Your browser is not supported. Please update it.

4 September 2014

VideoFlow momentum grows in patent pending video delivery

Whenever scientists look closely at a highly specific problem, they can come up with highly specific methods for solving it. So while the generic problem of sending video across the internet has given rise to some of the biggest mathematical breakthroughs in Forward Error Correction, like the Raptor codes and Luby transforms, solving the problem of preventing buffering in Adaptive Bit Rate steams is a far more specific problem and we met this delightful bunch of Israeli scientists at IBC last year, at a company called VideoFlow, who now think they may have the answer.

The way their system had been architected is not dissimilar to how a number of other video protection systems have been designed – and it involves storing a video buffer in a hardware device, just say 100ms of backlog, and taking over responsibility of resending any late or missed packets, where any video packets are lost.

We remember Cisco once introduced something like this for resending lost multicast packets, to individual set top boxes, but sending them unicast, fast enough to arrive before a set top box buffer was full. It never caught on, but the system on the D servers of Microsoft Mediaroom did something similar, whenever anyone decided to change channel and it needed the I-Frame from 500ms ago, in order to resolve interim frames, so that was sent unicast at the beginning of a channel change and then the multicast information made sense and could be resolved when it arrived, in under 100ms – a breakthrough.

When we met VideoFlow last year it was applying its patent pending approach to the problem of contribution streams, going head to head with companies like Net Insight and in particular its recent introduction of an open internet version of its Nimbra product line.

If you can send live football matches back to the studio over the open internet, rather than leased lines, your costs are going to be dramatically lower, but you really have to be sure that the video packets will ALL arrive and all arrive in time.

The wider uptake of ABR protocols also makes this type of technology even more desirable in the final stages of video delivery to end user devices and the problem is to eliminate jitter (which if it creates too much of a lag has the same effect as packet loss) and packet loss which comes from a variety of sources, anything from the poor quality of copper access lines, to buffer sizes of routers “en route.”

The VideoFlow solution is to apply an FEC to the resent packets which is proportional to the amount of loss, so that it adds the smallest amount of bandwidth overhead. It gives this example, if a content stream is 10 Mbps per second, the content distributor might add 20% to the bandwidth in order to provide some protection. But this means that distributors will be forced to pay for an additional 2Mbps of bandwidth. Deploying the what VideoFlow calls its Digital Video Protector (DVP) will reduce that overhead bandwidth to 0.5Mbps , keeping down costs.

The DVP  talks to another box at the receiver end which it calls a Sentinel (we assume that or last mile access this can be at the edge of the CDN) so that all packets have an arrival guarantee.  It is now offering what it calls it Controlled Adaptive Rate (CAR) and the key in this it says the trick is for the decision of which ABR stream to use does NOT to reside in the end user device. What happens instead is that the remote device sends data about what is getting through back to the origin, and then transmission side decides what bit rate it can support.

It is ABR or very like ABR, where the receiver probes the network’s capacity and the video stream health in real time, but sends that data back to base where CAR dynamically sets the highest bit rate that each receiver can achieve without this interfering with the video stream.  CAR is encoder independent and will work with any encoder supporting changing the bit rate on-the-fly.

Using the protector and adaptive FECs, the VideoFlow technology can maintain a higher quality video stream in a lower bandwidth, so temporary fluctuations do not result in a shift to a lower bit rate stream.

This year in the lead up to IBC we have already received about 6 press releases from VideoFlow, with client wins including one at Sapec in Spain which makes video transmission systems, who says it will launch a range of encoders and decoders with the technology built in for contribution delivery.

Another win is with Teleradio Padre Pio, an Italian broadcaster, using the DVP technology to replace an expensive satellite backup solution to its fiber connection and the ROI offers payback in a few months.

The company said that the system had to prove itself when the main fiber connection went down a couple of times for several hours and the VideoFlow connection took over without any break in transmission or change of quality.

VideoFlow now says it will be launching an addition to its DVP systems at IBC with a new DVP100X handles up to 80 concurrent streams while maintaining 80 concurrent VPN tunnels, each at up to 300 Mbps.  Again this is targeted at the original application of large live contribution feeds over the open the Internet with no packet loss.

Another design win that VideoFlow is touting is at Alma TV a cable TV operator in Kazakhstan offering 120 TV channels and which it gathers from around the world over an MPLS network.
Dmitry Basin, CTO at ALMA TV, explained,  “We decided to use MPLS, which is considered the most reliable form of IP network currently available for the video delivery.  However, we experienced video quality problems so we tried almost every standard approach including Pro MPEG FEC (Forward Error Correction) to resolve it, but none of these legacy technologies could deliver the quality that we expect to provide to our customers.” The VideoFlow system was tested and subsequently employed.