Best of NCTA Tech Papers: Part 2
As promised: More tasty tech nuggets from this year’s compilation of Technical Forum papers, selected and aggregated by the National Cable Television Association and the Society of Cable Telecom Engineers (SCTE).
Starting with what it took for Cablevision Systems to build a network-based DVR service. Sure, everybody talks cloud DVR now, but Cablevision was first to market, arrows and all. How it all came to be, and what it took, is compiled in a paper simply titled “Remote Storage DVR,” by Stephanie Mitchko, SVP of Video Infrastructure Software.
A term you’ll learn: “COTS,” which stands for “Commercial Off-the-Shelf.” As in servers. Putting video “in the cloud” requires lots (and lots) of them.
Surprise, surprise: People record more than they watch. If you’re into a graphical representation, Mitchko includes two charts showing daily volume of both recordings and streaming activities.
This year’s collection is heavy on the SDN – Software Defined Network. It’s a cloud thing, and not the easiest concept to comprehend. Which makes this explanation, from Cisco’s David Lively, particularly handy: “The general definition of cloud is resources, abstracted from the physical hardware, delivered elastically and on demand.”
Because of that, Lively continues (in a paper titled “Software Defined Networking and Cloud – Enabling Greater Flexibility for Cable Operators”), applications can be deployed and scaled without having to physically configure and deploy new hardware.
And because of that, network devices can be outfitted with “programmatic interfaces to define network behavior, utilizing business intelligence and policies.”
I try not to miss anything written by cable modem oracle Gerry White, Chief Architect of Network Infrastructure for Arris. This year’s Gerry-paper (“Can DOCSIS Networks Leverage SDN?”) is a little jarring, because of what it means to people and plant (goodbye, headend!), but such is life.
“In this architecture, the headend in its current form can be replaced by a data center, an Ethernet distribution hub and a simple node,” White writes in his assessment of what it would take to move parts of the cable modem specification out of hardware and into a software-defined network.
Comcast Labs figured out a way to predict the popularity of TV programs, by algorithmically combining DVR scheduling data (gleaned from its xfinity app), social media and historical Nielsen data.
In “What’s Hot: Linear Popularity Prediction from TV and Social Usage Data,” authors Jan Neumann, Xiaodong Yu and Mohamad Ali Torkamani detail how they popped the predictive capabilities of top-10 programming by 4 percent, over Nielsen-only methods.
Plus they drop this keeper of a quirky phrase: “In the end, we got the best results using shallow random forest regression trees.” (As opposed to deep deliberate desert progression flowers.)
We’ll spare you a Part 3. But as tech paper compilations go, excellent haul this year!
This column originally appeared in the Platforms section of Multichannel News.