US WEST planned fiber rings with a multi-year integer program

By the early 1990s, interoffice fiber-optic networks were already a multi-million-dollar business for US WEST Communications, and demand was expected to expand through its 14-state territory.

The company built a mixed-integer programming model for metropolitan fiber networks. Instead of treating each link as a one-off engineering job, the model produced explicit optimal expansion plans across multi-year horizons.

The model was tested successfully in 1991 on Colorado Springs network and demographic data. A descendant of it was then used to design and plan fiber-optic ring growth in Seattle, Denver, Minneapolis and other major US WEST metropolitan markets.

Plan the ring before digging

Fiber expansion has path dependency: today's trench, ring and capacity decisions constrain later growth.

A multi-year integer program made the cost of future rebuilds visible before construction began.

Using demographic data tied network design to likely demand, not just to existing facilities.

Infrastructure choices get cheaper when the model sees time: a route that is suboptimal this year can be right if it avoids rebuilding the network next year.

The published article reported that an outgrowth of the model was being used for fiber-optical ring networks in multiple major metropolitan areas across US WEST's region.