Back in April at the US NAB show, Harmonic began virtualizing the process of encoding, making it a pure IP play for generic cloud servers. This week it is extending that vision and because we didn’t give it much of a write up before, we thought we’d at least go over the Harmonic announcements.
Essentially VOS is a virtualized environment for cloud servers, so that Harmonic can offer encoding, graphics and playout on a single server configuration, because all of the features are carried out in software. It can be an encoder one minute, then when the encoding is done it can lay on graphics over the video and when that is done it can playout the video to consumers.
The demo at the HP event, HP Discover, this week, was obviously showed on an HP server. But we remain a little puzzled and surprised that Harmonic has gone down this route. It was back in April that its Electra XVM package was launched, the first VOS app.
We can see how this represents a good route for anything up to HD encoding, and that a private cloud or a full open cloud formation can be used for all the wide and varied steps in video preparation, and that this can be used across both private installations and the cloud, in a hybrid. That may make big savings for Harmonic customers who previously had to decide how much processing power they needed for each job in the video production chain and now they only have to calculate total capacity.
What we can’t understand is how this works in HEVC. That’s because so far Harmonic has said that software only is not enough and has gone down the specialized hardware route for HEVC, in particular 4K. It announced around the same time in April that it would use Altera’s new 4Kp60 capable H.265 enhanced motion estimation engine (EME) based on an FPGA as part of a server co-processing package also featuring its own HEVC software. Harmonic claimed this will enable the system to evolve, while also offloading the computational intensive aspects of HEVC encoding to the dedicated FPGAs.
So it would be able to do some HEVC processing in a cloud formation, but it would need to install the special FPGA to get the full system working, which would make it a non-cloud system, and less useful for virtualization. There wouldn’t be much point virtualizing the processing power of the FPGA to other applications, because it’s optimized for a special function, not even the entire encoding process, just part of it.
But that explanation aside, we agree that more and more encoding, transcoding, preparation of ABR streams, graphics editing and playout and other functions will all need to be run in software on virtualized hardware.
The shift to VOS is very similar, but not identical to the way Ericsson went with its own NAB announcements in April. There it talked about virtualizing only encoding (Ericsson Virtualized Encoding or Eve) and allowing the encoding resource not only to be an Ericsson device, or software process, but allowing other encoder makers to certify against this, with Elemental being the first to jump on the bandwagon.
The Ericsson system arbitrates among any encoding resources you have access to, whether these are in-house or cloud based, and selects the right platform for a particular video output on the basis of priorities set by the operator, relating to quality, cost and performance.
In that environment a company like Harmonic could either admit that its own encoders are a little behind the times, and use the most advanced out there, or install the Altera chips inside servers in your IT department and offload just motion estimation to them. That way really complex, but efficient installations, of hybrid cloud video resources could be put together, that use state of the art in both software and hardware breakthroughs.
After all the year old Ericsson AVP 4000 video encoding chip looks like it is not very different from an FPGA, and so would have to be used in much the same way as the Altera chip in an Harmonic installation.
To us it looks like both companies are heading down the same path, perhaps emphasizing different aspects first, creating an operating environment that will be tempting to fall into, especially as more tasks become virtualized, but even more difficult to get away from once you have loaded up all the resources and your workflow has become entrenched in a set of virtualized devices. Departing from VOS or EVE will be the problem, not replacing actual encoding resources.