In March 2012 with a headline which said, “Tversity throws content at the set top, with just software,” we met this company at IPTV World Forum and talked about how tricky it was for Apple’s Airplay to be copied without using any chips and working via a set top.
This week the same CEO, Ronen Mizrahi, called again with the news that after working for 3 years on the product in conjunction with multiple operators, one extension of that product, which he called AppFlinger had made it into the Dish Hopper 3 which was announced last week at CES. He used the platform to relate an entire vision of how operators need to behave in an app dominated world and how cloud processing would change the entire video market.
Essentially all that the Hopper 3 and the Hopper 2 will do with App Flinger is allow casual games to be played on the TV screen of Dish Network customers. But this is not a small thing, Mizrahi explains and the explanation is worth sharing because it takes in the entire transition to cloud. “Dish used to have a gaming option before and it made them a lot of money, but they lost it when they moved from one device to another. There were too many problems getting each game ready for the next platform,” says Mizrahi warming to his theme.
What AppFlinger does is it render in the cloud anything written for an HTML 5.0 browser, whether that is graphics, a dynamic UI or video. That is then sent as not one, but multiple, video streams to the device to play.
This sounds very familiar and two names leap out at us in Active Video and Metrological – one that renders a UI in the cloud and overlays it into video using MPEG Stitching in a single video feed, while the other plays apps on a TV which have been rendered in the cloud. Mizrahi nods and talks about the detail.
“ActiveVideo was designed to place a new UI onto an old machine, to preserve that machine’s life. We never set out to do that, we wanted to build apps and video for new set tops, but move as much processing as possible into the cloud, but make it responsive to improve the user experience. We also wanted to make it cost effective.”
Mizrahi is referring to the concern we have come across proposed as a limitation of ActiveVideo, which is that if it has to go into every single MPEG frame and overwrite where the UI is, this will lead to a large amount of cloud processing.
Before our friends at Active Video complain, it’s not a problem if you, like Charter Communications in the US, want to invoke an entirely brand new UI, in order to miss out on an entire generation of device purchases – that lack of capex is more than worth the extra cloud processing.
But Mizrahi disagrees, “When you combine a UI with video as a complete overlay, there are not good economies.” So how does App Flinger do it?
“We create two videos, one of the video material and one of the UI, and if you have a source video like YouTube, there is no transcoding, just send it through, and if your UI needs to be shown, you have the device to render the second video of the UI over the video.”
But what about devices which don’t have the manpower to cope with two video streams? “Well first we are not using hardly any cloud processing power, because the video that lays the UI on top only needs to work at something like 3 frames a second, so it is one tenth of the processing power that you need for video. Secondly when a set top is not man enough to do the job, we use a 3rd channel of communication to tell the local media player to switch between the two. Take a single YouTube app – when the YouTube video stops playing the UI can begin again. If there is not enough processing power you turn the UI into JPEG images which sit over the video on certain areas of the screen, while the video is playing.”
So apart from using less cloud processing, why is this better? To Mizrahi this is all about how operators need an architecture which allows them to join in with the App world and yet not inherit the problems that come with it. What happens when app providers issue updates – you don’t want the update going down to a set top to run, especially not while you are watching TV. And you don’t want to write a different version of the app for every chip that you have in your installed set top base, and you don’t want different apps taking up different amounts of processing power on the target device – you want to be insured from all of that. He also talks about it be the most “secure” way of offering apps, and operators also don’t want a security problem compromising their CPE in the field resulting in expensive help desk calls.
It is even tougher for an operator that embraces smart TVs directly, instead of a set top – these change from device to device, and sometimes the chip or OS can change from one generation to another.
“We want to isolate the set top from updates or processing, and this makes them reasonably predictable in terms of how the app works, how much power the device uses, and if prevents a fragmentation problem for the developer.”
It was then when we said, “Oh, rather like Metrological,” to which Mizrahi said, “We are complimentary to Metrological, and we are partnered with them, but they work slight differently from us, using an abstraction layer, and they need to optimize their system for each platform they run on, whether that is new chipsets of new cloud platforms.”
The Hopper 3 is based on the 21000 DMIP Broadcom 7445, Quad core ARM architecture, so it has a ton of power. So it’s quite different from previous MIPS based Broadcom chips, so if the Dish gaming system had still been in use, it would have required an entire re-write for the new chip, so we can see Mizrahi’s logic working in this decision.
Mizrahi adds, “With our implementation it takes one developer a couple of weeks to port our system from one chip to another, no longer. And right now HTML 5.0 is still beyond most chips to run natively. Certainly no set top chips and even the larger processing power of smartphones and tablets, it doesn’t run on smoothly yet. Even the most powerful chipsets are too slow. Which is why Dish in the end decided to use our system to run these games apps in the cloud.”
Finally, his vision, “The emergence of Apple TV and Roku and Chromecast sets a real challenge for pay TV operators. They don’t want to lose control of their environment, and unless they provide an alternative to the major internet App architectures, they will become submerged under someone else’s apps. They need to think of themselves as an aggregator of apps, but without all those fragmentation and update problems, and with unified search and discovery and UI.” Mizrahi clearly thinks that Tversity has part of that answer with its AppFlinger
About a year ago Tversity was scratching around for investors, today on the back of this Dish deal, it is likely it can take its pick of funding sources, and its plan to develop in Europe (because it is US based) is to target the Zenterio set top operating system and it has a partnership there. Given that Zenterio has already made inroads at Deutsche Telekom, and DT longs to work in a pure cloud environment where it doesn’t even have set tops, the chances are that we may see Tversity riding into more tier 1 accounts on the Zenterio wave.