Date post: | 24-May-2015 |
Category: |
Technology |
Upload: | cpqd |
View: | 123 times |
Download: | 0 times |
Faster Networks Faster to the Abyss?
Maximilian Ott CTO, Incoming Media
Sr. Principal Researcher, NICTA [email protected]
Objective
• Carriers need to provide a different service
• From “packet/bit delivery” to “content logistics”
• Outline
• Why nobody likes Carriers
• The Standoff between Carrier & Service Providers
• Incoming: Solving Mobile Video with Data Science
• A Proposal for a new Carrier Service
2
Nobody likes Carriers
3
Voice
SMS
http://bit.ly/1bOmlVl
Not too long ago they actually provided a service customers wanted
Nobody likes Carriers
4
Voice
SMS
http://bit.ly/1bOmlVl
YouTube
Carrier
… Even the Financial Markets
5
… Even the Financial Markets
• Carriers look more & more like utilities
• Valuation (P/E Multiple) will drop accordingly
6
Mexican Standoff
7
Service Provider
Carrier
Advertiser
But it is a Symbiotic Dependency
• Increased User Experience, increased Income
• Great User Experience requires great network
• Service Providers need to incentivize Carriers to help maximizing User Engagement
8
How to Share?
9
Service Provider
Carrier
Advertiser ???
How to Share – Shapley Values?
• Model as a cooperative game which derives a distribution among the players of the total surplus generated by their coalition.
• http://en.wikipedia.org/wiki/Shapley_value
• Ma, et al, Internet Economics: The use of Shapley value for ISP settlement, CoNeXT ‘07
10
More pragmatic – Let’s look at Cost
• What is the REAL cost of delivering a Movie?
• Infrastructure (CapEx) & Operation (OpEx)
• Dominated by Capex
• Capex driven by PEAK demand
• Leads to Congestion Pricing
• Cost varies greatly based on congestion – Orders of Magnitude!
• Same for service providers (servers, connectivity)
11
More pragmatic – Let’s look at Cost
12
Real delivery cost of a Bit can vary by
Orders of Magnitude!
Lower the Peaks, Filling the Valleys
13
Time- shifting
Lower the Peaks, Filling the Valleys
14
Time- shifting
But Data Services cannot do that!
What is the traffic mix?
15
We
b Games
Voice M2M
Live
Anything special about Video?
16
t
“Created”
Hours – days - months
Consumed
FedEx – Packet delivery?
FedEx – Logistics Company!
FedEx – Logistics Company!
The ideal place for zero-latency delivery
Content Logistics
20
Content Logistics – Control Peaks
21
Lowering Cost by Increasing Traffic!
22
Incoming Media
23
The Phone IS part of the network!
• Lots of storage
• Lots of processing
• Big screen
• Lots of sensors
• ONE, very loyal user
• The FIRST screen
24
Incoming Media
• Smart on-device content pre-positioning
• Create great mobile video experiences
• Increase engagement and satisfaction
• Save money doing this
25
Incoming’s Solution
• Pre-fetching content on the mobile utilizing spare network capacity when available.
• Decision what and when to pre-fetch is based on our ability to learn (and predict):
• user behavior
• user’s social network behavior
• mobile’s context and network environment
• operator’s preferences
26
Incoming’s Architecture
27
Incoming TV App
• Proves value proposition of Smart Caching to user
• Over 1.3M installs
• 80% cache-hit ratio
• Top 20% users watching 10 to 20+ mins per day
• Zero mobile data cost
• Instant start, zero stalling, zero buffering, HD video
28
What should Carriers do?
29
Content Logistics Service
• New content-centric network API
• This is a bit different to CCN proposals
• Smart distributed storage
• This includes mobile device (terminals)
• Opportunistic pre-positioning of content
• Less-than-best-effort traffic class
• Only forward when there is nothing else to
30
Content-centric network API
• Declaring of intent
• Probabilistic intent of future use
• Declaring meta-data
• Prediction algorithms rely on discovering correlations
• Providing feedback
• What was the final utility of retrieved content
31
Smart distributed storage
• Storage before bottleneck
• Using meta-data and past behavior to predict:
• what to store
• when to move between storage
• Building on IEEE 2200 (HQME)?
32
New Traffic Class
• Less-than-best-effort traffic class
• Focus on keeping pipes full
• Using IP Header DiffServ Code Points (DSCP)?
• Technology is there
• May require new Traffic Management insights
• Various large carriers are thinking about it
• Interest in exploiting LTE Broadcast
33
Conclusion
• Carriers need to provide a different service
• From “packet/bit delivery” to “content logistics”
• Video is THE dominating traffic source
• Need new technical approach & business model to tackle it
• Move from ‘dumb pipes’ to ‘smart logistics’
34
Faster Networks Faster to the Abyss?
Maximilian Ott CTO, Incoming Media
Sr. Principal Researcher, NICTA [email protected]