ARK Multicasting does just what their name implies. Today about 80% of home broadband usage is for video, and ISPs unicast video, meaning that the send a separate stream of a given video to each customer that wants to watch it. If ten customers in a wireless node are watching the same new Netflix show, the ISP sends out ten copies of the program. Today, in even a small wireless node of a few hundred customers an ISP might be transmitting dozens of simultaneous copies of the most popular content in an evening. The ARK Multicasting technology will send out just one copy of the most popular content on the various OTT services like Netflix, Amazon Prime, and Apple TV. This one copy will be cached in an end user storage device, and if a customer elects to watch the new content they view it from the local cache.
The net impact of multicasting should be a huge decrease in demand for video content during peak network hours. It would be interesting to know the percentage of video viewing in a given week comes from watching newly released content. I’m sure all of the OTT providers know that number, but I’ve never seen anybody talk about it. If anybody knows that statistic, please post in reply comments to this blog. Anecdotal evidence suggests the percentage is significant because people widely discuss new content on social media soon after it’s released.
The first trial of the technology is being done in conjunction with a Microsoft partner wireless network in Crockett. Texas. ARK Multicasting says that it is capable of transmitting 7-10 terabytes of content per month, which equates to 2,300 – 3,300 hours of HD video. We’ll have to wait to see the details of the deployment, but I assume that Microsoft will provide the hefty CPE capable of multi-terabyte storage – there are no current consumer settop boxes with that much capacity. I also assume that cellphones and tablets will grab content using WiFi from the in-home storage device since there are no tablets or cellphones with terabyte storage capacity.
To be effective ARK must be deleting older programming to make room for new, meaning that the available local cache will always contain the latest and most popular content on the various OTT platforms.
There is an interesting side benefit of the technology. Viewers should be able to watch cached content even if they lose the connection to the ISP. Even after a big network outage due to a storm, ISP customers should still be able to watch many hours of popular content.
This is a smart idea. The weakest part of the network for many fixed wireless systems is the backhaul connection. When a backhaul connection gets stressed during the busiest hours of network usage all customers on a wireless node suffer from dropped packets, pixelization, and overall degraded service. Smart caching will remove huge amounts of repetitive video signals from the backhaul routes.
Layering this caching system onto any wireless system should free up peak evening network resources for other purposes. Fixed wireless systems are like most other broadband technologies where the bandwidth is shared between users of a given node. Anything that removes a lot of video downloading at peak times will benefit all users of a node.
The big OTT providers already do edge-caching of content. Providers like Netflix, Google, and Amazon park servers at or near to ISPs to send local copies of the latest content. That caching saves a lot of bandwidth on the internet transport network. The ARK Multicasting will carry caching down to the customer level and bring the benefits of caching to the last-mile network.
A lot of questions come to mind about the nuances of the technology. Hopefully the downloads are done in the slow hours of the network so as to not to add to network congestion. Will all popular content be sent to all customers – or just content from the services they subscribe to? The technology isn’t going to work for an ISP with data caps because the cashing means customers might be downloading multiple terabytes of data that may never be viewed.
I assume that if this technology works well that ISPs of all kinds will consider it. One interesting aspect of the concept is that this means getting ISPs back into the business of supplying boxes to customers – something that many ISPs avoid as much as possible. However, if it works as described, this caching could create a huge boost to last-mile networks by relieving a lot of repetitive traffic, particularly at peak evening hours. I remember local caching being tried a decade or more ago, but it never worked as promised. It will be interesting to see if Microsoft and ARK can pull this off.
3 replies on “Improving Rural Wireless Broadband”
The idea of basically extending the CDN into the customers homes is great. Both ARK and Microsoft will need to navigate the content licensing rules and contracts to enable this – historically the content owners get twitchy when a bunch of full digital copies of popular shows and movies are stored on customer equipment (part of why it took so long for Netflix offline viewing to roll out, for example).
I have to comment on the multicast part though. As someone who spent years participating on the Internet2 Multicast Working Group, I think it’s great to see multicast rise from the ashes – but I’m skeptical. The use case makes sense; the CPE cache nodes can all be loaded at the same time, overnight or something, with the same content dump. But multicast works notoriously poorly over wireless or anything but loss free networks. I could see a situation where a lot of local copies of shows are corrupted. Hopefully they are doing something smart like sending everything in chunks a-la torrents, such that client nodes can go back and request corrupted chunks only, then put everything back together.
This also reminds me of Qwilt which is basically a dynamic CDN that sits close to the edge, usually at the ISP. https://qwilt.com But if ARK and Microsoft can crack the nut to bring the CDN all the way home, more power to them – and I would love to see this technology for high bandwidth users too, like to get access to UHD content with less compression.
I was wondering the same thing about multicasting. The whole concept doesn’t work if people have to search for missing packets at prime time as they watch cached content. It’s one thing to cache content to one local server, but Microsoft is talking about making every home into a cache server, and there are bound to be problems caused by inside wire, WiFi and other local impediments and inconsistencies that pop up even with high-quality fiber networks. It’s hard to make homes a consistent environment.
Even a unicast delivery of a top 10 favorite shows delivered at 2am to 6am seems like it would make a world of difference. The 7PM to 10PM peak is real. I also like the concept of even if the ISP connection is down you’d still be able to have access to some content. All around, very interesting!