Best advice: read this sample chapter from Grover's book on mesh restoration. Otherwise, continue with my brief words below.
Objective function to minimize: In comparing different network designs, we normally choose a cost function or heuristic that represents an approximation to the bulk of the costs of building the network. We assume the design of the network to carry working traffic is fixed and the cost of restoration is composed of the cost of separate links allocated for spare capacity which is used to carry traffic on restoration paths after failure. The design restores 100% of traffic for each failure in an explicit set of failure scenarios, normally representing single failure events.
It is critical to note that it is the responsibility of the human planner to ensure that these failure scenarios represent comparable failure probabilities. There are many network transformation steps applied to raw fiber network topology data to generate such failure scenarios -- these are not described here.
If the number of switching nodes is constant and all links represent the same bandwidth and cost, then
LINKS = link_count summed over all edges in the network
is a common objective function in mesh restoration studies. In some cases, though, link cost is roughly proportional to the length of the link, so total
LINK_MILES = link_count*edge_length, summed over all edges in the network
is another common objective function in that case, although it is common to use kilometers rather than miles as units of length in optical networks.
There are many different proxies for cost that are useful in heuristic algorithms for network design. Some, like LINKS, make formulation as a MIP (mixed integer program) easier than others. More accurate models of network cost include modularities which introduce nonlinearities into the objective function that complicate the formulation of a MIP.
Similarly, measures of the efficiency of the resources added to provide network survivability also vary depending on the algorithm and types of costs included in the model. Spare-to-working ratio (of the number of links in the working network and additional spare links) is the most common measure, but there are many different ways to calculate it if the working design has been customized for a particular restoration method. Be aware that the absolute value of this ratio is not always useful in judging the optimality of a particular algorithm; comparisons of this ratio between different algorithms across several networks is a more reliable method.
St*Mesh was motivated by early attempts to model cost modularities more accurately than heuristic or MIP techniques, and in those early attempts the spare-to-work cost ratio was superior even when the (raw) spare-to-working bandwidth ratio was not.