Cable Adapting to Video's Streaming Future
Adaptive video streaming is already commonplace within the wilds of Web video, but the cable industry is also interested in taming that technique as it chews on TV Everywhere strategies and ponders ways to deliver video to PCs, smartphones, and broadband-connected tablets.
Cable's not deploying adaptive streaming in a big way yet, but the concept is clearly on the cable radar. At The Cable Show in May, Comcast EVP and CTO Tony Werner remarked on a panel that the MSO was conducting some proof-of-concept work on "fragmented MPEG-4" as it weaves together an IP video strategy. The MSO hasn't said much more about it since, but one industry source says work is underway at the Comcast Media Center (CMC) in Denver, and Harmonic Inc. (Nasdaq: HLIT), whose MediaPrism product would seem to fit the bill, is one of the key vendors that's involved. (See Harmonic Goes Multi-Screeen With MediaPrism .)
Werner's comments suggest Comcast is more interested in handling the transcoding and other heavy lifting on the network alongside switched digital video (SDV) techniques, rather than relying only on a new breed of souped-up residential video gateways that will be able to translate a primary video stream into various formats. (See Docsis 3.0 Enters the Gateway Era .)
Using the Web TV wheel
If Comcast and other MSOs end up heading in the direction of adaptive HTTP streaming, they would jump on a wave that's already rolling through the Web TV world. It's understood that all content distribution networks are standardizing on HTTP-based delivery methods, with many leaning toward support of adaptive streaming platforms developed by Microsoft Corp. (Nasdaq: MSFT), Adobe Systems Inc. (Nasdaq: ADBE), and Apple Inc. (Nasdaq: AAPL). Ironically, Comcast-backed Move Networks Inc. helped evangelize the idea of chunk streaming. It's now evaluating strategic alternatives, including an outright sale. (See Comcast Moving On Move Networks Alternative.)
HTTP adaptive streaming would present a new technical angle for cable, which has largely relied on MPEG-2 transport and User Datagram Protocol (UDP) streaming for traditional, set-top-based video-on-demand (VoD) services. That current method is hampered by timing restrictions, as every frame has to leave a server at its appointed moment, explains Santosh Krishnan, VP of product strategy at Verivue Inc. , a maker of Flash-based media switches that are expected to factor into the cable CDN movement. (See Verivue Tech Gets a Toehold at Shaw and Verivue Surfaces With Comcast Backing .)
Adaptive streaming also strays from the progressive downloading techniques YouTube Inc. , Netflix Inc. (Nasdaq: NFLX), and others started off with for Internet-delivered VoD. Netflix has since found adaptive streaming religion as well.
Although progressive download served a purpose, it's "not a best-effort type of model," Krishnan says, noting that the source file is encoded once, and the technique allows for content to burst into the buffers and play back on its own accord.
Adaptive HTTP streaming is a much different animal that allows a movie or TV to run smoothly as bandwidth fluctuates. It does this by taking the TV or movie asset -- an MPEG-2 transport stream or a mezzanine MPEG-4 file -- and chopping it into tiny chunks. In the case of Microsoft's Smooth Streaming approach, each chunk is about two seconds in length.
Once those chunks are set up, an associated software client (in the browser or installed as a plug-in) in the PC, iPad, or some other end device, puts out requests for those fragments and stitches them together like a playlist. The client also monitors the buffer levels and checks bandwidth and server conditions. If there's a drop or rise in bandwidth, it can make decisions in these two-second increments and adjust to a lower or higher bit rate version.
"The client and server have to be synched," says Yuval Fisher, CTO of RGB Networks Inc. , a company that's tackling this with a dense transcoder that can spit out a wide range of bit rate profiles off a source piece of content. (See RGB's TV Everywhere Offer: A Video God Box .)
These adjustments are supposed to be subtle. "When a client asks for a subsequent chunk, on screen you don't recognize there's a blip," Krishnan says.
And that model can come in handy especially in the world of smartphones. "If [bandwidth] quality changes from a 3G cell to a 3G cell, the client can adapt and request the chunk that's suitable for the bandwidth it has," Fisher says. In the TV realm, an MSO could also create adaptive streaming profiles for 3D programming.
And it's not just for VoD. The process would be "perfectly viable for linear content," says Fisher. Widevine Technologies Inc. , a company that specializes in adaptive rate streaming, demonstrated how that would work at the recent CableLabs Summer Conference in Keystone, Colo.
However, Widevine, a firm that has been taking the CE and service provider markets by storm, says it takes an entirely different approach with an adaptive streaming system that aims to be device and network agnostic. Widevine likewise claims that its approach, which does involve device-side clients, doesn't require any special technology in the CDN. (See Boxee Teams With Widevine, Widevine Protects Sling's Streams , Widevine Locks In Dish 'TV Everywhere' Deal, and Widevine Ties Up Web TV Deals.)
But it's also being tight-lipped about how its approach is different, other than stressing that it is different. "We don't chunk," says Widevine CEO Brian Baker. "We're in another category, and arguably one that's being much more widely adopted. We don't want to advertise to our competition why we're taking all the deals."
Adaptive streaming, at least at a high level, is supposed to eliminate the need for a purpose-built architecture, though operators would still need to support centralized, dedicated encoding farms where they could generate the fragments that will end up on the CDN caches. "But you can use the standard HTTP CDN model and yet support streaming that [offers] quality that's close to older UDP technologies," Krishnan says.
Adapting for TV Everywhere
And the same technique could help cable support myriad video platforms. In addition to supporting fragmented streams based on how much bandwidth is available, subsets of those chunks could be made to address different device "profiles" for PCs, and Android phones, iPhones, and iPads.
Those fragments would typically be generated by passing them through transcoders that would flush out the different required resolutions and device profiles. That process might end up generating 14 or more variants from the original mezzanine file to ensure that any bit rate adjustments aren't jarring to the viewer.
Another benefit is that cable wouldn't have to worry about proprietary streaming formats or specialized hardware or software. HTTP Web caches would do the trick because they would hold all the necessary fragments for various bit rates and device profiles. Plus, content streamed this way has no troubles traversing home networking firewalls.
"It's really a revolution on video delivery," RGB's Fisher says of the HTTP-based approach. "The server is simple and all the intelligence is in the client. You don't need a complicated server anymore."
But don't look for cable to start supporting this sort of thing overnight. That work remains in the "early stages," according to Krishnan. "But they are thinking about building networks in that fashion."
Baker is more bullish, confident that MSOs will start to support a handful of non-traditional mobile devices, including tablets, by the first quarter of 2011.
Comcast, as noted earlier, is showing interest in the technique, and it has a big CDN project underway, with scaled commercial deployment in its "Freedom Region" (Philadelphia and Cherry Hill, N.J.), with Washington, D.C., on deck. However, the initial focus will be on beefing up the MSO's traditional VoD lineup rather than on TV Everywhere and supporting gobs of different mobile devices. (See Comcast's 'Project Infinity' Takes Flight and Cable Thinking Big With Video-Focused CDNs .)
Despite all the benefits, there are drawbacks and tradeoffs with the approach. For starters, MSOs would have to boost their encoding capacity to handle all the supported bit rates and devices. That would likewise boost storage requirements. It also presents a fairly significant content management challenge because the backoffice must now handle multitudes of these two-second files. Goofs in fragment metadata or device profile information could cause the targeted quality of experience to go downhill.
— Jeff Baumgartner, Site Editor, Light Reading Cable