Internet and CDNs
The Internet Market Trends
Source: AT Kearney: A Viable Future Model for the Internet
The Internet Value Chain
Market Segments Capitilization
Internet Structure
• BE, connectivity market, two-sided
End user trends and Demand (I)
• High broadband penetration
• Increasing rate of Internet subscriptions – 18 million households in Europe came online between
2005 and 2008
• Internet overtakes infotainment business – User spend online more time than watching tv, reading
press or playing video games
– Internet consumption in 2010 will average 14,2 hours per week compared to 11,5 hours per week for TV
• Trend for Internet applications with need for QoS – IPTV, video on demand, gaming
End user trends and Demand (II)
• “To Generation Digital, TV means video, delivered on demand. In fact, one in seven 18-24 year olds now watch no live TV at all” – “Europe Logs On”, European Internet trends of Today and Tomorrow,
Microsoft, April 2009
• Demand for video, gaming, social networks – Growing volume of mobile phones integrating these applications
– Social networking integrating virtual goods, games
– Increasing market value of Over-The-Top providers active in these domains
• Content tailored to market segment and needs – E.g. short videos that make them easy for people on the move to watch,
targeting for mobile customers
• Mobile users typically pay to download content and apps from third party providers
Related Data and Projections (I)
• Hours per week spent on Internet increasing while other infotainment segments shrink
Related Data and Projections (II)
Social Networking versus Email
Major Consumer Trends
• Online services increasingly richer and migrating to a plethora of devices – Different viewing capabilities, QoS requirements
• Seamless entertainment, IPTV
• Mobile part of the entertainment mix – Smartphones increasingly popular and with built-in OTT services
capabilities
– “Mobile Internet will soon overtake Fixed Internet”, Mary Meeker of Morgan Stanley
• eReaders expected to increase in popularity – But still in infancy…
• Increasing popularity of video and voice services over the Internet, high content generation and downloading rates
Pressures on the Internet
Performance Pressure (I)
• Exponential growth of traffic
Performance Pressure (II)
• New services – More demanding in resources:
Internet-TV, Catchup-TV, streaming
• Increasing penetration of multimedia devices – Smartphones
• Changing usage patterns – Always on
• Rise in mobile Internet
• Few to many flows – E.g. Google 3rd in overall global
traffic carried > Tier-1 (such as Sprint, Cogent)
– Asymmetry • 1:10 or 1:20 for content
Economic Pressure (I)
• Structural problem: pricing disconnected from usage (flat fee)
Economic Pressure (II)
• Investments required impact cost
– CAPEX
– OPEX
• Methodology:
€9Bn
Mobile Traffic Growth
• Traffic growth requires deployment of new technologies and services (e.g. 3G)
– More capex and opex expenditures
Economic Pressure of Mobile
• Main findings:
Policy Pressure
• Regulated pricing
• Competition, unbundling
• Net neutrality debate
Key Issues
• Increased cost of handling the exponentially increasing traffic – is not matched by additional revenues for a key market
segment: network operators
• Pressures on Internet will only increase, calling for more difficult decisions on where to invest and when – Identified limitations could hinder the future innovation of
new services for business and critical applications due to high costs incurred
• E.g. telemedicine
• VoIP does not provide emergency calls
Potential solutions
1. Continue with the current commercial model and raise additional revenue required by modifying retail pricing schemes to increase average revenue per user (ARPU)
2. Introduce a data-conveyance charge to be paid by traffic senders based on the total volume of traffic they send
3. Develop optional enhanced traffic delivery services over the public Internet in a coordinated manner, with a price premium based on quality of service delivery
4. Develop new services where Connectivity Providers offer Online Service Providers managed services over their networks on a bilateral commercial basis
Key issues • Does the model improve overall economic efficiency?
• Could the model influence end-users’ behavior (such as increasing/decreasing usage and penetration)?
• Could the model affect Online Service Providers’ behavior (for example, increasing/decreasing entry barriers, innovation and content accessibility or increasing prices excessively)?
• Does the model ensure the Internet value chain remains open and competitive?
• Does the model encourage more efficient use of available network capacity?
• How feasible is it to implement the model and what is the wider impact on key stakeholders?
Make users pay more
• No signals to online providers
• Need for differentiated user tariffs
• Application adaptations to lower bit rates
• Hard to implement in a highly competitive market
Make the Service Provider pay
• Need for a collective agreement among ISPs
• Sender pays model
• Revenue sharing
Differentiated quality services
• Need for industry-wide standards on QoS delivery
• Agreements among ISPs, billing etc.
Bilateral agreements CDNs-ISPs
• Bilateral agreements: CDNs cache content, access ISPs provide
for QoS in last mile (like VPN services)
• No need for wider consensus
• Already happening
CDN interconnection
CDN provider A
CDN provider A
CDN provider B
CDN provider B
CSP A CSP A CSP A CSP A
User agent User agent
CSP A makes only one business agreement with provider A
By interconnecting, CDN providers obtain geographic extension
• without compromising the quality of delivery
• without incurring additional transit and other network costs that would result from
serving content from geographically or topologically remote locations
CDNI
ISP
access CDN
authoritative CDN
ISPs handle third-party content:
• offer better QoE to their customer
• authoritative CDN reduces HW capacity and footprint
• ISP derives incremental revenue by “deep” cashing in access CDN
• ISP reduces traffic load on its access network, controls traffic
Also:
• handling of nomadic users
• overload handling and dimensioning
• resiliency
End 2 End CDN QoS
The evolution and future business models for CDNs
Sources
• http://drpeering.net/
• http://gigaom.com/2007/08/06/cdn-price-wars/
• On Cooperative Settlement Between Content, Transit and Eyeball Internet Service Providers
• Complexity of Internet Interconnections: Technology, Incentives and Implications for Policy
• X. Tang, J. Xu, On replica placement for QoS-aware content distribution, In IEEE INFOCOM, March 2004.
• D. Applegate, A. Archer, V. Gopalakrishnan, Optimal content placement for a large-scale VoD system, In ACM CoNEXT, Philadelphia, USA, Dec. 2010.
The world as we know it
• Layering of ISP’s
• Transit & Peering schemes
• Vanilla CDN’s
• QoS only inside same AS/ISP
Problems/Challenges
• Lack of Inter-AS QoS
• Do-it-Yourself CDN
• Embedding Properties on CDNs
• Real World Business Examples
• Reversed Value Flows
Layered ISP’s
• Lower Levels(Tiers) buy transit(connectivity) from upper or same level ISPs
• Tier1 ISPs don’t buy transit
• Full Mesh between Tier1 ISP’s
• Peering agreements between them
Transit & Peering
• Transit price $/Mbps
• Price drops constantly
• b/w almost commoditized
• Settlement-Free Peering
• Over existing links
• Rules about peering policy of ISP
Vanilla CDNs
• Overlay Networks, designed to deliver Content from the optimaloptimal location
• Points-of-Presence in many ASes & geo-locations
• Requests re-directed to closest to client CDN PoP
• Content Providers pay to host content on a CDN under promise of better/lower latencies
Vanilla CDNs (cont.)
• Network Owned
– Level3, Limelight, AT&T
• all/most servers in same ASN
• Or not
– Akamai, MirrorImage, CacheFly, PantherExpress
• ..CDNs place their servers(PoPs) directly in other ASNs
• Source: 2008, Patrick Gilmore, akamai, Peering with other Networks
CDN pricing
CDN findings
• Closest node is NOT enough • Google,2009 research.google.com/archive/imc191/imc191.pdf
• Price of Content hosting almost commoditized
• Price Wars(akamai vs. LLNW,EdgeCast,Level3) • http://www.datacenterknowledge.com/archives/2010/01/20/akamai-price-cuts-bad-news-for-cdn-rivals/
• Weight on new services
• Relations between cost & content & end-users
So the current world..
• The Transit and CDN Markets are doomed
– The days of simple Internet Transit or vanilla CDN services are numbered
– and any business model that is based on margins on something going to zero are doomed
Content Distribution Model 1: Traditional Cable
• Paramount offers an early release Pay-Per-View Movie Premiere. Assume that Paramount advertises: – an opening night pay-per-view movie for $13 on the cable system or – on a subscribers Internet service, and the end-user clicks on the movie
to purchase it
• With the traditional pay-per-view cable revenue split model, Paramount would typically receive 50% of the revenue from the movie, and the rest would go to distribution chain (the cable company), about $6.50 per video
Content Distribution Model 2: BestEffort Internet
• Studio pays ISP for Internet access and/or a CDN paying on a metered basis. The upstream ISP or CDN pays for distribution as needed along the distribution chain for best effort service
transit
transit peering
$0.40
$0.30
Content Distribution Model 2: Potential Challenges
• Despite the larger profit margin here, there are challenges: – BE Internet service for large long-lived streams under
conditions of network stress results in poor QoE • … and loss of revenue.
– Congested peering links and peering disputes
– BE transit service does not scale well for content with unpredictable demand characteristics.
– ISPs often prefer more circuitous traffic path to maintain their peering traffic ratios. As a result, the shortest path from content to a CDN is sometimes not used
• It may make sense to pay a little more money for a better end-user experience!
Content Distribution Model 3: MediaGrid Overlay Network
• Paramount is willing to pay Tata a little extra for a high-quality "Media Grid" service (“ASQ path”)
• Assume Tata offers Media Grid Service to Paramount for 75 cents per GB -> $3 in total to deliver the movie
• Two scenarios: – Cascading revenue flow (without caching)
– Revenue flow with cache
Model 3: Cascading Scenario
Model 3: Caching Scenario
• In this scenario, revenue is proportional to the cache hits; everyone along the path is compensated for the first distribution, and after that, only when their caches are hit.
– This incents big caches at the edge for popular content, and the widest dissemination of caches to increase cache hits and corresponding revenue.
Conclusions
• End-to-End QoS is possible under cooperation of all stakeholders
• As end user pays, reverse value flows unfold
– Additional revenues should compensate additional costs of all value chain actors
• ISP and CDNs morphing
– Deep caching as an emerging business