Optical backbone networks are required to be highly dynamic in supporting requests with flexible bandwidth granularities to cope with the demands of new broadband wireless and fixed access networks. To provide this flexibility, services are offered by taking requested bandwidth profile into consideration, instead of assigning a fixed amount of bandwidth to each request. New techniques are developed for the resource management of the elastic optical networks to realize services with a specified bandwidth profile, consisting of minimum, average, and maximum required number of spectrum slots, in addition to holding time. In this work, two new schemes are proposed to realize such services, exploiting a probabilistic spectrum partitioning approach. This new probabilistic spectrum partitioning scheme is devised to enhance the chance of accommodating requests and consequently lower request blocking probability. It enforces different probabilities to contributing spectrum partitions in a certain service realization. Taking advantage of this probabilistic spectrum partitioning and a profile-based routing, we introduce two multistage spectrum assignment methods to make a certain lightpath meet the requested service profile constraints, considering the time-weighted average of the assigned spectrum slots. The results indicate that our algorithms can successfully realize the requests with the probability of 0.993, for erlangs lower than 400.

翻译：光骨干网络需具备高度动态性以支持灵活带宽粒度的请求，从而满足新型宽带无线和固定接入网络的需求。为提供这种灵活性，服务交付不再为每个请求分配固定带宽，而是考虑其请求的带宽轮廓。针对弹性光网络的资源管理，本文开发了新技术以实现具有指定带宽轮廓的服务——该轮廓除保持时间外，还包含所需频谱时隙的最小值、平均值和最大值。本研究提出两种利用概率频谱分区方法的新方案来实现此类服务。该概率频谱分区方案通过为服务实现中不同频谱分区赋予差异化的概率值，提升请求接纳概率，从而降低请求阻塞率。结合概率频谱分区与基于轮廓的路由，我们提出两种多阶段频谱分配方法，通过考虑已分配频谱时隙的时间加权平均值，使特定光路满足请求服务轮廓约束。结果表明，当话务量低于400爱尔兰时，所提算法能以0.993的概率成功实现请求。