Bell Labs: there's a data center coming to a street near you

Bell Labs presents a study on data traffic (free download). It is based on a reference network architecture of metro networks interconnected by a backbone, US end-user traffic trends and certain assumptions.

Bottom line:

• Data traffic to grow by 560% from 2012 to 2017 (5 years). Sources of growth: demand for video (720%) and proliferation of data centers (440%), i.e. cloud traffic.
• Metro traffic growth 2x faster than backbone traffic growth.
• 75% of traffic to stay on metro networks (today: 57%) as a result of more data centers. This is called 'north south' traffic (between end-user and data center), as opposed to 'east west' traffic (between data centers).
• Video caching closer to the end user reduces overall traffic. Bell Labs distinguishes two cases: caching centrally in the metro network, or 'deeper' caching (closer to the end-user), which leads to a (unsupported) 41% reduction of traffic.

Some of the underlying assumptions (partly from Gartner):

• End-user data demand x3.7 (fixed and mobile, consumer and business).
• Pay-TV traffic x2.3.
• Non-traditional pay-TV traffic x7.5.
• Number of metro areas with data centers grows at a 10-20% CAGR.
• Number of data centers in metro areas grows at a 60% CAGR.

On the side: the number 560% probably is incorrect. The report (page 5) clearly mentions growth by a factor of 5.6 (see figure below), which translates into 460% growth (in which case 720% should possibly read 620% and 440% should possibly read 340%).

It remains to be seen how this translates to non-US architectures, but the number of data centers may explode as they are decentralised. What will it mean for ISPs, access networks, on-net traffic and net neutrality?