Project topics and materials
Get Complete Project Material File(s) Now! »
Overview of the Content provisioning value chain
The value chain of content provisioning is, in a simplified view, composed of three main families of players: content producers and providers (professionals, e.g. BBC [43], YouTube [127], Netflix [93]…), content distribution service providers (Akamai [32], Level3 [80]…) and network service providers. A CP owns a proprietary infrastructure that, in addition to content and metadata management (including popularity estimation) and storage, enables customer related functionality like authentication, authorization, accounting, billing, profiling, content adaptation. Content popularity assessment and users profiling facilitates an optimized usage of resources. CPs also may offer some value-added services to end users such as portals personalization per end user and provision of intelligent tools for browsing the content catalog.
As the volume of contents and requests exponentially grows, CPs should find a tradeoff between investing in their infrastructure thus increasing their costs, and delegating content delivery to third parties. Since the datacenters of most CPs are centralized, the long distance to end users often induces a degradated QoE due to high routing delays and eventual traffic jams on the internet. Scalable and efficient content delivery requires CDNs. CDNs consist in clusters of servers distributed in one or many geographic locations or PoPs and connected to ISPs networks through Internet exchange points [101] [100]. There is a moving equilibrium between CPs and CDN providers. In fact, CPs usually delegate their content delivery (at least for the most popular content) to CDN providers but this is not always the case. YouTube and Netflix are well-known exceptions [4] [8]. The CDNs deployed by CPs are usually destined for proprietary use and are not open to third parties. Some CPs may rely on more than one CDN in order to ensure a better quality of experience and master costs (before implementing its own CDN, Netflix had agreements with three CDNs [33]). However, CPs may lack good control solutions to facilitate the management of the overall (composed) system.
CDNs, the systems on which content distribution rely, can be classified in two main families: those that use the Internet for the connectivity requirements, denoted by pure-play CDNs, and those based on end-to-end managed networks, denoted by Telco CDNs. A CDN performs two essential functions. First, it caches content at the edge of the network, closer to end users, to reduce the IP traffic traversing the core network and, ideally, deliver a higher quality experience to viewers. Second, a CDN positions multiservice, multiprotocol content delivery capabilities at the network edge, allowing a dynamic delivery of non-linear and linear contents to a multitude of devices and end users [101] [100]
Overview of the CDN Market
Since first commercial CDNs were launched in 1996 [58], the CDN market has witnessed an exponential growth worldwide [58]. The evolution, since 2010, of the CDN market is shown in Figure 2.6. As can be noticed, CDN providers are mostly present in North America, Western Europe and the Asia Pacific region. Today, the CDN market handles around 35 % of the global internet traffic [19] [23]. This share is expected to increase in the upcoming years. In particular, more than 55 % of the worldwide traffic is expected to be delivered by CDNs in 2017 [19].
The CDN market is composed of two main categories of players: those that use the internet for their connectivity requirements, denoted by pure-play CDNs, and those relying on end to end managed networks denoted by Telco-CDNs. In its early days, the CDN market was limited to a reduced set of pure-play CDNs with a high capacity and a global reach or footprint. These include Akamai [32], Limelight Networks [82] and CDNetworks [50]. Akamai, for instance, owns more than 127 000 servers distributed in 1150 clusters and in 87 countries worldwide [32] [51]. Even though these players still own a large share of the CDN market (Akamai alone owns today 80 % of the CDN market share [19]), recent years have witnessed the rise of specialist and regionally-focused CDNs joining the ranks of pure-play heavyweights. For instance, BitGravity (founded in 2006, owned by and integrated with Tata Communications) is a CDN provider that is specialized in the delivery of on-demand video and High Definition (HD) live streaming over India and Australia [46]. In order to cope with evolving users’ trends, giant pure-play CDNs like Akamai enhanced their basic functionalities (listed in section 2.2.2) for supporting the delivery of mutiple types of contents (video, live, web pages…) to a multitude of devices (fixed and mobile users) [58]. Value-added and customized services are also being proposed by many pure-play CDNs to their customers among Content Providers and businesses. These include online security, bitrate adaptation and portals personalization. On the other hand, the CDN market has also witnessed the rise of a new category of players: the Telco-CDNs. The number of Telco-CDNs is growing through years and has reached 120 players in 2012 [58]. The majority of Telecom operators dedicate today their CDNs for their internal usage [2] [53]. However, more Telco-CDNs include the ability to manage over-the-top (OTT) Internet traffic. The monetization of the CDN service can be through a Business to Client (B2C) model (Retail CDNs [53]) or a Business to Business (B2B) model (wholesale CDNs [53]). In the B2B model, a Telco-CDN can propose a last-mile content delivery service to CPs and to pure-play CDNs. Telco-CDNs attractiveness in this context is limited to local and regional CPs. AT & T is an example of a Telco-CDN that has partnered, through a B2B model, with local CPs in the US [2]. Given their leadership position in the market, pure-play CDNs has not found so far enough incentives to delegate their content delivery to Telco-CDNs. Nevertheless, this trend is evolving as will be observed in the next section.
Toward a closer and lower content distribution
One way of dealing with users’ higher QoE constraints consists of placing content closer to them. Content providers and pure-play CDNs are already aware of the impact of distance on the delivery of sensitive contents like videos. This awareness is expected to widen in the coming years [85].
In fact, many OTTs including CPs and pure-play CDNs are bypassing the internet backbone through directly connecting their servers to Tier3 ISP networks. YouTube [127] and Netflix [93] are part of this trend and the recent agreement between Netflix and Comcast in the US goes in this direction [6]. Similarly, Akamai is today within a distance of one autonomous system of 90 % of worldwide internet users [64]. It is important to mention that OTTs like Akamai pay local ISPs for providing them with a direct paid peering connection [64], others like Netflix claim that they should not pay.
Table of contents :
1 Introduction
1.1 Context and Background
1.2 Collaboration of CDNs
1.3 Thesis Objectives
1.4 Contributions
1.5 Overview of the dissertation
2 State of the art of Content Distribution Services
2.1 Overview of the Content provisioning value chain
2.2 Overview of Content Delivery Networks (CDN)
2.2.1 Overview of CDN Topology
2.2.2 Overview of CDN Operation
2.2.3 Overview of the CDN Market
2.2.4 Overview of CDN pricing schemes
2.3 Evolution Trends
2.3.1 Transformation of the Internet Edge
2.3.2 Network transformation
2.4 Impact on CDN Stakeholders
2.4.1 Toward a closer and lower content distribution
2.4.2 Content Delivery Networks re-defined
2.4.3 Collaboration of CDNs
2.4.4 Joint CDN-Network Operation
2.5 Summary: Motivating our contributions
3 CDN Federation: Control Architecture
3.1 Federation Computation
3.2 Federation Provisioning
3.3 Federation Dynamic Control
4 CDN Federation: Static Decision-making
4.1 Variables
4.2 Objective Function: the point of view of CDNs
4.2.1 Revenue Function
4.2.2 Cost Function
4.2.3 Objective Function
4.3 Service Constraints
4.4 Capacity Constraints
4.5 Revenue Model
4.6 Rational Fairness
4.6.1 Separate Revenue
4.7 Computing Optimization and Outputs
4.7.1 Established CDN Federations
4.7.2 Content Distribution Strategy
4.7.3 Requests Routing Strategy
4.8 Federation Re-provisioning
4.9 Model Performance and Scalability
5 Use Cases of Federation
5.1 Traffic Dataset
5.2 Use Case 1: Federation of Local Telco-CDNs
5.2.1 Use Case Presentation and Inputs
5.2.2 Evaluation Methodology and Outputs
5.2.3 Discussion
5.3 Use Case 2: Federation of pure-play CDNs
5.3.1 Use case Presentation and Inputs
5.3.2 Evaluation Methodology and Outputs
5.3.3 Discussion
5.4 Use Case 3: Federation of pure-play and Telco-CDNs
5.4.1 Use case Presentation and Inputs
5.4.2 Evaluation Methodology and Outputs
5.4.3 Discussion
6 CDN Federation: Dynamic Decision-making
6.1 Request Routing
6.2 Control of Peak Events
6.2.1 Context and Motivation
6.2.2 Framework 1: Adaptive Load Balancing-based approach
6.2.3 Framework 2: Resolution Adaptation-based approach
6.2.4 Framework 3: Mixed approach
6.2.5 Evaluation and discussion
7 Telco positioning in the CDN ecosystem
7.1 Overview of the Telco assets
7.2 Overview of potential Telco Roles
7.2.1 CDN provider
7.2.2 Federation Controller
7.2.3 Added-value services provider
7.3 Overview of the Telco Control Plane
7.4 The Evolution of the Telco Control Plane: A Roadmap
7.5 Enabling Telco-based added-value Services: a Use Case
8 Conclusion and future work
8.1 Conclusion
8.2 Future Work
Bibliography
