Broadpeakâ€™s nanoCDN solution will be launched commercially early this year, giving telcos and cable operators the opportunity to reduce bandwidth consumption and simultaneously increase the QoE for customers watching live streaming video on multiscreen devices. This product has the potential to change the economics of streaming but until now there have been only limited details available about how it works and what it can achieve. Broadpeak has just started releasing more details.
The fundamental change that nanoCDN brings to the market is the ability to utilize existing multicast IP networks to deliver content into homes but continue to make use of adaptive bit rate (ABR) streaming into the devices themselves. Notably, this is achieved without needing any new customer premise equipment. No changes are needed to the â€˜playersâ€™ on the display devices; although the video is multicast across the delivery network, it is unicast inside the home as ABR.
A second innovation is the ability to harness hard drives inside the home as a local cache so that the time-shifting you typically find with live streaming, like pause and 60 second rewinds, is accommodated with some local storage. This feature will be included from launch.
The net result is that customers could watch a live football game on their Apple iPad as an ABR stream inside the home but it will be delivered using multicast on the network. Then if they rewind, service providers can still avoid them making a unicast request to an origin server by using a very local cache.
Live football on a tablet is a good example of how nanoCDN improves QoE while reducing bandwidth demands. Currently there are two main ways for a network operator to deliver this. You could broadcast a normal TV channel to a home gateway device, then turn that broadcast stream into IP and transcode it for use on multi-screen devices on the home network. This requires a new breed of home media gateway. The alternative is to stream the channel across the broadband network as a unicast ABR stream, through the broadband router/gateway without changes and direct to the tablet.
The aim of nanoCDN is to combine the best of these two approaches and avoid having to deploy new customer premise equipment (CPE). The technology relies on the service provider having a multicast-enabled network and the target markets today are telcos who already offer IPTV and cable operators with DOCSIS networks. It uses multicast to send one copy of the live channel over the local network. As with all multicasting, if two people using the same DSLAM want to see the channel, the video stream is duplicated at a local level and each home receives a copy of the stream.
nanoCDN client software is installed in the broadband home gateway. This acts as an intelligent go-between that sits between the multi-screen devices and the network (and ultimately the origin servers where the video streams come from). If you are watching a fairly niche channel or an on-demand stream on the tablet, like catch-up TV, the tablet would make its normal request to see an ABR unicast stream. In this instance, the nanoCDN client would pass this request to a unicast origin server and the video would be streamed to the home via the broadband gateway (and nanoCDN client) as unicast ABR direct to the tablet.
If you then switch to a popular linear channel, like the sports channel showing the football, the tablet will make its usual request for a unicast ABR stream but this is intercepted by the broadband gateway. The nanoCDN client software realizes that the channel requested has been assigned as a multicast channel, either permanently because it is always popular or at least for this event. The gateway client then makes its own request to join a multicast stream instead.
The channel is streamed as multicast video (from a multicast origin server) to the broadband gateway. But because the tablet is expecting to receive a unicast ABR stream, the nanoCDN client converts the multicast into unicast adaptive streaming and sends this stream to the tablet (in this example). No changes are necessary on the DRM.
So without any changes to the end display device, and without any new CPE, it has become possible to combine unicast ABR inside the home with multicast across the service provider network.
If you can get hundreds of thousands of people watching multicasts rather than making the usual ABR unicast requests to origin servers, there is potential for major bandwidth savings. But Nivedita Nouvel, VP Marketing at Broadpeak, points out that bandwidth savings are only one of the benefits.
Another is improved Quality of Experience (QoE) for the viewer. Because the nanoCDN client effectively terminates the ABR stream request from the tablet (or other multi-screen client device) and converts this to a multicast request, the ABR bit rate profile the tablet requests is determined by the network conditions in the home but not by what is happening in the service provider network. Any congestion, contention or failures in the service provider network are discounted from the equation.
The operator can provide multiple bit rate profiles for the multicast channel, and these could match the bit rate profiles used for the adaptive bit rate streaming inside the home. So as an example, you might take the channel Sports One and multicast it four times, once at 0.5Mbps, once at 1Mbps, once at 1.5Mbps and again at 2Mbps. Moreover, these multicasts can be managed so their bandwidth is guaranteed.
If there is little contention on the home network, the tablet could request a 2Mbps ABR stream of Sports One (showing the live football game). The nanoCDN client converts this to a request to join the 2Mbps multicast of Sports One. If someone starts watching YouTube on a laptop in the same home, and the home network bandwidth availability drops, the tablet might request a drop to the 1.5Mbps stream. Then nanoCDN requests a switch to the 1.5Mbps version of the Sports One multicast.
If nanoCDN was not terminating and converting the ABR request in the broadband gateway, the â€˜playerâ€™ in the tablet would take account of the new bandwidth contention in the home plus any issues in the service provider network. Rather than seeing a drop from 2Mbps to 1.5Mbps, a consumer might get a bigger drop from 2Mbps to 1Mbps instead.
According to Nouvel: â€œWith NanoCDN you will always receive a stream [on the multi-screen device] that is equal to or better than what you would have received as a unicast. This is because it erases all [service provider] network issues that could affect the bandwidth evaluation and maybe convince the player to switch to a lower bit rate stream.â€
An operator could even expand the number of bit rate profiles available on the multicast. They could offer 3Mbps or 4Mbps for use by tablets or perhaps Smart TVs or laptops if they are confident that, because devices only consider home network issues when determining the final ABR stream profile, they will be able to request (and then maintain) these higher bit rates.
Providing multiple versions of a channel that is being multicast obviously duplicates the content on the service provider network. But this additional bandwidth demand is more than offset by the ability to minimize unicasting across the network.
â€œIf operators believe that NanoCDN can save bandwidth we think they will multiply the number of profiles they offer [over the multicast network] so there is a higher level of granularity when it comes to switching [multicast] streams,â€ says Nouvel. â€œThis means you can always achieve the best quality possible, based on the home network conditions.â€
The second innovation in the nanoCDN solution is the use of hard drives in the home as a local cache. The aim is that much of the time-shifting people do when watching live streamed TV could be accommodated with some local storage.
Without a local cache, if you are watching the football match on your tablet and you want to rewind to see a goal again, you must leave the multicast (which is streamed in real-time) and begin your own unicast session across the service provider network. The nanoCDN client takes care of this, forwarding a request to begin an ABR unicast session with the origin server.
The nanoCDN clients knows when you are pausing or rewinding (or indeed, changing channels to something that is not assigned as a multicast channel). The stream between the broadband gateway and the device is always unicast ABR anyway, but in this instance the stream across the service provider network would become unicast ABR as well. If the viewer returns to live (e.g. by fast-forwarding), the nanoCDN client would request to join one of the multicast streams again.
Nivedita Nouvel says Broadpeak, working with telco partners, has studied how people use trick-play and therefore time-shifting during live television. The behaviour varies by operator and the type of content being viewed but for something like football, people leave the live stream only for short periods, often to watch a specific incident again. â€œIn sport people will probably not want to go back in time by more than 20 minutes,â€ she points out.
This is where the local cache can help improve bandwidth efficiencies further. When you press pause or rewind, if a hard drive within the home network has been recording from the multicast (and only keeping the last 30 minutes, so there is always a 30 minute buffer), this recording can then be used as the personalized viewing stream. The recording is the equivalent of a unicast and avoids the need for the nanoCDN client to make a unicast request to the origin server. The more unicast sessions you can avoid across the service provider network, the better.
The local cache is certainly not designed to give access to content that was broadcast 24 hours ago. This is for short duration recording. â€œWe think that for live sports, the local cache could handle 80% of the time-shift requests,â€ says Nouvel.
Broadpeak is working with customers to determine where the storage should reside. DVRs have plenty of storage but are not always on. Broadband gateways are always on and may have some storage, depending on their model. The nanoCDN client can be hosted by any Linux based device that resides in the home network.
The nanoCDN solution was previewed and demonstrated last year and has received a lot of attention but details about how it works have been limited. With the commercial launch approaching, the finer details are now being revealed. If nanoCDN does what Broadpeak says it will, it could have a significant impact on how telcos and maybe other service providers deliver live content to PCs, laptops, tablets, smartphones and various connected TV devices for in-home viewing.