WP76 Superguide 1 | ADVANCED DISTRIBUTION: CDNS, ANALYTICS, QOS, AND QOE JANUARY/FEBRUARY 2017 SPONSORED CONTENT
By agreeing to use the same metrics, based on the same
definitions, each tool and service provider can compare their
outcomes and place each service into usable context. Some of
the more important of the metrics proposed are as follows:
• Video Start Time (VS T): The Video Start Time attribute
refers to the time, in milliseconds, that it takes for the first
frame of video to be rendered after the Play Initiation.
• Play Failure Count: The Play Failure Count is the absolute
number of play failures observed in the period.
• Playing Bitrate: The Playing Bitrate attribute specifies the
playback bitrate of the media being rendered, in Kbps. The
value is calculated as the video stream playback bitrate +
the audio stream playback bitrate.
• Labeled Bitrate: The Labeled Bitrate attribute is the value
passed to the video player, labeling the bitrate of the media
being rendered, in Kbps.
• Re-buffering Ratio: The percentage of time that a viewer
experiences re-buffering issues (i.e. when video stops
playing because of buffer underflow, and not due directly
to user intervention such as scrubbing or pause).
• Dropped Frame Count: The Dropped Frame Count is a
count of dropped frames that a viewer experiences during a
period of video playback.
• Video Transition Time: The Video Transition Time is the
amount of blank screen time between one clip ending and
another clip starting.
• In-stream Error Count: The In-Stream Error Count is the
count of errors occurring during the playback experience
by the player (i.e.: after the first frame is rendered).
Network metrics like these allow services to acquire
an actionable view of the QoE being experienced by their
consumers. Note that they are all defined as counting the
experience being enjoyed by the real user, rather than by
system resources: from the perspective of the consumer, the
technology assets in use to make the experience possible
Sometimes the insights that can be gained from
tracking these metrics do an excellent job of confirming
what we would expect. In the following graph, for instance,
As we might expect, when there is enough free bandwidth
to respond quickly, there is plenty of bandwidth to deliver
As we dig in, and particularly when we combine
different dimensions—especially when we narrow our
observations to particular geographies, CDNs or ISPs—
we can reach confounding conclusions that challenge our
Let’s take a look at some of the discoveries we can make,
and how they might guide capital investment decisions. The
following graphs come from an actual Cedexis client, and
compare the results tracked for each of two major global
CDNs. First let’s take a look at Video Start Failures (FIG 2):