One of Streaming Media West’s most contentious presentations came from a retired broadcast engineer – someone that cannot let sleeping dogs lie when the video streaming industry is getting so many things wrong.
“VoD is an avoidable and epically wasteful consumer of bandwidth,” declared Ted Staros, Executive Director of an initiative coined imediat and formerly of Harris Corporation and GatesAir.
While Staros succeeded in raising eyebrows with this opening gambit, there was no shortage of eye rolling too, as the talk at times bordered on directionless rant – personifying the virtually insurmountable mission to save the media and entertainment industry $billions through fixing the inefficiencies in video delivery.
Unicast is the source of this waste, Staros summizes, and multicast is the savior.
There go those eye rolls.
Let’s start by looking at the basis of the problem, before expanding on the proposed solution.
Unicast, the one-to-one communications mechanism, is the basis of streaming content in near real-time, from a single server (or software thread), to a client device. VoD comes with what Staros clunkily calls operational immediacy, whereby viewers can start, stop, rewind, and fast forward at will. The content provider uses a large number of individual sever threads to support a large number of end users, with each thread using a unicast IP stream.
In contrast, multicast takes a one-to-many approach, where a group of server threads can deliver content to a group of end users concurrently – reducing network traffic. Multicast is not without its complications, notably that multicast does not support operational immediacy.
Staros notes that supporting high volumes of unicast is not exactly a piece of cake either.
The issue is that multicast itself has failed to gain adoption. Local ISPs never saw an opportunity for multicast in the IPv4 landscape, largely due to obstacles in residential NAT (Network Address Translation) routers. The proliferation of IPv6 and 5G New Radio usurp this obstacle, claims Staros.
Support for multicast in ISP networks has been poor, in part because multicast is also an unfamiliar beast to many video service providers in the on-demand game.
A prime example is AT&T’s U-verse, which broke the mold by using multicast for live TV delivery – traversing the interdomain challenges of multicast by streaming in a private CDN. U-verse still used “wasteful” unicast for VoD and catch-up.
Enter Hybrid VoD. This is Staros’ theory for taking the operational opportunities available with multicast together with the operational immediacy of VoD – creating a highly accessible and scalable means for video delivery. There are a lot of ideas here, but not a lot of substance. This is little more than a concept about merging two deficient technologies to create a superior hybrid monster.
In danger of being unaffectionately coined the Dr. Frankenstein of video, Staros dropped some number crunching to support the arguments, claiming that the multicast model can drop capex and opex to 20% of a company’s total media assets.
Diving into the underlying protocols, brute force unicast uses TCP which, as we know, is slower with high latency, while multicast uses UDP. UDP and RTP cannot recover packets but can renumerate packets using FEC, explains Staros, which reminds us of Nokia’s recently launched live CDN, which takes TCP’s packet retransmission and wraps it up in optimized RTP.
Live streaming is where multicast methods offer greater opportunities, an observation not lost on Staros.
There are numerous moving parts that need to slide into place before a multicast-based Hybrid VoD model can become anything more than the ideas of a rambling retiree. IPv6, 5G, and ATSC 3.0 are all ripe for multicast delivery, while the declining cost of solid state storage is another economic factor to consider, with Staros claiming that NAND Flash storage could break below $10 per TB soon, from an average of less than $12 today.
Another requirement is what Staros coins “speculative caching” – ideal for fat-tail content, i.e. the opposite of long-tail distribution, where niche content finds preference over generic hit content.
In a Hybrid VoD world, speculatively cached VoD assets are stored in large Flash memory, perhaps making use of an internal HTTP server to facilitate playout through the incumbent VoD service TCP/HTTP stack.
Speculatively cached targeted ads are a must, although the nature of multicast means this cannot be done server-side, making client-side ad insertion your only option.
One transitional technology developed to stimulate native multicast implementation is Automatic Multicast Tunneling (AMT). There are two components. AMT Relay envelops the multicast content with a unicast wrapper, and listens to discovery messages from AMT Gateways. An AMT Gateway initiates connection to one or more AMT Relays, then unwraps and redistributes multicast content.
Staros namedrops Juniper Networks as an advocate of AMT, describing it as a stepping stone to fully realized interdomain multicast.
Above all, there is need for an alliance or standards body to create a common framework for CSPs, ISPs, CDNs, and set top vendors.
Staros calls out the Streaming Video Technology Alliance (SVTA) as being best situated to take on this mantel, mentioning that the work on open caching standards could be augmented to address the needs for Hybrid VoD server side management.
We learn that Staros hasn’t even reached out to the SVTA yet, but was hoping to build bridges with the organization in Huntington Beach at Streaming Media West. Unfortunately, it was just Staros’ luck that SVTA Chairman Jason Thibeault was not present at the event.
Here, Faultline has only scratched the surface of the technical details behind the Hybrid VoD concept, which are explained more fully in this whitepaper. We would never write off anything completely as a lost cause, but this comes pretty close.