Content Delivery Networks Get Personal


  • Service providers can combine the CDN replication model with personalization
  • Adding 2 key components increases CDN intelligence and processing power
  • Manipulating content at the network edge increases scalability

CDNs make sense for personalization

Because content delivery networks (CDNs) are at the heart of next-generation IP video infrastructure, they are the logical place to add personalization. With the right approach, service providers can take advantage of the CDN replication model while delivering personalized content. At the same time, they can ensure the network scales to achieve quality expectations. There are many good reasons to personalize content, including:

  • Targeted ads: Advertisers may want to target ads at groups or individuals to increase their effectiveness. With targeted ads, end users watching the same TV show or video-on-demand receive different ads based on their profile.
  • Emergency alerts: Local, regional, or national authorities may want to use TV messages to alert the public to an impending or current emergency in their area. Messages could include weather- or disaster-related notifications, health and safety bulletins, or missing persons alerts.
  • Blackouts: Broadcasters may not want to air certain TV programs, such as major sports events, in a certain market. Content delivery may be blocked or substituted with alternate content for people living in the geographic area where the event is taking place. Managing blackout zones becomes more complex as content rights are fragmented across screens and as nomadic users get access to TV on their mobiles devices.
  • Quality adjustments: Service providers may want to dynamically increase or reduce content quality based on contextual data such as subscriber profile, device used, and content watched. Adjusting content quality is a helpful tactic when bandwidth is scarce and when several end users are sharing the same bandwidth. Consider home networks and mobile cells where the access technology typically limits bandwidth to a few megabits per second.  Service providers may want to control exactly how bandwidth is shared among the devices and give priority to premium subscribers and wide screens. Subscribers may also want to specify their preferences. 

In all of these cases, end users receive different versions of the same content, or they receive alternate content for a limited time period.  Most of the time, most of the content is common to all end users, allowing service providers to take full advantage of CDN efficiencies. To use their CDN for personalized content, service providers need a way to partially or temporarily manipulate the content at the edge of their network. This means they must augment their CDN with capabilities that allow them to enhance HTTP generally and HTTP adaptive streaming (HAS) specifically.

HAS in a CDN environment

The industry has adopted HAS for delivering multiscreen video to connected devices. HAS adapts the quality of the video stream in real time to match available bandwidth, helping service providers deliver video to more customers in a cost effective way. One advantage of HAS is its compatibility with the standard HTTP caching mechanism. Service providers are building their own CDNs to replicate content at the edge of the network. The goal of the content replication mechanism for HTTP/TCP delivery is the same as that of multicast for Real Time Streaming Protocol (RTSP)/User Datagram Protocol (UDP) delivery in the IPTV world: They are both meant to reduce network traffic by serving individual end users from the edge of the network with a unicast stream. To achieve this goal, a CDN puts replication points in the network to copy the multicast tree.

CDNs serve end users from the edge of the network

Figure 1. Like IPTV networks, CDNs serve end users from the edge of the network

Serving end users from the edge of the network allows service providers to scale the network and improve video quality for popular content. Service providers no longer need to send individual streams of the same content to every end user from a central point in the network. Instead content is sent only once across the network to a point as close as possible to the end user. However, because HTTP is a stateless protocol, the streaming server does not manage the state of the client. The client is responsible for maintaining its own state. To the server, all client requests are equal, whether they come from the same client or multiple clients. As a result, service providers have little control over what is being delivered to the client. This makes it difficult for them to scale delivery cost effectively, and impossible for them to tailor content delivery for individual users or devices.

New CDN components increase content control

To take control of the delivery mechanism, service providers must build a server-side version of the client-based adaptive streaming concept. Introducing two new components will bring more intelligence and processing power into the CDN to improve overall HTTP behavior and performance:

  1. A session manager retrieves the contextual information needed to customize the content and tell the cache in the CDN which changes to apply to the content.
  2. A video processor in the cache generates the new content to be sent to end users based on the original content and the information delivered by the session manager.

A session manager and video processor increase control over content

Figure 2. A session manager and video processor increase control over content

A session manager increases awareness A session manager tracks each session and application. It helps service providers gain greater awareness of individual content requests and video streaming sessions. This awareness allows service providers to manipulate and personalize each video stream while accounting for current network conditions, subscriber profile, device type, and content being viewed. The data is collected from a variety of sources:

  • Network conditions are retrieved from a real-time network analytics system.
  • Subscriber profile and location are retrieved from a subscriber management system based on attributes such as device IDs and IP addresses.
  • Device type is retrieved from HTTP GET requests sent by the devices.
  • Content being viewed is retrieved from a content management system.

The session manager also interfaces with third-party systems that make decisions about changes that must be applied to the content. The type of system needed depends on the type of personalization required. Examples include:

  • Ad campaign managers
  • Emergency alert centers
  • Broadcaster management systems
  • Systems that support a Policy and Charging Rules Function (PCRF)

The session manager processes the information from the third-party system and passes on specific recommendations to the cache. A video processor customizes content The video processor in the cache is a sophisticated engine with deep knowledge of the underlying video codecs and transport protocols. It uses the recommendations from the session manager to dynamically manipulate the original content for each end user or device. The content manipulation performed by the video processor can include:

  • Completely substituting the original content with other content during a limited time for a blackout or targeted ad.
  • Slightly modifying the content by adding new flows to the original stream to display alerts or support trick modes such as pause, fast forward, and fast rewind as an overlay.
  • Sending a different bit rate than requested or providing a limited list of available bit rates to control the quality delivered to end users.

Content manipulation creates new opportunities

Content manipulation techniques also open the door to new opportunities to adapt, optimize, and secure video distribution to connected devices.  Content protection The ability to manipulate content at the edge of the network allows the CDN to better secure the content. Most content protection technologies rely on encryption techniques such as scrambling to restrict content usage to authorized users. Encryption can be applied at content preparation time, but its effectiveness is much higher if it is applied separately for each session. This technique is known as session-level encryption. Inserting an invisible mark in the content before it is delivered to the end terminal is becoming a very valuable technique to protect content.  It can be used as a complement or an alternative to classic digital rights management (DRM) solutions. The premium version of the invisible marking technology can personalize the watermark per video session. This makes it possible to track the end user who retrieved a specific asset from the network. Fast channel start Unicast content delivery offers fast startup times that are comparable to IPTV with fast channel change capabilities. When an end user requests a video, the HAS client simply sends an HTTP GET request to the video server, and the server replies instantly with segments of the requested video. Video rendering begins as soon as a few video segments arrive in the HAS client buffer. Unfortunately, important delivery functions such as entitlement checks, content manipulation, on-the-fly content encryption and decryption add delay to startup times. The delays are influenced by how responsive the systems providing these functions are, and by how well the end-to-end solution is designed, integrated and engineered. The additional delays can increase startup times by up to a couple of seconds or more. End users accustomed to the fast channel change capabilities that today’s IPTV offerings provide will be disappointed by this performance. Manipulating content in the CDN reduces startup time to improve end-user quality of experience. For example, the cache can serve the first few video segments as clear content to remove the delays associated with content encryption. Encryption is a time-consuming task. Bypassing this step for delivery of the first few video segments brings the benefits of the HAS fast startup time to the end-to-end content delivery solution.

Improve your CDN edge

Service providers should treat the CDN as a key component of their core video network and not as an independent overlay. This approach exposes the CDN to intelligence held within the network. Innovations from Alcatel-Lucent enable the CDN to customize the content based on end-user preferences and context to deliver a better quality of experience. The ability to manipulate content at the edge of the network allows any number of personalization possibilities while supporting scalable delivery. The same content is distributed cost efficiently to all end users across the CDN, and all content personalization occurs in the cache, close to the end user. In contrast, solutions that personalize content in a central location dramatically increase traffic volume and storage requirements. Because they must support an individual unicast stream for each end-user request, centralized solutions are also harder to scale.

Related Material

The Future of IP Video: From Pay TV to Cloud TV ebook

To contact the author or request additional information, please send an e-mail to