As high-bandwidth low-latency mul-
timedia applications such as videocon-
ferencing, online games, and virtual re-
ality applications become mainstream
on the Internet, they will saturate the
network, especially in mobile environ-
ments, much as other high-bandwidth
(but high-latency) applications such
as peer-to-peer file sharing do now.
Videoconferencing, for example, is be-
coming a common service provided by
social-networking sites (for example,
Hangouts in Google). More than ever,
real-time video streaming is becom-
ing an integral component in many
applications, such as live broadcasting
of big events (for example, the Olym-
pics, Super Bowl, and World Cup), dis-
tance learning, and on-demand video
(Netflix). In addition, the global video-
games industry was a $64 billion busi-
ness in 2012, growing larger than the
movie industry. Fast-paced large-scale
games have high bandwidth require-
ments, 1 so they do not adhere to the old
wisdom of network games having thin
communication streams.
end-to-end tcP Latency analysis
Since interactivity and transport latency are a key focus here, let’s look
at the sources of TCP latency and set
the context for Paceline. As depicted
in Figure 1, end-to-end transport latency is commonly broken down into
four components: processing delay, as
