Use of Spare Capacity to Support Local Traffic

Traffic data and models for both metropolitan and wide-area networks indicate a significant degree of locality for certain types of traffic.  The model proposed by Dwivedi and Wagner [1], for example, uses a voice component inversely proportional to internode distance and a transaction data component inversely proportional to the root of internode distance.

We have presented a new method for providing link recovery in mesh networks. Our method allows for rapid recovery, easy extensibility when new links are added, and distributed operation. Moreover, we can assign the capacity in secondary fibers, which are generally reserved to provide backup capacity in case of failure, to carry unprotected traffic. This reassignment of capacity allows for more efficient use of capacity in networks. The other main approach to providing rapid link failure recovery in mesh networks is through networks of rings. We compared our method with one type of cycle cover approach and showed that algorithms based upon our new method for link failure recovery afford significant advantages in terms of spare capacity. However, these gains in spare capacity for unprotected traffic do not come without drawbacks. Even though for primary traffic a single failure on any link is recoverable, robustness and reliability measures for failures of two links are adversely affected.
 

In conventional high-speed systems, only switching nodes perform recovery. The purpose of recovery for a fault, say between nodes A and B, is to re-establish a connection between nodes A and B. If access nodes are placed along the link [A,B] and the method of recovery is not changed, then a fault will disconnect access nodes along the link [A,B], even though those nodes remain physically connected to the network by the remaining portions of the link [A,B]. The figure below illustrates the effect of link-based restoration using loop-back without taking into account the presence of access ports. Note that the access ports become isolated from the rest of the network. In part b of the figure, the switching nodes still perform loop-back recovery, but they take into account the presence of access nodes to maintain connectivity with the access nodes. The access ports, in turn, will maintain connection with the rest of the network, using directions that depend on their location with respect to the fault. The networks considered there are composed solely of switching nodes, so that recovery of the type shown below is sufficient. We propose to extend this very general type of network recovery to allow for the presence of access ports.

Conventional link-based restoration and link-based restoration in direct access networks. Large nodes are nodes with switches and smaller nodes are access ports. Full lines represent fibers and dashed lines represent connections, e.g. wavelengths. Short dashes show the original connection and long dashes show the backup route or restoration route. The original connection is shown from D to C, with possible access from the access nodes to that connection.