Author: John Holobinko, Vice President of Strategy and Business Development
Five years ago, before video streams began to make up a large percentage of the overall packets on data connections, it was relatively easy to predict and manage IP capacity. While data streams can benefit from high data speeds, they are very bursty in nature and are quite tolerant of slow downs, delays and packet losses. With the rise of YouTube and other over-the-top (OTT) services, video viewership leads the pack with the greatest percentage of traffic during peak periods on IP access networks.
We’re in for another change—soon.
As YouTube and other similar sites have matured, videos are becoming increasingly longer in duration, meaning that the persistence of these streams is growing. Netflix, Hulu, TV Everywhere and similar services have much longer persistency than YouTube. At the same time, the number of smart TVs and other IP-enabled consumer devices capable of accessing this content is mushrooming. As a result, the overall persistency of IP video streams is growing, thereby reducing any benefits derived from oversubscription of the access network.
Another complication is that the types of video traffic are expanding, and IP device behaviors are not conducive to efficient utilization of access network bandwidth. A case in point is IP video devices that support adaptive streaming. Basically, these devices tell the network what IP video transmission speed they desire and the network responds. However, this mechanism is ripe for misbehavior. Why? Many manufacturers have designed their IP consumer devices to request far more bandwidth than is necessary to support their screen resolutions. The result is that network congestion can occur with far fewer devices than would be the theoretical ideal.
To make things even more complicated, the need for new video formats will drive further variations in traffic types and speeds. Currently, Motorola and others are working to establish the next generation compression standard, HEVC, which is intended to reduce further the bandwidth required for high definition video and to support “4HD” TV. The first TVs capable of displaying 4000-line television were shown at the Consumer Electronics show this year. Therefore, there will be both negative and positive effects from HEVC on network bandwidth consumption.
One thing is perfectly clear: the old data oversubscription models that were used to estimate network capacity are no longer relevant.
Motorola is continuing to model IP access network behaviors in light of these evolutionary changes. While we can correlate the results to historical data from actual networks, we will need to conduct advanced simulations to confirm predictions of how networks will behave with the added complexities of new device behaviors and new video formats.
Related Articles:
5 Reasons Why Multicast Video Won’t Save Bandwidth in an IP World—Part 1 and Part 2
From CPU to Silicon—Video Transcoding Reaches a Tipping Point
{ 1 comment }
