Limitations of scaling switching capacity by interconnecting multiple switches

Carriers are often faced with the need to scale the switching capacity of a central office site by installing multiple switches at that site. In this paper, we first provide a strong argument that multiple colocated switches should always be interconnected together because of the uncertainty in future traffic requests. However, this results in an inefficient use of a portion of the switching capacity as a number of ports are used to interconnect the switches together (the so-called "interconnect penalty"). We develop and apply a quantitative model that analyzes the interconnection capacity required to achieve certain performance criteria as a function of traffic uncertainty. We also derive bounds on the interconnection capacity required between multiple switches for different interconnection approaches and traffic patterns. The practical implication of this work is the realization of the need to eventually deploy an additional layer of higher-capacity switches of higher switching granularity in the context of optical networks to handle high bandwidth services and relieve the capacity needs of lower granularity switches.