A unified approach to on-demand real-time media delivery

January 23, 2009

On-demand real-time video delivery

Network-based on-demand service must be available from multiple source points, and those points must be able to change as demand shifts around the network.

By Deepak Kataria
Director of Systems Solutions
Networking and Telecom Practice, HCL America

The digital content market is growing rapidly. Market research reports indicate digital video content is expected to grow to $130 billion worldwide by 2011. This includes on-demand digital video streaming, television services such as Video on Demand (VoD), Subscription Video on Demand (SVoD) and Pay-Per-View (PPV). Also included in this market is on-line video advertising, Internet Protocol Television (IPTV) and mobile TV.

Growing demand for media-rich, just-in-time content and an expanding universe of mobile devices capable of real-time viewing are placing great demands on networks. The inexorable network migration from a centralized, core-based architecture that assumes low-capability access devices, to an edge-based, distributed network architecture that is accessed by intelligent devices is also changing the underlying networking paradigm. Also, shifting is the manner in which content is delivered and consumed. Traditionally, broadcasters determined unilaterally which content they would distribute and when. And while there remains demand for this distribution modality, it is rapidly falling out of favor in large swaths of the viewing population because of growing demand for on-demand, just-in-time delivery.

However, the challenge of delivering the right content in front of the right person at precisely the right time, using whatever device and access modality the customer wants to use to view the desired content is placing unprecedented challenges on the network delivery infrastructure. In other words, real-time, unicast content needs to be delivered on-demand to any of the three screens, be it a TV, a PC or a Mobile appliance as the customer desires at the highest Quality of Experience. The content could range from short clips and TV episodes to full movies and can be professional grade, conversational or user generated content popularized by the growth of social networking. The content has a popularity quotient attached to it, and has spatial and temporal distribution in its request. The challenge that network designers face is how to build a highly scalable, efficient and robust real-time content distribution network.

The forces shaping the on demand video and real-time media delivery market are as varied as the players in the market. The market comprises traditional telco operators, cable providers, wireless players, ISPs, satellite companies, and content delivery networks (CDNs) to name a few. To one degree or another they are all affected by the evolution that is underway today. Telco's are marshalling their resources to offer fiber-deep networks capable of delivering all forms of content with equal quality to all subscribers across a single converged optical infrastructure. Verizon FiOS and AT&T U-Verse are prime examples of this evolution. They recognize that customers want access to content and their response is to ensure that the network they operate in the near-term and beyond will be up to the task. In the cable side of the industry, Data Over Cable Service Interface (DOCSIS) 3.0 is the most visible, high impact force at work.

As part of the move from analog to digital, cable channels are shifted, freeing up channels that can then be bonded to provide faster Internet access -- currently as much as 100 Mbps, soon to be 160 Mbps and beyond. The move also allows cable operators to provide IPTV over DOCSIS-based networks, thus accelerating the move to IP across the board. For bandwidth efficiency, DOCSIS 3.0 allows operators to isolate and dedicate a downstream video feed to any and all users who want to watch it at the same time, simulating the way linear TV works and making the network significantly more efficient. ISPs benefit from this transition as well.

Equally important is the continued use of switched digital video (SDV) as a delivery option. SDV has the ability to free up as many as 50% of the available bandwidth by sending only the channels that are actually being watched at that point in time, thus supporting a "content-on-demand" model.

On the wireless side, the evolution to both mobile IP and broadband are beginning to create excitement. Fourth-generation wireless (4G) is, by definition, IP-based, bringing the advantages of IP all the way out to the mobile device -- its ultimate destination. And the move from Code Division Multiple Access (CDMA) and Global System for Mobile communications (GSM) as perceived end state technologies toward the real end states Universal Mobile Telecommunications System (UMTS), High-Speed Packet Access (HSPA), Long term evolution (LTE) will deliver on the promise of these emerging technologies. HSPA offers the equivalent of DSL, with 3 Mbps of downstream bandwidth and half that upstream, while LTE promises much more -- 300 Mbps of download speed, enough to deliver any service available today to a mobile user, including content-on-demand, high-def video, and other media-rich, bandwidth-intensive services.

The satellite companies have not been sitting idly by, either. Today they offer far and away the largest number of high-definition television channels, and will continue to for some time to come. The satellite companies are facing the challenges of distance and technology demonstrated by the distance-related delay inherent in satellite propagation, which means that interactive services will always suffer from latency. By entering into terrestrial content hosting agreements with Content Delivery Networks (CDN), this issue is somewhat circumvented, however, challenges remain nonetheless.

CDNs are a collection of networked servers that use the Internet as their transport backbone and co-operate to deliver hosted content to subscribers from whatever source makes the most sense for each download transaction. The advantages that CDNs offer to the customer are improvements in performance, scalability, and cost efficiency. Figure 1 shows a unified view of a video distribution network.

View full size

Figure 1: Unified View of Video Distribution Network

There are a number of commercially viable CDNs in the market today including Akamai, Limelight, CDNetworks and Level3. Their existence -- and success -- are directly related to the two factors discussed earlier: First, growing demand for media-rich content delivered to mobile devices. Second, the migration of networks from centralized to edge-based and peer-to-peer (P2P). Supported by the success of such early entrants as BitTorrent, LimeWire, Napster and KaZaA (which ultimately morphed into the highly successful Skype VoIP service), peer-to-peer went from the shadowy side of legitimacy to full-blown success, culminating in contracts between P2P providers (BitTorrent) and major Hollywood movie providers (Warner Brothers) for the legal distribution of content across their networks.

Next: Network architecture

The challenge is to promote a network architecture which makes the transition to on-demand real-time distribution not only possible but also profitable in the context of a new business model where the owner of the content is as much a driver as the consumer. Ultimately, consumers of media have three requirements in terms of content:

Further, the network must be able to adapt to customer requirements implying that network-based service must be available from multiple source points, and those points must be able to change as demand shifts around the network.

View full size
Figure 2: Rich media content delivery infrastructure

The content may originate at a variety of sources. These include standalone Internet content providers such as YouTube, AOL, Yahoo!, or online video providers -- Blockbuster.com or NetFlix. These sources own the rights to the content but have no distribution network of their own. Alternatively, the content may come from a head-end, super headend source or video-on-demand provider such as a traditional broadcaster (BBC, NBC, Fox), telco operators (BT, Vodafone, Verizon) or cable provider (Comcast, COX, CableOne). They have content ownership or access rights as well as a distribution network. Finally, the content may originate at a headend content center or digital content data center such as Akamai or Limelight; organizations that own a network but have no ownership position or use rights to content.

Content Data Centers serving as edge-based hosting facilities and using the Internet as a pipe into the mainstream network are another class of players. They often form relationships with CDNs to serve as an alternate source of in-demand content. Similarly, a video subscriber office performs much the same task but does so at the edge of the network and is often used to host both traditional video-on-demand content as well as commonly requested and local content.

Traditional service providers have a strong desire, for good business reasons, to offer premium content over their networks. This desire, however, must be tempered by a clear understanding of the processes, risks, and costs involved in creating, producing, distributing and acquiring the rights to intellectual property (IP). The main goal of the content owner, naturally, is to maximize the profitability of the content they own, which translates into careful control of release windows across the various market segments they sell into. Most movie properties leave the big screen and go to airplanes (captive audience), hotel pay-per-view (bored captive audience), DVD/tape, first-tier cable, second-tier cable, and finally to alternate delivery outlets (telephone companies). They are then delivered according to exhaustively studied market segments: continents, countries, regions, followed by commercial and residential distribution. This supply chain relationship is very linear, highly controlled and until recently, a model that worked well.

Next: Telcos vs. content owners: a confusion of aims

Telcos vs. content owners: a confusion of aims
Telephone companies want to offer premium pay-per-view as an added service to their services lineup. Traditional content owners, however, see pay-per-view as a powerful cannibalization force, destroying what they believe to be their primary revenue generator: DVDs. To them, pay-per-view removes the potential for a profitable sale (the disc), when in fact it identifies a distribution modality that is in growing demand: content availability on the viewer's terms rather than on those of the provider. Ironically, even an illegal download is sending a powerful message to anyone willing to listen: If you won't give me the content when I want to see it, I'll get it myself. And while no one is fond of piracy, it tells the market that the product being pirated is in demand. Equally important is the issue of licensing control: Those who control licensing do NOT want to lose their role in the overall value chain.

Much of the change the media industry is finding itself forced to deal with revolves around the evolution from a centralized to a peer-to-peer distribution model. This is a wrenching yet unavoidable change, and it carries with it the need on the part of the service provider to view itself in a very different way and take steps to evolve in lockstep with the evolution of the network. Today, telcos preach a philosophy of scarcity: broadband is expensive to deploy, customers aren't always as frugal as they could be in terms of their use of it, so expensive, scarce bandwidth is the order of the day. Meanwhile, companies like Google come into the picture, preaching a philosophy of abundance: Help yourselves! It's free, eat all you want! If you run out, we'll get you more!

The need for media knowledge
In reality the telcos need to bring in more media-savvy resources -- people who understand the media industry and all of the many ways in which it differs from traditional telecom. They must become more media-aware by forging alliances with content owners and rethinking their network architectures to place more resources at the edge rather than deep in the core. They must learn to appreciate the business drivers behind digital content, recognizing, for example, that the entire distribution chain must be both logically and physically secured and that the proposed audience must be large enough to secure a sufficiently broad level of exposure for the property if network-distributed media is to be profitable. As third party distributors, they bring a higher degree of risk to the equation, one consequence of which may be that content owners will demand pre-paid sales up front as an offset to the potential for risk associated with an as-yet unproven relationship. Furthermore, distribution quality must be iron-clad.

Of course, the telcos are not alone in their confusion: content owners do not understand the business of network-based distribution, nor have they quite realized that the model they use today for distribution of digital content is not viable in the long-term.

Consider this: content owners have huge content libraries that comprise two tiers of material:

Box office receipts and rental revenues can be tracked easily for the first tier of content, but because the second, more obscure tier is harder to track, the ability to quantify the degree to which it can be monetized is fleeting.

The irony is that the second tier of content, often referred to as the long tail, is enormously valuable and potentially highly lucrative. In fact, most industry analysts believe that if handled properly, the long tail of older content could generate revenue at a level that would eclipse the returns seen for first tier properties. How to make this happen?

The answer lies in aggregation. If the telecom and media industries can come up with a way to aggregate both content and demand for that content and put the two together, the so-called "long tail" -- that vast store of un-monetized, older content -- can be transformed into a money-maker. This is because products that have what is perceived to be low demand can, when taken in aggregate, generate revenue streams that rival or even exceed best-selling properties, provided the distribution channel is large enough. Consider the success of such services as NetFlix and iTunes. Studies have shown repeatedly that niche content is in higher demand than premium content in terms of revenue. So, two issues are at play here. How do suppliers make the market aware of the available content and how do they get that content to the right people within a particular market? This creation of a pull strategy for on-demand media delivery is of paramount importance in the marketplace today and must receive the full force of the industry's considerable resources to make it successful.

Future of TV
So what does this mean in the final analysis? It means that the future of television (media consumption) is not in a single subscriber base of millions, but rather within thousands of niche bases of varying size and value, that are aggregated to yield high revenue volumes. At issue is not what can be distributed, or how, but is the ability to get the right content in front of the right person at precisely the right time, using whatever device and access method the customer wants to use to view the desired content. Unfortunately, today's model does not have a standard methodology for browsing, searching, discovering content, receiving content recommendations, or subscribing to content feeds. What is needed are the following:

Of course, all of this comes at a cost, both monetary and otherwise. None of this happens without analysis of usage patterns, content matching to a user profile, and enhanced user awareness. Furthermore, some models may emerge in which the content is given away free of charge in exchange for revenue derived from video-rich, interactive, personalized, context-based advertising -- clearly a new model for the industry. And the key benefit? For customers, awareness of and access to a breadth and richness of content that they never knew existed. For content owners, an abundance of new revenue streams. And for network providers, enhanced long-term relevance, and elimination of concerns about net neutrality.

But there is another factor that must be considered, and that is demographics, in particular the latest addition to the workforce. A group of people known as the Millennial Generation. Defined as those people born roughly between 1982 and 2004, they, like all generations, share a common set of circumstances. In the case of the Millennials, however, those characteristics make them particularly interesting to us:

Just how important are these social networking environments? Put it this way: According to a recently published report by Citigroup, YouTube alone could generate nearly $500 million per year, simply by placing a non-interactive banner ad on each of its pages. Furthermore, from the perspective of a market analyst, these sites represent the ultimate gold mine. They reflect a tremendous amount of information about the people who post content, but they do something even more important. They don't tell us who these people are -- they tell us who they want to be.

And what about the general consumer in all of this? What do they want? The answer is simple. Any content, to any user, anywhere in the world, over any network and access modality, and to any device they decide to use. And one more thing: They want it billed the way they want to be billed. How hard can that be? We'll see in the following section.

Demands associated with media consumption are constantly shifting like sands in the desert -- and are different depending on one's point-of-view. CDNs, network providers and consumers all have different requirements and priorities. The only thing they have in common is the infrastructure over which the content is delivered.

Next: Toward a consumer-driven media model

Toward a consumer-driven media model
In April of 2007, Bear Sterns conducted an online survey of video consumers, asking 1,000 participants in the U.S. for answers to five questions:

The questions were designed to gauge market acceptance of new media and the degree to which the market would be willing to pay for Internet-based media services. Eight high-level findings resulted:

Online video demand is stronger than originally believed
Just over 50% of consumers watch video online, with 30 percent of them using streaming video at least once a week. Usage of these services is strongest among young men, with 62% of men between the ages of 18 and 24 viewing online video at least once a week.

Online video does not cannibalize TV
According to those surveyed, online video does not cannibalize traditional television. Roughly 75% of those surveyed continue to view the same amount of television as they did before. In their minds, online video augments the viewing experience.

On-demand video is snippet-based
According to those surveyed, the most common use for on-demand video is to view user-generated content such as YouTube videos, movie trailers, and music videos. Designed as short subjects anyway, these forms of entertainment lend themselves to short-subject format.

User-generated content is here to stay
Across the entire population surveyed, user-generated content is far and away the most popular form of on-demand media. Among the target market of men between the ages of 18 and 34, it is the second most popular content category. There is no question that this form of content availability has long-term staying potential.

Context -- then content
The volume of available content out there is overwhelming, even for the savviest of users. The ability to find (or better yet, the ability of the network to recommend) media that are ideally matched to individual users is importance and will contribute greatly to relevance and revenue in the future. Content is important -- no question. But the right content is even more important.

Advertising-supported video is acceptable
Relevance comes from one thing and one thing only: Self-interest. If advertisements can be sent to a customer that present a product or service they are interested in, the viewer is far more likely to watch the ad and far less likely to be offended by it -- or worse, to turn it off or skip it. According to Bear Sterns, the length of the advertisement is less important to the consumer than its relevance.

Selection, convenience, quality -- in that order
The survey concludes that most users prefer using a single, central site like YouTube for video search and viewing, particularly if the site provides good search capabilities, a healthy selection of content and an easy-to-use interface.

More speed, less delay
Most vexing to those surveyed is the need to wait for content as it is streamed to a buffer on the receiving machine. Adequate access bandwidth, therefore, is of critical importance to the market.

So what do these results tell us? They tell us that on-demand media is a force to be reckoned with and that network and content providers must evolve to support it in addition to their pre-existing services. But what are the challenges presented by the current network?

The challenges of real-time media delivery
Ultimately, consumers of media have three requirements in terms of content. They want the ability to view short snippets of video such as those seen on YouTube. They want access to IP-based Video, such as IPTV. And, they want the ability to download an entire film or program on a pay-to-own basis using Amazon's Unbox or Apple's iTV or iTunes service.

The question is this: Because these services place widely varying demands on the network, how must the network be designed to respond to them appropriately?

So, with all of this in mind, what technologies present themselves as ways to address the challenges described above?

Next: Key technologies

Key technologies
To address the issues listed above -- proper placement of the content for optimal access by customers, design of the underlying transport and access network, content analysis and recommendation, traffic analysis, and billing -- a collection of interrelated technologies have emerged. Remember this list noted earlier?

The list actually identifies the key technologies that providers must focus their attentions upon. Those technologies must provide the capabilities shown in the list above. So what are the seven key attributes?

First is universal, easily accessible broadband. Given the media-centricity of today's network, it is critical that access bandwidth mirror the bandwidth available in the core if network performance is to be perceived as adequate to the task.

Second, but closely related, is mobility. It is clear that as the wireless network evolves beyond the GSM-CDMA battle toward UMTS, HSPA and LTE, the ability to deliver media-based content to a mobile device will occur in lockstep with the customer's demand for mobile capabilities that mirror those they have when attached to a tethered network.

Third is the ability to deliver highly-customized and targeted content to each customer as required. This is done most effectively through the judicious application of data collection and data analysis. Data collection implies the ability to capture, archive and protect customer information. Data analysis implies the ability to examine and permute the data so that it is transformed into competitive intelligence that benefits both the customer and the service provider.

Fourth is the ability to gather key data and drive knowledge from it. Data collection comprises the three functions of data capture, data storage, and data protection. Data capture is performed most effectively through the use of such technologies as Deep Packet Inspection, an offshoot of IP-based networks that provides the ability to capture and analyze packet information in real-time so that decisions can be made down the road about such functions as billing, advertising, and QoS/QoE. Data storage is the ability to archive the data so that it can be safely and confidentially accessed, examined and manipulated to yield knowledge about the customer that can be converted into competitive intelligence and ultimately, customer intimacy.

If data analysis is performed properly, the information it yields can become advertising content and product and service recommendations. By all rights the best place to perform this analysis is as close to the customer as possible, which means at the edge of the network rather than deep in the core as it is done today. And since the network edge in many cases today resides inside the customer's home or business (the set-top box, DSL modem, cable modem or router), the point of data collection can be very close to the customer indeed. By placing a Deep Packet Inspection-based client inside the device that is locally resident, the service provider can monitor and analyze all IP traffic that flows into and out of the customer's premises.

Fifth is data protection, which falls into the broad category of security. Data can be defined two ways. Data-In-Flight is data that is being transported from a source to a destination by a network. Data-At-Rest is data that is archived somewhere. In both cases the data is vulnerable to unauthorized access and must be protected at all costs against manipulation.

Sixth is the need for a service-relevant network architecture. Today's centralized, core-based network, built on the assumption of an intelligent core and a less-than-intelligent edge, is not relevant to the task at hand. The critical nature of content, particularly paid-for and personal content, implies the need for a network-based storage and backup function, neither of which is presumed in the legacy network. Finally, the ability of the network to adapt to customer requirements (instead of the other way around) implies that network-based services must be available from multiple source points, and that those points must be able to change as demand shifts around the network.

Seventh is the need for the network to adapt to different access devices. A customer watching a high-definition movie on their widescreen television will have a very different experience than the customer watching the same movie on their mobile. That much is assumed. What is often overlooked is the fact that the network (or functional elements within the network) must in some way modify the content for each session to ensure that the user's quality-of-experience is as good as it can possibly be. Some kind of digital signal processing event must take place to adapt the delivered signal to the "least common denominator" of the device receiving it.

Each of these elements -- broadband, mobility, custom content, network architecture and device adaptability -- are part of the technological future that must be put into place if the promise of a media-centric network is to be realized. How to do that? Read on.

Next: The path ahead

The path ahead
If we consider the overall systemic architecture of the networked media ecosystem we find that it has four functional "regions" that work together to ensure proper delivery of content from a source to a customer. These include the actual source of the content; a stage where content manipulation takes place; the edge and core regions of the access and transport network; the content packaging region; and finally, the user domain, which comprises a diverse collection of receiving devices. Each of these regions depends on specific technologies (many of them described earlier) to carry out their various operational functions. Let's examine each of the regions in turn.

View full size
Figure 4: End-to-end secure content delivery over managed network (from LSI Lab Demo)

View full size
Figure 5: End-to-end secure content delivery over the Internet (from LSI Lab Demo)

Content sources
The content may originate at a variety of sources, as mentioned earlier, including YouTube, AOL, Yahoo!, online video providers; Blockbuster.com or NetFlix; a traditional broadcaster (BBC, NBC, Fox), telephone company (BT, Vodafone, Verizon) or cable provider (Comcast, COX, CableOne); or headend content center or digital content data center such as Akamai or Limelight who have no ownership position or use rights to content -- they are merely a conduit.

The list continues, including content data centers and content delivery networks (CDNs). Many believe that their popularity will grow as they evolve to offer better customer-centric services. A video subscriber office at the edge of the network is often used to host both traditional video-on-demand content as well as commonly requested and local content. Both the content data center and the video subscriber office rely on DSPs for code conversion; network processors (NPs) for content classification, Quality of Service, traffic management, security, and switching; and deep packet inspection for analyzing flow-through data and creating customer profiles.

Finally, the Broadcast-Multicast Service Center (BM-SC) serves as a gateway or inter-working facility between the content manipulation region and the distribution network. It receives manipulated content and repackages it in two ways. One is for distribution to the network, in which case the data is packaged for distribution in much the same way as the other data types we've examined; the other is for 3GPP-based mobile TV applications -- Mobile Broadband Multicast Service (MBMS), for example -- in which case the data is packaged for handoff to a Serving GPRS Support Node (SGSN) for ultimate handoff to a GSM-based cellular network. Once there it is delivered, where possible, to a device that can display mobile TV. Clearly a full range of NPs and DSPs are integral to the functionality of this particular corner of the network.

Content manipulation
The technologies that support the functions carried out at the "content origination" region of the ecosystem are diverse but highly interrelated. First of all, the content must be encoded, typically in a non-real time fashion, to ensure that it can be delivered to a plethora of device types. But there are also real-time functional considerations. The media stream must be transcoded, and in some cases advertising content must be inserted. In commercial environments that support the delivery of broadcast television content to a mobile device, support for the Digital Video Broadcast to a Handset (DVB-H) protocol must be performed, and it is typically performed here to ensure that the information is properly packaged for delivery.

There are also several functions that must be done in what is often called "quasi-real time." Perhaps the most important is the set of analytical functions that must be performed if customer behavior is to be monitored, analyzed, and acted upon to create competitive advantage. This is done through a combination of Deep Packet Inspection (DPI), Deep Content Inspection (DCI) and edge-based storage. At the same time, the hosting Web site must be maintained and any middleware applications (billing, maintenance, etc.) must be operated.

Clearly, specialized processing must take placed in this region of the network, including real-time data modification, content security, and parental control. Best performed in silicon, this is the domain of specialized Digital Signal Processors (DSPs) that manage these functions throughout the network.

Next: Network edge and core

Network edge and core
Inasmuch as the network is migrating from a layer two, switch-based model to a layer three, IP (packet) based model, the technologies involved here are relatively obvious: IP, Multiprotocol Label Switching (MPLS), Session Initiation Protocol (SIP), Media Gateway Control Protocol (MGCP), and all of the hardware-related technologies that show up in the infrastructure of the network as it evolves from the layer two PSTN to the layer three IP and MPLS-based network of the near future. These include high-speed switching, routing and multiplexing functions, broadband access technologies such as DSL, cable modems and wireless (GSM/GPRS, CDMA-1x, WiMAX, High-Speed Packet Access (HSPA) and eventually Long-Term Evolution (LTE). At the same time there is a related need to ensure compliance with and deployment of full-featured network management, operations support and billing support software.

Content packaging and end-user devices
In the content packaging region of the network we find the functions that ensure interoperability between the content that is delivered to the end-user and whatever device is being used by that user to access the content, be it voice, video, data or some other form of user-accessible media. At this point the network has a choice to make because of the diversity of end-user devices to which the content can be delivered. Depending on whether the ultimate destination is a television, telephone or computer (the now well-known three-screen strategy), the data stream will be encoded as required by that particular device. If the access technology is Hybrid-Fiber Coax (HFC), the data must be encoded using a multi-bit encoding scheme, such as Quadrature Amplitude Modulation (QAM); otherwise it will be encoded as required for DSL access or fiber-to-the-node or fiber-to-the-home delivery schemes. In the evolving network, a device known as a Multi-Service Access Node (MSAN) becomes centrally important because of the rise and anticipated large-scale success of IPTV.

Keep in mind that at the device level the resolution of the transmitted signal must match the display capabilities of the device that is receiving it. The spatial resolution of a typical video signal ranges from a relatively low-resolution MPEG-1 signal of 192 x 144 pixels to the gorgeous 1920 x 1080 pixel images required by HDTV. Given that the average neighborhood to which the content is to be transmitted comprises a heterogeneous mix of users with diverse devices, content must be made available to them in a variety of resolutions and encoding schemes. Of course, the ability (or desire) on the part of the hosting company to keep multiple copies of the same content encoded in different formats will be limited because of the cost of doing so.

Storage may have gone down in price in the last few years, but the volume of content that we are talking about here is massive. Storing copies of the content in all possible formats will be prohibitively expensive -- not to mention inordinately complex from a management point-of-view. Furthermore, the ability to perform inline transcoding to and from all possible formats will place an unacceptable load on the devices in the network and will necessarily limit the number of simultaneous downloads that can be performed, thus limiting network capability but more importantly limiting customer ability to buy. An alternative approach is required, that is designed to balance the requirements for bandwidth, content processing and archival storage.

Even though a large number of different devices can access the network and request content, the actual number of requested formats is finite. The list includes:

Because the list of formats is relatively short, one approach makes copies of the most commonly requested formats -- essentially those on the list, above -- and stores them at subscriber offices and regional hub-offices to ensure that the most requested formats will be made available in near-real-time. Other formats must be inline transcoded and will have to wait a short period of time for delivery, but the wait time will not be onerous. Within the domain of formats that require on-demand transcoding, a difference can be observed between "fixed quality requests" and "flexible quality requests." "Fixed quality requests" are those that cannot accept an alternate stored format and have hard quality requirements. "Flexible quality requests" are those that can accept an alternate stored format and have soft quality requirements. If, however, a "fixed quality request" becomes recurring beyond an established demand threshold, it will be added to the list of stored formats and made available on an immediate basis.

It is important to note that the headend -- that is, the ultimate source of the content -- will transmit to all downstream locations in MPEG-4 HD format to ensure that the quality of the transmitted signal is received in the most bandwidth-efficient and most appropriate format possible. The formats required in each region to satisfy fixed quality demands will be derived locally, thus ensuring efficient network operation. This service delivery architecture lies at the heart of an approach for the new media-centric network.

About the author
Deepak Kataria is the Director of Systems Solutions consulting for the Networking and Telecom practice at HCL America. Prior to his current role, he serverd as the Director of Advanced System Prototyping for the Network and Storage Products Group, LSI Corporation. His responsibilities included the prototype development of advanced networking systems for both wireless and wireline applications. Recent initiatives included video networking and distribution, femto cells and the utilization of regex and xtm engines for IPS/IDS, anti-virus and xml threat management solutions. He also supported the LSI Business Service Gateway initiatives with various OEMs. Prior to this position, Deepak spent more than 17 years serving in various capacities as a network consultant, systems engineer, systems architect, systems applications manager, and systems integration manager during his tenure with AT&T Bell Labs, Lucent Technologies, Agere Systems, and LSI. The scope of technologies covered during this career include ATM/IP/MPLS networks, high-speed switching and routing, metro Ethernet, wireless access (node B network interface, RNC), wireline access (DSLAM, MSAN), network security (IPS/IDS, DoS), hybrid P2P (peer-to-peer), home networking, residential gateways, small/medium business (office-in-a-box), fixed-mobile convergence and storage area networking. A noted industry expert, he is the holder of 4 US patents in networking, with 6 others in pending status. His work has been published in numerous publications including trade magazines and journals, has presented at numerous industry conferences, and currently serves as Industry Co-Chair for the IEEE ANTS and ICCBN conferences. He also serves on the Industrial Advisory Board for an NSF-funded project NSF 0531507 at University of Texas A&M on wireless sensor networks along with TI and Cypress Semiconductor. Deepak holds a B.S. in Electronics and Communications Engineering, and M.S. and Ph.D. degrees in Electrical Engineering from Rutgers University. He can be reached at dkataria7@gmail.com.