul Sagan, Akamai CEO, Jun 2012
2 / 15 Gwendal Simon Scadoosh
Live stream delivery
Content Provider
encodersingestserver
CDN
originserver
edgeservers Clients
disclaimer : ISP strategy is out of the scope of this talk
3 / 15 Gwendal Simon Scadoosh
Live stream delivery
Content Provider
encodersingestserver
CDN
originserver
edgeservers Clients
disclaimer : ISP strategy is out of the scope of this talk
3 / 15 Gwendal Simon Scadoosh
Rate-adaptive video streamingHTTP streaming server client
video segments
requests
time
representation 1
representation 2
representation 3
bit-rate
time
available bandwidth
4 / 15 Gwendal Simon Scadoosh
Rate-adaptive video streamingHTTP streaming server client
video segments
requests
time
representation 1
representation 2
representation 3
bit-rate
time
available bandwidth
4 / 15 Gwendal Simon Scadoosh
Rate-adaptive video streamingHTTP streaming server client
video segments
requests
time
representation 1
representation 2
representation 3
bit-rate
time
available bandwidth
4 / 15 Gwendal Simon Scadoosh
Rate-adaptive video streamingA1 50 kbps 320 x 240 D1 900 kbps 1280 x 720A2 100 kbps 320 x 240 D2 1.2 Mbps 1280 x 720A3 150 kbps 320 x 240 D3 1.5 Mbps 1280 x 720B1 200 kbps 480 x 360 D4 2.0 Mbps 1280 x 720B2 250 kbps 480 x 360 E1 2.5 Mbps 1920 x 1080B3 300 kbps 480 x 360 E2 3.0 Mbps 1920 x 1080B4 400 kbps 480 x 360 E3 4.0 Mbps 1920 x 1080B5 500 kbps 480 x 360 E4 5.0 Mbps 1920 x 1080C1 600 kbps 854 x 480 E5 6.0 Mbps 1920 x 1080C2 700 kbps 854 x 480 E6 8.0 Mbps 1920 x 1080
one video stream = a dozen of representations= more than 20 Mbps
5 / 15 Gwendal Simon Scadoosh
Rate-adaptive video streamingA1 50 kbps 320 x 240 D1 900 kbps 1280 x 720A2 100 kbps 320 x 240 D2 1.2 Mbps 1280 x 720A3 150 kbps 320 x 240 D3 1.5 Mbps 1280 x 720B1 200 kbps 480 x 360 D4 2.0 Mbps 1280 x 720B2 250 kbps 480 x 360 E1 2.5 Mbps 1920 x 1080B3 300 kbps 480 x 360 E2 3.0 Mbps 1920 x 1080B4 400 kbps 480 x 360 E3 4.0 Mbps 1920 x 1080B5 500 kbps 480 x 360 E4 5.0 Mbps 1920 x 1080C1 600 kbps 854 x 480 E5 6.0 Mbps 1920 x 1080C2 700 kbps 854 x 480 E6 8.0 Mbps 1920 x 1080
one video stream = a dozen of representations= more than 20 Mbps
5 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers last-mile ?
last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?
last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning
6 / 15 Gwendal Simon Scadoosh
Identifying the issue
Where is the bottleneck in today’s CDN ?
ISP 1 ISP 2 ISP 3
origin servers
reflectors
edge servers
last-mile ?last-mile ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mile
peering link ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
in edge-servers ?
adaptive streaming → last-mileedge servers in ISP → peering link
commoditized hardware→ server under-provisioning6 / 15 Gwendal Simon Scadoosh
Our objective
Finding a trade-off between :infrastructure capacity for CDN providersquality of experience for clients
Typical services : multiple simultaneous live streamsuser-generated live video servicesmultiview live eventssecond-screen services
7 / 15 Gwendal Simon Scadoosh
Our objective
Finding a trade-off between :infrastructure capacity for CDN providersquality of experience for clients
Typical services : multiple simultaneous live streamsuser-generated live video servicesmultiview live eventssecond-screen services
7 / 15 Gwendal Simon Scadoosh
Scadoosh
8 / 15 Gwendal Simon Scadoosh
Main idea in a nutshell
To not send all representations to all edge servers
9 / 15 Gwendal Simon Scadoosh
Formal Scadooshue
ij = utility score of representation i ofchannel j for edge server e
x eij =
1 if repr i of chn j is delivered to e0 otherwise
objective :max
∑e
∑j
∑i
ueij × x e
ij
10 / 15 Gwendal Simon Scadoosh
Formal Scadooshue
ij = utility score of representation i ofchannel j for edge server e
x eij =
1 if repr i of chn j is delivered to e0 otherwise
objective :max
∑e
∑j
∑i
ueij × x e
ij
10 / 15 Gwendal Simon Scadoosh
Formal Scadooshue
ij = utility score of representation i ofchannel j for edge server e
x eij =
1 if repr i of chn j is delivered to e0 otherwise
objective :max
∑e
∑j
∑i
ueij × x e
ij
10 / 15 Gwendal Simon Scadoosh
Scadoosh Principle
Scadooshcoordinator
1
edge serversreport toScadoosh
coordinator abouttheir activities
during last period
12 Scadooshcoordinatordecides utility
scores, computesdelivery forestoverlay and
notifies sources
2
3 sources use theforest overlays todeliver the live
representations tothe edge servers
3
11 / 15 Gwendal Simon Scadoosh
Scadoosh Principle
Scadooshcoordinator
1
edge serversreport toScadoosh
coordinator abouttheir activities
during last period
1
2 Scadooshcoordinatordecides utility
scores, computesdelivery forestoverlay and
notifies sources
2
3 sources use theforest overlays todeliver the live
representations tothe edge servers
3
11 / 15 Gwendal Simon Scadoosh
Scadoosh Principle
Scadooshcoordinator
1
edge serversreport toScadoosh
coordinator abouttheir activities
during last period
1
2 Scadooshcoordinatordecides utility
scores, computesdelivery forestoverlay and
notifies sources
2
3 sources use theforest overlays todeliver the live
representations tothe edge servers
3
11 / 15 Gwendal Simon Scadoosh
Scadoosh Principle
Scadooshcoordinator
1
edge serversreport toScadoosh
coordinator abouttheir activities
during last period
12 Scadooshcoordinatordecides utility
scores, computesdelivery forestoverlay and
notifies sources
2
3 sources use theforest overlays todeliver the live
representations tothe edge servers
3
11 / 15 Gwendal Simon Scadoosh
What we have done
Formulate an optimization problemPascal Frossard (EPFL)Hervé Kerivin (Clemson then Clermont-Ferrand)
Design optimal and approximate algorithmsJimmy Leblet (Lyon) and Fen Zhou (Avignon)Hanan Shpungin (University of Waterloo)
Design and test a systemCatherine Rosenberg (University of Waterloo)Jiayi Liu (Telecom Bretagne)
12 / 15 Gwendal Simon Scadoosh
What we have done
Formulate an optimization problemPascal Frossard (EPFL)Hervé Kerivin (Clemson then Clermont-Ferrand)
Design optimal and approximate algorithmsJimmy Leblet (Lyon) and Fen Zhou (Avignon)Hanan Shpungin (University of Waterloo)
Design and test a systemCatherine Rosenberg (University of Waterloo)Jiayi Liu (Telecom Bretagne)
12 / 15 Gwendal Simon Scadoosh
What we have done
Formulate an optimization problemPascal Frossard (EPFL)Hervé Kerivin (Clemson then Clermont-Ferrand)
Design optimal and approximate algorithmsJimmy Leblet (Lyon) and Fen Zhou (Avignon)Hanan Shpungin (University of Waterloo)
Design and test a systemCatherine Rosenberg (University of Waterloo)Jiayi Liu (Telecom Bretagne)
12 / 15 Gwendal Simon Scadoosh
Results Overview
120%100% 80% 60% 40% 20%0
0.2
0.4
0.6
0.8
1
provisioning
ratio
ofre
ceiv
eden
codi
ngs
mobile
120%100% 80% 60% 40% 20%0
0.2
0.4
0.6
0.8
1
provisioning
xdsl
120%100% 80% 60% 40% 20%0
0.2
0.4
0.6
0.8
1
ftth lowest representationlow representationnormal representationhigh representationhighest representation
13 / 15 Gwendal Simon Scadoosh
Conclusion
14 / 15 Gwendal Simon Scadoosh
Conclusion
Rate-adaptive video streaming :widely adopted in video servicesthreatening CDN infrastructure
Scadoosh is in preliminary stage :reduce infrastructure by a factor of 5maintain quality of experience
Stay tuned :http ://perso.telecom-bretagne.eu/gwendalsimon
15 / 15 Gwendal Simon Scadoosh
Conclusion
Rate-adaptive video streaming :widely adopted in video servicesthreatening CDN infrastructure
Scadoosh is in preliminary stage :reduce infrastructure by a factor of 5maintain quality of experience
Stay tuned :http ://perso.telecom-bretagne.eu/gwendalsimon
15 / 15 Gwendal Simon Scadoosh
Conclusion
Rate-adaptive video streaming :widely adopted in video servicesthreatening CDN infrastructure
Scadoosh is in preliminary stage :reduce infrastructure by a factor of 5maintain quality of experience
Stay tuned :http ://perso.telecom-bretagne.eu/gwendalsimon
15 / 15 Gwendal Simon Scadoosh