By Beecher Tuttle
A recent blog post by Bell Labs’ Harish Viswanathan and Mark Clougherty, "Optimizing HAS for Mobile Wireless," looked at technical issues surrounding the delivery of quality video over mobile networks. There they discussed in detail why HTTP adaptive streaming (HAS) is capable of doing so but not at its default settings.
This is an important finding since HAS currently is used for delivering acceptable video quality over a wide range of typical network conditions, and thus is a strong option for wireless networks where data rates can vary substantially.
With the explosion of smartphones, tablets and PC with cellular network cards, the ever-increasing demand for streaming mobile video applications has forced service providers to examine the manner in which on-demand content is delivered. With this in mind, the Alcatel-Lucent Bell Labs’ researchers conducted a study to assess the ability of HAS to deliver optimized video content through wireless networks while providing a high-end quality of experience (QoE).
As mentioned, HAS is commonly deployed and well understood in wireline networks. However, very little has been known about HAS in wireless environments. Viswanathan and Clougherty discovered that using the default HAS parameters led to significant difficulties due to its propensity to overreact to transient spikes in download bandwidth. The issues included undue panic drops, variations in streaming data rates and generally poor video quality.
So, the researchers toyed with the parameter settings to discover if any changes would result in improved average quality levels, better overall client performance, as well as controlled panic rates and rate variations. The team also looked for ways to improve bandwidth utilization.
After tireless efforts, the researchers found that when several key HAS client parameters are optimized for mobile settings, HAS transforms into a highly reliable delivery method for on-demand mobile video.
These tuning variables (shown below) include the maximum buffer size, the steady state threshold, the dead zone width, the slope threshold and the download bandwidth averaging window, among others. With changes to the default settings of each of these parameters are made, mobile video content delivered through HAS gets drastically improved.
In its simulation, Bell Labs discovered that "the magnitude and frequency of the chunk rate variation is reduced…and panic events are completely eliminated, resulting in improved QoE," the authors note.
Each specific parameter setting provided unique benefits. A larger client buffer enabled more effective use of bandwidth, while the deeper buffer resulted in a more stable chunk rate. The narrower dead zone and the increased slope, meanwhile, allow the video client to mitigate panics.
The authors stress that HAS is "well-suited" for video optimization over mobile wireless networks, but only when client parameter settings are customized to compensate for the evolving network conditions.
"They must allow for more cautious rate increases and more agile rate decreases than are required for typical wireline applications," they add.
Viswanathan and Clougherty say that, hopefully, HAS will one day advance to the point that it can recognize whether it is working over a wireline or wireless connection, and adjust accordingly. Until then, adjust those settings.