Thursday, February 25, 2016

50-80 Peak Busy Users Contend for the Advertised Rate (Revised)

Chart 23 is from the  
page 50 

Chart 23 shows that data consumption follows data rate.

For a scalable service,  offer a certain "data rate" that will manage the total usage, and from that share that certain "data rate" with more than one user.  Each service can be shared by a few users without degrading the customer experience expected from the advertised data rate.  A contended rate of 50:1 to 80:1 peak busy hour users is applied.

Service to home users grows dramatically up to about 5 Mbps, and continues to grow at a moderate rate with no end in sight.  For the purpose of this discussion, I am branding a 5 Mbps service as a good service based on it being a break point, both to rates above and below.
  • Going from 5 Mbps to 1 Mbps means a five-fold decrease in data consumption,
  • Going from 5 Mbps to 12 Mbps is about a 50% increase in data consumption, 
  • Going fro 12 Mbps to 25 Mbps is aout a 15% increase in data consumption.

For a 5 Mbps service

  • Keep available data rate more than 5 Mbps (or consumption dries up)
  • Average throughput per user is 63-100 kbps

For a 12 Mbps channel

  • usage would go up about 50% from the 5 Mbps channel
  • Average throughput per user is 150-240 kbps

For a 25 Mbps channel

  • usage would go up about 15% from the 12 Mbps channel
  • Average throughput per user is 313-500 kbps

For a 1 Mbps data rate (a dreary service)

Some applications are more tolerant of disrupted transactions.

A network does not have the capacity to queue up data link effectively, and will drop packets when oversubscribed.  Quality of Service is applied to favor those packets in some sensible manner to preserve the highest user experience.

Some applications are not very tolerant and can timeout, triggering annoying error messages.

Attempts to stream or large file transfers can grind to a halt and fail after watching for an eternity.

The user can adjust their behavior.

With a 1 Mbps service:
  • Usage would be suppressed to about 20% of the 5 Mbps service
  • Average throughput per user is 13-20 kbps

NOTE: this is all about the forward channel (to the user), and ignores latency, which can reduce throughputs, and cost, which can limit usage.

Peter Lemme

Copyright 2016
All Rights Reserved.

The following are extracted from the FCC report, verbatim


Test traffic data use is tracked and subtracted from each consumer panelist’s personal data usage, which allows us to include a chart demonstrating consumer data consumption. 

The data was taken from a subset of 5760 measurement devices that were active during the measurement period which reported a total of 301 terabytes of data consumed, which represents the amount of data uploaded and downloaded through all measurement devices across the panel, minus traffic associated with the program. 

Chart 23 (above) shows the average amount of data traffic consumed by users in each speed tier, normalized as a percentage of total traffic generated by all consumers. This normalized view of user traffic shows a correlation between data consumption and speed tiers. 

In general, we found a correlation between higher speed tiers and greater data consumption by the average user. This could mean that, as higher speeds are made available to consumers, consumers increase the amount of data they consume through some combination of greater use of the Internet and adoption of more data-intensive applications and services, or that consumers who use more data- intensive applications on the Internet tend to subscribe to faster speed tiers.  

Chart 24, above, shows the cumulative distribution of traffic by technology. 

One important note about the data consumption information presented in this Report: the panel methodology specifically attempted to exclude both users with high consumption profiles and very fast tiers that had relatively low subscription rates. For these and other reasons, while the data do show a correlation between speed tier and data consumption, no conclusions can be drawn about total data consumption by broadband subscribers. In other words, while Chart 24 does not show data consumption above roughly 160 GB, that does not mean that typical broadband subscribers do not consume more than that amount each month, just that such subscribers would be excluded by the methodology of the Report. In addition, data consumption within the sample population has increased from the previous study.