The evolution of the network digital video recorder (nDVR) can be likened somewhat to the development of the automobile. Even with the steering wheel, driveshaft, engine and seats all properly assembled, its success, in volume, relies on a stringent set of rules of the road. Without them, expect plenty of accidents to occur at those lightless, four-way intersections.

Still, the nDVR has taken several steps forward compared to the early days of Time Warner's MystroTV project, whose original aim was to put everything on TV on-demand. Now that some of the tricky copyright issues are starting to be ironed out, we're starting to see, if not everything, then plenty more programming becoming available on-demand than there ever was before. And the quality of the available programming is improving rapidly.

Although the nDVR evolution has a ways to go before it reaches the goal of making "anything available, at any time," here are just a few recent examples that demonstrate how things are driving in the right direction—forward:

  • Time Warner Cable recently announced it will extend the reach of "Start Over," a service that allows customers to restart select programs (i.e. those with copyright clearance) already in progress. Following the initial launch in South Carolina last November, the operator is launching it this summer to customers in Greensboro, N.C.; Rochester, N.Y.; and San Antonio, Texas. Time Warner is also working on a next step, called "Look Back," that will allow customers to access shows after they've aired.
  • Cablevision Systems Corp. is taking a much different approach with its Remote Storage DVR (RS-DVR) trial in Long Island. Rather than trying to obtain recording rights directly with programmers, Cablevision is allowing trial customers to set their own recordings on server space on the network.
  • There's plenty going on internationally, as well. In the U.K., for example, NTL and the BBC are experimenting with "Catch Up TV," a service that serves up select shows that aired sometime during the previous week.
Advantage: nDVR

In addition to providing a competitive foil, there are several other built-in benefits with the nDVR concept, according to Gil Katz, director of cable solutions and strategy for Harmonic Inc.

For starters, operators can serve up multi-room DVR functions without having to manage the tricky communications path between a primary media center and the boxes that feed off of it. With nDVR functionality, a $70 set-top can now provide many of the functions a $250 set-top with a hard drive can provide, he says.

"The cost of the nDVR, we think, is much more attractive," adds Phil Simpson, director of product marketing of on-demand solutions for SeaChange International. He notes that it's not economically feasible to offer a home-side DVR with 200 tuners, but that's exactly the type of thing a network-based version can offer.

nDVR architecture
A step-by-step look at a typical nDVR architecture.

Storing and controlling sensitive and popular content on the digital cable network can also provide an extra layer of protection from pirates.

"I'd rather have my content on servers with operators I can keep an eye on versus millions of DVRs I can't control," says Mark Evensen, VP of product development for video server and asset management firm Entone Technologies.

But there are also possible downsides to consider. The one challenge that Cablevision will likely face, if it ramps up the RS-DVR, is the amount of storage and streaming it will have to support. Under its model, if 5,000 customers request to copy tonight's episode of American Idol, the system will have to make 5,000 individual copies of it.

Most agree that the technical pieces for the nDVR are already in place. Here, we'll look at several technical and operational aspects of the nDVR and how they differ from traditional video-on-demand.

The importance of ingest

One of the big technical differences between VOD and the nDVR of the future is ingest. Ingest, in the nDVR world, needs to occur much more rapidly, almost in real-time, rather than days or even a week in the case of movies or shows tagged for traditional VOD.

"Less than five seconds is the threshold that the MSOs feel is most acceptable and [with which] the customer doesn't feel the perception of waiting," says Jim Brickmeier, the director of applications engineering for Concurrent Computer Corp.

There is, however, a tradeoff between ingest and stream capacity. Generally speaking, as the amount of ingest increases, the available capacity for streaming decreases.

"You have to be heads up on tradeoffs; that's what our customers are looking at now," says Rick DeGabrielle, president and CEO of Arroyo Video Solutions, which is supplying the servers for Cablevision's RS-DVR project.

To deal with the challenges of the nDVR, most VOD vendors have architected their systems so storage and streaming can scale independently.

They have also built their systems to rapidly ingest content and have it quickly ready for on-demand access.

For example, Tandberg Television, in partnership with Arroyo, showed off a platform at the recent National Show that can take any program off a live stream, encode it and create its appropriate metadata information, and then push all of that onto the video server, cataloged and ready to go within a few seconds. That way, the 6 o'clock news would be ready for on-demand access by 6:01, explains Braxton Jarratt, Tandberg's senior vice president of marketing & business development.

In Tandberg's case, it is supporting nDVR applications with a version of its "Xport Producer" product that includes a recorder/splicer. In addition to immediate playback, Xport is also outfitted to offer a program just after its linear run ends. If the 6 p.m. news ended at 6:30 p.m., for example, it could have a copy ready for access by 6:31 p.m. The version of Xport that provides instantaneous access and playback requires a "minor software upgrade," and could reach deployment by mid- to late-2006, Jarratt says.

Perchance to stream (a lot)

The network DVR will certainly have quite an impact on overall usage, particularly as more quality (read: popular) content becomes available for on-demand viewing. But, until there are deployments to draw on, the exact severity of that change remains a mystery.

Instead of scaling concurrency to 7 percent to 10 percent, the general rule of thumb for traditional video-on-demand, that percentage should climb higher if usage is anywhere near what it is for home-side digital video recorders.

But how much will it go up? Based on the success Time Warner has had with Start Over so far, it will be well above what's seen in traditional VOD. Citing figures from its initial Start Over launch in South Carolina, Time Warner notes that it has already generated more than 25 million additional viewing minutes that otherwise would have never been seen by consumers.

"With the network DVR being used as an alternative to a set-top-based DVR, you could easily get upward of 50 percent simultaneous usage," predicts Mitch Auster, senior director of product marketing for Ciena Corp., which handles VOD transport for some cable operators. In homes with multiple TVs watching different time-shifted programs over the network, "you could get over 100 percent simultaneous usage," he adds.

Bigger stream counts should also drive usage of RAM-based technologies.

Broadbus has built its nDVR architecture around D-RAM, which scales particularly well for ingest. Broadbus' system, says Senior Director of Marketing Tom Kennedy, can ingest just under 1,300 channels simultaneously. That should be more than enough. Time Warner's Start Over system, by comparison, has the rights to ingest just north of 60 channels.

D-RAM, though expensive relative to disk-based storage and streaming, is well-suited for popular content. A fully-loaded Broadbus B-1 could produce 20,000 streams off of just a single asset—something that might come in handy in the Super Bowl scenario, where a large audience is watching and pausing and rewinding all at the same time.

To deal with such scenarios, Concurrent and other video server makers are providing more RAM—either as a stand-alone product or coupled hybrid-style with disk-based servers.

SeaChange is also offering RAM capabilities through a blade that can be retrofitted with the company's disk-based systems. Today, those blades can only support a few hours worth of memory storage. SeaChange, however, has a fully RAM-based server on the roadmap. But not everyone believes RAM is going to be the way of the on-demand world anytime soon.

"I don't think we've hit a level where RAM is an absolute requirement yet," Concurrent's Brickmeier says. Still, that hasn't stopped his company from unveiling a RAM-based server of its own, the MediaHawk 4500. "RAM is for future-proofing for when nDVR content is freed up, and there are rights [available] for better content. From a VOD perspective, the benefits of RAM are not quite as obvious," he argues.

Tricky trick files

Yet another challenge the nDVR must address involves "trick plays"—that is, the function of rewinding, fast-forwarding or pausing the stream. Home-side DVRs begin to rewind content almost the moment the user presses the corresponding button. If the nDVR truly aims to replicate that experience, the responsiveness of trick plays should be very similar to what one expects from a TiVo, for example.

Arroyo claims it's pretty close to that—when the VOD backoffice is taken out of the equation. With a direct-to-server connection, it takes about one-third of a second from the time a customer hits the fast-forward button to the time the video is fast-forwarding on-screen, DeGabrielle says, but acknowledges that more latency is present with the backoffice in place. And that second scenario is a distinct possibility if operators want to use the backoffice system to track trick mode usage and other reporting.

But the delays in trick modes for VOD are the product of the network and the number of jumps the communication has to make.

The nDVR "should work like the TiVo," says Mark Crandon, VP of marketing of Kasenna Inc.

Paying the freight

In addition to obtaining rights from programmers (unless the Cablevision approach is invoked), the nDVR is in need of establishing a true business model that programmers can believe in.

"Frankly, we've had network DVR functionality for close to a decade at this point," says Joe Matarese, SVP of global advanced technology for C-COR Inc. What's been holding it up is the business end, namely the creation of an ad-supported model. "We're talking about television content, after all," he adds.

Matarese believes the cable industry has to continue to migrate toward a unicast model that will allow for more targeted advertising, as well as a more dynamic advertising rotation model. Some ads, he explains, can become irrelevant if they promote an event for the particular evening they originally aired. "If you time-shift more than 24 hours, then that ad is completely wasted," Matarese says.

C-COR, he notes, has at least one pilot slated this year (with Atlas, a company that has recently applied its Internet advertising acumen to the world of VOD).

SeaChange is also involved with the ad side of VOD, having released a platform called "AdPulse." Initially, AdPulse is targeted to insert ads within traditional VOD content, perhaps at the start, middle and endpoints of a program or movie. The system is in two "major" trials with undisclosed operators.

The rules of the road

The nDVR will also have to support business rules for recorded content.

In the Cablevision example, the system will allocate disk space and could determine how long a given recording can remain on the server.

Other system intelligence and program discrimination tactics might also ensure that customers can't record a VOD movie to their allocated space on the nDVR server farm.

Those rules might also disable the fast-forward feature so that viewers have to watch the ads.

Broadbus, for example, is building a suite of applications that generate the "rules engines" for the network. That should help to automate the business end of the nDVR so that it can be determined which shows are okay to be copied for playback and, if so, determine what their availability window is (24 hours, 48 hours, a week, etc.). SeaChange is also working on a business engine that could be built into its own VOD system.

Kennedy acknowledges that the Broadbus approach is still in the early demonstration stages, and is considered more "proof of concept" at this juncture. But once the concept is ready, then the proof will be in the pudding.

Once the business end is settled, "the floodgates are going to open," Simpson says. "Vendors with those [nDVR] solutions will be selling them the next day."