Most service providers guarantee end-to-end service availability through service-level agreements with subscribers. A useful strategy for maintaining service, or restoring service quickly, in the event of network element failure, is to develop a sparing plan that is closely tied to end-to-end availability objectives, from which service-level agreements are derived. This paper proposes a new model for sizing the spare inventory to meet an end-to-end service availability objective. A simple approach is to compute a stockout probability (the probability of not having a spare available when it is needed) for a field replaceable unit, and then use the Erlang C formula to compute the number of spares. This is illustrated using components in a dedicated Internet access network. A second approach takes the cost components associated with spares into consideration. The number of spares is obtained by minimizing the life cycle cost subject to an availability constraint. This is illustrated by sizing the number of optical amplifiers in an optical transport network. The two approaches proposed in this paper are better than the standard approach because the stockout probability is not selected arbitrarily; it is derived from an end-to-end availability objective.
展开▼