Like life, the Internet evolves. Environmental pressures drive this evolution and so it is with video protocols.
The amphibians of the past -- proprietary RTSP, RTP and RTMP streaming protocols -- relinquished their dominance to the more cost-efficient and scalable HTTP-based progressive download (PD) reptiles. Though the reptiles thrived and can still be routinely spotted, a new class entered the world in 2008 with Microsoft Corp. (Nasdaq: MSFT)'s introduction of Smooth Streaming: mammals. These mammals combine the desirable streaming attributes of amphibians with the HTTP-based scalability of reptiles. This mammalian adaptive bitrate video (ABR) class is now comprised of numerous species that have proved adept at mastering their domain. However, they're constrained, like all of us, by their DNA. We'll get back to that in a moment.
In the meantime, lest anyone forget, mobile video continues to burn through subscriber data plans the world over. According to the most recent Cisco Systems Inc. (Nasdaq: CSCO) Visual Networking Index, mobile data usage is set to rise by a factor of ten between 2014 and 2019 to nearly 300 exabytes, equivalent to more than 10 billion Blu-ray plays of Jurassic Park. Of this traffic, mobile video is expected to climb from 55% to 72%, with more developed markets expected to see even greater increases. In short, the mobile network is now a mobile video network.
The most recent Citrix Systems Inc. (Nasdaq: CTXS) Mobile Analytics Report suggests one of many reasons for the continued video gluttony. The findings: An average iPhone 6 Plus user consumes twice as much mobile data as the typical iPhone 6-owning subscriber. The difference in data volume can reasonably be attributed to the longer duration and higher resolution and consumption associated with a larger screen size.
The good news (if one can call it that) for mobile operators is that the growth in data consumption is more or less locked in. As a consequence, operators seem likely to face a continuous increase in infrastructure costs associated with transporting this growing traffic. Given the current pace of evolution within the content ecosystem, it would appear for the moment, at least, that they are likely to capture little of the revenue destined for the pockets of YouTube Inc. , Google (Nasdaq: GOOG), Facebook and Netflix Inc. (Nasdaq: NFLX).
Given the harsh realities of this landscape, prerequisites for data services success include: (1) convincing subscribers to undertake their voracious consumption on your network, not your competitor's network; (2) nudging subscribers into the next data bucket, should it exist; and, critically, (3) ensuring that the revenue collected from those subscribers is sufficient to fund the continued supply of smooth video playback with enough profit left over to meet the business' objectives. The first two objectives are addressed by delivering a better experience; the last by making more efficient use of available network resources.
We've discovered a common (though not universal) misconception among operators regarding the ability of ABR to contribute to the resolution of the challenges above. The misconception is that ABR as a rule ensures a consistent, high-quality video experience for all subscribers and ensures efficient use of available bandwidth.
In other words, it's often believed that the previous challenges to mobile video delivery associated with lesser-evolved PD protocols have now been solved by the more highly evolved ABR protocols.
These ABR mammals possess, without question, superior adaptations for the mobile network environment than do their PD predecessors. The PD protocol is a simple creature that follows the same routine, day in, day out. Its client downloads a video in its entirety as fast as possible at a fixed quality level. Knowing this, operators have long turned to video optimization as a means of managing PD video to ensure a better individual and collective subscriber experience. Stated simply, the techniques used to do this change a PD video into what is, in effect, an ABR video stream.
Indeed, ABR is more sophisticated. It can respond to changes in its environment, adjusting its quality level on the fly as available bandwidth varies. But it faces one critical challenge: It can't see very well. It's genetic. Inside the forest of the mobile network, the mammal -- let's call it a bear -- known as the ABR client, is aware only of the conditions in the patch of ground upon which it stands. Because the ABR client lacks this information, it exhibits sub-optimal behavior both for its own good and for other clients sharing the same resources. Examples we've observed:
- A small percentage of network-wide seconds of video (8%) are responsible for a significant portion of video data volume (40%). This small percentage of video seconds is delivered at high-quality levels (bit rates up to 10 Mbps) while a significant minority of video seconds (54%) is delivered at low-quality levels (bit rates less than 600 kbps)
- The ABR client frequently changes quality levels between the reference ranges above, yielding an inconsistent experience for the subscriber
- The client discards buffered video at one quality level; then, re-downloads the same content at a higher quality level, wasting as much as 80% of the total downloaded content
These examples are indicative of the ABR client's inherent inability to make good decisions because of its blindness to the real-time conditions in the network-at-large. This bear has no knowledge of what's transpired in the past, nor of what is likely to transpire in the future. It's unaware that it's about to collide with a tree. The bear can be helped, though. What if the forest could pool the knowledge of all the bears within it and convey that knowledge to one particular bear?
The operator has the ability to facilitate just such a knowledge transfer. Because of its unique comprehension of the goings-on in the entire forest, the operator can:
- Observe the network conditions affecting all ABR clients and subscribers
- Observe all data usage on the network
- Measure and record network performance and subscriber experience in real-time
- Use all of this information to enable the ABR client to make better decisions
Fortunately for operators, many of the challenging attributes of ABR that arise as a consequence of their genetic constraints can be minimized by the operator itself through the use of appropriate analytics and optimization techniques applied within the network itself. In this way, ABR's evolutionary advantages can be enhanced for the good of both the mobile subscriber and the mobile operator.
— Mark Davis, Senior Director, ByteMobile Product Marketing, Citrix