Implications of Real-World Shifting Experiences through Start Over and San Diego TV-to-PC

Keith V. Nichols, Senior Director, New Product Deployments, Time Warner Cable
Kenneth Gould, Senior Director, Consumer Technology Development, Time Warner Cable
Yaron Raz, Director, Network Architectures, BigBand Networks

1.0    Introduction to Shifting

Time-shifting and device-shifting are two ways for cable companies to demonstrate to subscribers that network functionality is evolving to accommodate their busy, changing lifestyles. Time-shifting refers to the ability for viewers to have some degree of control over when they begin watching programs; device-shifting provides subscribers with the opportunity to watch television programming on personal computers, cell phones and other devices, not just immobile TV sets. Embracing these technologies allows the more innovative cable operators to build customer loyalty, differentiate themselves from their competitors, and explore new revenue opportunities.

Time Warner Cable launched its version of time-shifting, called "Start Over", in November 2005. With Start Over, digital cable customers who miss the beginning of certain shows, but who tune in before the end, can push a button and go back to the start. They also can pause and rewind, but can't fast-forward through commercials. Device-shifting capability is also being offered to some customers, in this case the ability to watch television programming on PCs inside subscribers' homes. Time Warner currently offers this service in the San Diego market.

This paper examines the implications of these experiences, and describes new opportunities and limitations identified as a result of these deployments. It begins, though, by exploring how Time Warner was able to implement time-shifting functionality in its network.

2.0    Time-Shifting: A Case Study

Time Warner Cable introduced Start Over in the Columbia, South Carolina market in 2005. The trial initially involved two primary hubs and 63 channels, and has now evolved to the entire division and with over 90 channels participating in this market. From the early stages, TWC (Time Warner Cable) recognized that a successful deployment would require the participation of the programmers, along with the actual owners of the content.  Although a sister company to many content providers, TWC has business relationships with a wide array of other owners of programming. TWC's business strategy, therefore, was two-fold:

• create a product that attracts and retains consumers:
• ensure content owners are comfortable with how time-shifting affects viewership.

Start Over successfully addresses both of these goals and has, consequently, benefited TWC by driving creation of business rules that determine how content is utilized. These rules have been jointly developed with content providers. 

For example, TWC currently distributes Start Over content from central locations, with programmers dictating what content is used.  This is accomplished using scheduling information created by the content providers.

TWC only provides Start Over to viewers when the program is "in window", meaning that we only allow Start Over to be activated during the duration of the show (for instance, for a one-hour program, Start Over is activated for 59 minutes and 59 seconds of that one hour window).

2.1    Start Over Network Architecture

The following figure depicts the logical blocks comprising the Start Over service:

Figure 1: Key components of the Time Warner Cable Start Over architecture

Analog sources, which are converted to digital feeds through real-time encoders, along with digital / satellite sources, are fed into a grooming device. The program lineup is created, and local ads are spliced as required. For broadcast video systems, shaped VBR (variable bit rate) multi-program transport streams are sent for bulk encryption, and then proceed to the QAMs for linear broadcasting. Clamping is applied to incoming video feeds, creating CBR (constant bit rate) single program transport streams over IP for each channel separately. In addition, the video streams are adjusted for VOD compliance, by re-mapping PIDs (packet identifiers) and adding metadata for on-demand storage locations in the VOD servers. These CBR SPTSs (single program transport streams) can be used both for switched broadcast systems as well as for short-term storage for the Start Over service. Using CBR streams guarantees predictable aggregation of streams at the edge QAM, without running into bandwidth overflow issues.

The Edge QAMs may be dynamically shared between broadcast video, switched broadcast, VOD and Start Over traffic, giving operators further flexibility of service mix over time, and reducing operational complexity.

The recording processor also receives information on the programs in each channel, their metadata, and the contents that are approved for Start Over. Based on these data, and the CBR channel streams, the recording processor segments the streams that are marked for Start Over into VOD assets, and distributes these assets to VOD servers.  Standard VOD content distribution techniques can be utilized for getting the assets to multiple VOD servers in different locations.

When a subscriber has selected Start Over for a program, the contents are distributed from one of the VOD servers in a similar way to other VOD assets.

2.2    Implementing Start Over

Before Start Over could be deployed, several technical issues needed to be addressed, including modeling of service group sizes, prediction of VOD storage capacity needs and stream utilization, identification of the situations that could lead to excessive stream loads, and consideration of the tools available for supporting the service. This section describes the network / traffic engineering that TWC conducted prior to launching Start Over.

2.2.1    Service Group Sizes

To help predict usage for the new service, TWC completed a number of modeling studies using data collected during focus group surveys.  The modeling took into account the effects of SDV (switched digital video) rollout that was ongoing at the time, as the company realized that synergies could be gained from the similarities in infrastructure between SDV and Start Over.

TWC spent a great deal of time analyzing service group infrastructures to ensure that service would never be denied to any SDV or Start Over customer.  Modeling revealed the need for service groups to not exceed 1,000 tuners and, in some instances, not exceed 500 tuners per service group.

This modeling proved to be quite accurate across several sites. Consequently, TWC now has the tools in place to monitor peak session rates on each service group. It is anticipated that having tools to track usage patterns and forecast areas of potential congestion will enable the company to pro-actively split groups into smaller slices and prevent denials of service. 

Currently, SDV and Start Over have dedicated Edge QAM pools. TWC anticipates, however, that the introduction of the GSRM (global session resource manager) capabilities in the near future will allow the capacity of these UEQ (universal edge QAMs) to be shared across different service types, enabling gains in usage efficiency.  This work is currently underway with vendor partners to help realize these efficiencies.

2.2.2    VOD Storage Capacity

Up to the present time, capacity for VOD storage has not been an issue.  One of the metrics applied here is average hold time and, to date, the average hold time for any content has never exceeded three hours at any time. Once a show is determined to be "out of window" by a factor of two for those subscribers starting a show at the last possible moment, TWC "reaps" the asset from the system. "Reaping" an asset means deleting it from the server once the viewing window has closed, thereby ensuring that storage capacity is used efficiently. TWC's acquisition and content management system performs this process automatically. 

Even though 150 shows, each an average of one hour long, are available for Start Over, this equates to only an additional 300 hours of storage. Capacity within the TWC networks is sufficient to absorb this requirement at the current time.  However, as high definition content becomes more sought after by subscribers, and increasingly included in the Start Over service, storage capacity will need to be closely monitored.

2.2.3    VOD Streaming

Initially TWC estimated (based on focus group testing) that the streaming rates needed to support Start Over would be approximately double the VOD streaming rates.  Remarkably, a year after the project started, this is proving to be a good estimate of the streaming needs of the VOD servers.  As progress is made in providing Start Over in additional locations, TWC's operations team is starting with the premise that introduction of the service will double peak VOD streaming rates in every location. 

Looking to the future, as rights to additional content (especially high definition programs) are secured, TWC will need to re-evaluate and refine these starting points for streaming rates.  Fortunately, with today's newer architectures on VOD servers, there is now sufficient capability to expand server complexes with either storage or streaming capacity, as appropriate.  This has been a desire of TWC's from the early stages of the project, as the company wants to ensure that the flexibility needed to adjust to changing needs is maintained.

2.3    A Win-Win Service

Everyone benefits from Start Over, including subscribers, programmers and TWC. Here are some of the reasons why:

• Convenience for consumers – Start Over provides another way to make TV watching more convenient by introducing some flexibility as to when viewers can begin watching their favorite shows. DVRs rely on subscribers anticipating which show they want to watch later and taking action to program their device to make sure that those recordings happen. By contrast, Start Over doesn't require any forward planning by viewers and is, therefore, available for them at the press of a button.  TWC has also made the product very easy to find and use: a small icon, shown on the screen shot in figure 2, is all that is used to indicate that this is a Start Over enabled channel.  By simply pressing the select key, a subscriber activates a Start Over stream.
  
  Figure 2: Start Over makes only a small change to the picture seen by subscribers
  
  
• More eyeballs for programmers and advertisers – a service such as Start Over increases the number of programs that are viewed all the way through. Not only is more of a given program watched, but the number of commercials being viewed increases too.

Customer studies reveal that Start Over users are more satisfied and loyal to Time Warner Cable.  More than half of subscribers polled agree that Start Over makes watching television "more convenient", as figure 3 reveals.

Figure 3: Start Over is increasing subscriber satisfaction and loyalty to Time Warner Cable

Custormer research also revealed that Start Over channels have become preferable to traditional broadcast channels for increasing numbers of subscribers.

Figure 4: One-quarter of subscribers prefer to watch Start Over channels

In addition to its independent market research, TWC has also performed a customer study with one of the company's close partners on the project.  The key findings are:

• over 70% of digital households are using Start Over every month, and up to 7 times per month;
• 41% of Start Over users say they are more likely to retain digital cable;
• Start Over is driving additional outlets since increasing numbers of subscribers want the service in additional rooms of their homes.

Interestingly, research has revealed that customer see Start Over as complementary to TWC's on-demand service offering, and even to DVR usage. Growth of these services has continued unaffected to any significant degree by the roll out of Start Over.

Figure 5: Start Over provides benefits that do not cannibalize DVRs

Start Over has created a user environment that subscribers do not wish to relinquish. TWC believes that the service is causing a fundamental paradigm shift that differentiates the company, and cannot be matched by competitors.

Additionally, as Start Over matures, content providers will become increasingly comfortable with TWC's ability to protect content and use it advantageously. All Start Over schedules are rights-based with the partner providers, so that content provided for Start Over viewing is mutually agreed upon by TWC and its partners.

2.3    Lessons Learned from Time-Shifting

By the end of 2006, TWC will have seven markets in play with Start Over capability.  Within a year of its launch, the service is achieving over 100,000 "hits" per week, and recently reached the milestone of 20,000 "hits" in one night. One in four subscribers in the Columbia, SC market is using the service on a consistent basis.

Start Over is an excellent beginning for TWC to help realize the dream of "everything on demand".  TWC has learned that:

• the parameters are in place to expand the viewing windows;
• the parameters are available to offer more trick mode capabilities;
• we have the technology to create large archives of content;
• we have the ability to reduce the service groups sizing to provide more streams to the edge;
• our plant can be expanded to offer more time-shifted streams. 

The company plans to expand the number of markets significantly in 2007 and 2008 and our partners are committed to work with us on this effort.  The time spent building the partner relationships in place today are an investment toward expanded Start Over in the future.

The practical limitations are not technology anymore - the limitations are rights to the content that is delivered to consumers.  The real challenge is to show our partners in the content development community that their content is important and that it is preserved for true viewing and not copied.

3.0 Device Shifting in San Diego

While the majority of viewers are content to watch television on their TV sets, habits are changing. No longer is the PC seen as a "lean-forward" device, as increasing numbers of subscribers are choosing to access video content over the Internet. The recent acquisition of YouTube by Google testifies to an unmistakable trend – increasing numbers of subscribers want to watch video programming on a broader range of devices than simply television sets. Time Warner Cable's TV-to-PC service in San Diego is the first venture by the company towards offering broadcast television programming to PCs. In this section of the paper we explore some of the reasons that prompted the company to develop this service, and describe how it is being embraced by subscribers.

3.1    TV-to-PC Deployment Overview

In July of 2005, TWC launched a BBTV (Broadband Television) trial which delivers approximately 80 broadcast channels to subscribers over a HSD (high speed data) network using DOCSIS.  The goals of the trial included exploring:

• technical implementation issues;
• policy topics; and
• marketing strategies.

In order to receive the BBTV service, the customers had to be subscribed to one of the basic or expanded basic video tiers and to TWC's Roadrunner HSD service. This trial was accessible to customers residing in an area served by several hubs in TWC's San Diego division, at the time this paper was written in October, 2006.

The 80 channels of broadcast video are identical to those found on the San Diego division's basic or extended basic video tiers.  The channels on the BBTV service are on parity to those that customers can access on their television sets, meaning that, if they subscribe to the basic video tier, they cannot access the extended basic video tier on their PCs.

Figure 6 shows the network architecture used to support the BBTV trial.

Figure 6: Network topology engineered to support TWC's TV-to PC trial in San Diego

The RTEs (real-time encoders) are approximately 40 general purpose PCs with high-end encoder cards.  Each RTE is capable of encoding two video channels simultaneously.  A subset of the channels consists of programming containing secondary audio feeds.  For this subset, a single encoder is typically used to encode the video and both audio feeds.  The RTEs also run the "producer" program which encrypts the video and synchronizes the digital rights management keys with the Storefront server. The video client on the PC must receive an encrypted license from the Storefront that allows that unique client to first decrypt the license and then use the license to decrypt the video. Different licenses are used to protect the individual channels in the different video tiers. The license is unique per video client and cannot be successfully moved from one PC to another.

The video codec, video servers, and video player were provided by Real Networks, and the video/audio data rates for a single channel are below 1Mbps.

The user interface used to navigate between channels and to control the video playback is a series of web pages served to the users' browsers from the BBTV Storefront.  The Storefront verifies that the user is entitled to access the BBTV service (via the BBTV Authentication Server), determines to which video tier the user is subscribed, and provides an encrypted license to the RealVideo player that is embedded within the webpage (see Figure 7 below).  The BBTV Authentication Server works in conjunction with several cable-modem provisioning servers to ensure that the user is accessing the video service through a cable modem associated with the subscriber account that is also associated with an appropriate video service tier.

Each video channel is requested via a unique web URL.  In fact, the URL for each channel is specific to an individual instance of the video client running on a PC.  A user cannot "skip" the authentication process by sharing one of the video URLs with another person.  The Real Helix Video Servers receive the requests for video from the video client and streams the video to the appropriate PC.  If a particular channel proves to be extremely popular, multiple Video Servers are able to serve the same video channel to multiple video clients. The video is only delivered from an RTE to the appropriate video server if a user has requested the channel.  When a channel is unwatched, no video traffic is transported from the RTE to the video server.  Usage statistics are pushed from the video servers to the reporting server for correlation.

One driving goal of the project was to take the video content accessible via the TV and demonstrate that the same service could be replicated and delivered over an IP network (in this case, DOCSIS) to a PC. The BBTV service really is a replication of the analog TV viewing experience – the EAS messaging, closed captioning, secondary audio controls, parental controls and channel navigation found on the TV were replicated onto the PC platform.  The service does allow for full-screen mode, mute, and volume control, but does not allow PVR-like capabilities such a live pause, rewind, fast-forward, etc., nor does it support "Start Over" capabilities.

Figure 7: Example of programming viewed on a PC by a subscriber in TWC's BBTV trial

One goal of the project was to determine if the PC could viably be considered as just another video outlet in the home.  As the goal was to demonstrate device-shifting, and not place-shifting, checks were put in place to ensure that the user is accessing the service from behind the cable modem associated with their HSD account.  The customer could not provide access to the service to someone served by a DSL modem, nor behind a cable modem associated with an account that did not also subscribe to one of the appropriate video tiers.  While we will not go into all of the content protection details here, we can also state that the content is encrypted, DRM (digital rights management) protected, and that the licenses handed out to the PC to decrypt the video stream could not be used on another PC, nor could the license be used by the PC were it to be moved to another cable modem.

3.2    Lessons Learned from Device-Shifting

Why demonstrate device-shifting?   Better yet, why not put an RF tuner into a subscriber's PC and continue to deliver analog video to every device in the household?  The short answer is that while we continue to launch interesting and customer-rewarding products over the "classic" MPEG video plant (as described in the other sections of this paper), TWC wanted to identify potential pitfalls and triumphs that could come from delivery of video over a high-speed IP network.  Here are some of the insights gained from the BBTV trial:

• TWC spent quite a bit of time determining the least intrusive methods for authenticating the user, verifying their physical location, and authorizing the subscriber's video client to allow the display of the appropriate video tier.  Obviously this required a high degree of integration with the HSD provisioning systems and with the cable billing system.  All of these checks can be performed quickly enough that it does not adversely affect the consumer experience.
• While the co-ordination of the web-based navigation and video display client with the DRM-specific parts of the backend system are complex, to the end user it is seamless and non-intrusive.
• There are fundamental features of the traditional cable video service that have existed for decades that can be quite complicated to replicate in a PC video environment.  For example, TWC developed a system for delivering closed captioning data to the PC outside of the video stream, yet still keeping that data synchronized with the audio.  This was largely due to the fact that for the trial we elected to use a video codec that is not capable of accommodating the closed captioning data within the video stream.
• TWC chose to encode and encrypt our video after the EAS (emergency alert system) messages were overlaid on the video and after video advertisements were inserted.  This was acceptable for a field trial as the limited geographic area served by the trial was contained within a single ad zone and EAS zone.
• TWC constantly monitors the DOCSIS CMTS utilization to project when additional capacity will be required.  As with the introduction of cable telephony services, the mechanisms that trigger activities to add more capacity were also used for the BBTV trial.  We should point out that for our trial purposes, unicast video streams served each video customer.  We developed the ability to multicast the streams, but elected not to implement it for the BBTV trial and instead considered the unicast traffic to model the "worst case" traffic pattern.
• Usage reporting systems can be made very robust and will keep the marketing and product development teams busy forever.

By the time that this paper is published, TWC will have likely completed the San Diego BBTV trial.  While the particular implementation used for the trial will likely not be replicated elsewhere within TWC's footprint in the near term, the information gleaned from this trial is worth many times more than the cost of its implementation.

4.0 Looking Forward

One of the future capabilities that TWC wants to explore is an expansion in the role of programming guides.  Instead of a world of looking at a grid guide from the current time and to the right, the company is creating the potential to use a grid guide that looks both forward in time (the right) as well as backwards in time (to the left).  As our relationships mature with programmers and the uses of Start Over, our look back windows can increase and provide more of an "everything on demand" experience. 

The technology needed to achieve this is available, and includes:

• RTA (real-time acquisition) capabilities for hundreds of channels; 
• capability to expand the storage for increased video content due to RTA expansion; 
• the schedule mechanisms for selectable provisioning of content to be available for Start Over; 
• distribution networks for providing increasing numbers of sessions on the cable plant; and 
• the Digital Navigator technology to facilitate both the traditional looking forward in time for programming, as well as the looking back in time for programming that could be potentially available for viewing.

Additionally, once time-shifting functionality is in place, other synergistic technologies follow. Time-shifting introduces opportunities for a new advertising paradigm catering to both programmers wanting their content viewed and advertisers wanting their content being fresh. Remember that some advertising is only good for one day, and may be considered stale after the day of viewership. Time-shifting can be leveraged to insert "fresh" advertising into stored content on subsequent viewings as the "dynamic play list" concept with VOD servers matures. Although not yet ready for prime-time, this capability will become viable as usage of TWC's time-shifting service increases.

Real-time insertion of ads into IP video streams being delivered over DOCSIS to non-traditional video devices, such as PCs, will probably be a worthwhile goal for future device-shifting trials. Although doing so will be a challenge initially, there are significant benefits to doing so:

• cable companies can increase ad revenues by catering to marketers increasingly interested in using the Internet as an advertising vehicle;
• the concept of "targeted advertising" can be utilized to deliver to subscribers ads that more closely match their interests, resulting in higher response rates and leading to marketers willing to pay more to leverage this type of advertising; 
• the additional revenues obtained from inserting ads into BBTV streams can help fund broader device-shifting deployments.

Device shifting also requires that the conditional access systems used on traditional cable infrastructures are replicated to protect content delivered over IP networks,  and augmented with digital rights management to a degree that satisfies content owners and is transparent to subscribers.

5.0    Outlook for Shifting

By embracing the concept of network-based shifting, Time Warner Cable is retaining greater control of content, while providing subscribers with DVR-like capabilities. This bold step benefits the company in several ways such as the protection of advertising revenues and increased customer loyalty. Subscribers also benefit, however, because TWC must innovate to stay ahead of advances in CPE (customer premises equipment): services like Start Over and Broadband TV will evolve to offer expanded shifting capabilities.

Experience shows that network-based functionalities are received positively by consumers, due to the robustness and ease-of-use of such services. Some examples are:

• VPN services supported on platforms in a service provider's networks replacing VPNs run on enterprise equipment;
• residential voice mail systems managed by service providers becoming preferable to telephone answering machines in the home;
• the growing popularity of on-demand movies versus rentals from brick-and-mortar video chains;
• increasing amounts of music and video being downloaded via paid services, at the expense of peer-to-peer swapping.

Start Over is not, however, simply another example of consumers reacting favorably to a network-based service, but an opportunity for broader change.  The authors appreciate that the program provider community is concerned about video content in Napster-like environments. At the same time, the ability of video databases such as YouTube, Bolt and Grouper to attract millions of "eye-balls" represent a business opportunity that cable companies must contemplate. By continuing to work closely with its partners, TWC is ensuring that they are increasingly happy to embrace rights-based time-shifting and device-shifting initiatives. Slowly TWC is creating not just an "everything on demand" paradigm, but moving closer to the holy grail of making "television more like the Internet, and the Internet more like television":

• Start Over offers subscribers a live television experience but with some flexibility as to when they begin watching programs. An on-demand model is one of the attributes that attracts viewers to Internet video databases.
• BBTV provides subscribers with the ability to view broadcast-quality television programming on their PCs, offering them the added convenience of, say, watching their favorite TV shows while checking emails.

Future investments by cable operators in their DOCSIS infrastructures, and by computer manufacturers in PC technology, will enable the quality of the picture and audio experienced by viewers to evolve to the point where broadband TV viewed on a PC will be equivalent to what traditional TV sets offer.

Of course, many technical challenges remain, and TWC will need to find innovative solutions if the full potential of time-shifting and device-shifting is to be realized. These challenges include:

• storage formats – when storing contents for time-shifting, the video is stored in a single format that is compatible with the STB at home. However, when using different devices, different coding techniques, resolutions, clamping bit-rate and metadata may be required. As a result, any time-shifted material may need to be stored multiple times for usage with different devices, while maintaining meta-data synchronization between the different formats. Alternatively, transcoding and resolution conversions could be implemented on-the-fly when the time-shifted contents are actually being watched/consumed. However, the resources needed in this case are proportional to the number of active subscriber sessions, which may introduce scalability limitations and higher costs.
• encryption techniques – usage of different devices may impose usage of different DRM and encryption techniques. This, in turn, requires placing the encryption mechanisms after storing the contents, rather than before it as mostly done today.

Confident that these issues will find creative resolution, the authors anticipate many future possibilities that will benefit the end consumer as well as the business stakeholders.  The future for shifting is bright.

Abbreviations and Acronyms

BBTV	Broadband Television
CBR	Constant Bit Rate
CMTS	Cable Modem Termination System
CPE	Customer Premises Equipment
DOCSIS	Data over Cable Service Interface Specifications
DRM	Digital Rights Management
DSL	Digital Subscriber Line
DVR	Digital Video Recorder
EAS	Emergency Alert System
GSRM	Global Session Resource Manager
HSD	High-Speed Data
IP	Internet Protocol
MPEG	Moving Picture Experts Group
MPTS	Multiprogram Transport Stream
PC	Personal Computer
PID	Packet Identifier
QAM	Quadrature Amplitude Modulation
RF	Radio Frequency
RTA	Real Time Acquisition
RTE	Real Time Encoder
SDV	Switched Digital Video
SPTS	Single Program Transport Stream
STB	Set-Top Box
UEQ	Universal Edge QAM
VPN	Virtual Private Network
URL	Uniform Resource Locator
VBR	Variable Bit Rate
VOD	Video On Demand

Click here to download PDF