Aggregation of individual wavelengths into wavebands for their subsequent switching and routing as a single group is an attractive way for scalable and cost-efficient optical networks. We analyze the implications of this waveband hierarchy for a single optical node by analyzing two issues: the proper selection of waveband sizes and the assignment of wavebands for a limited set of input-output patterns of traffic. We formulate a general model and propose optimal algorithmic solutions for both problems. The performance of resulting sets of nonuniform wavebands is studied for several representative cases (a single node, an optical ring network, an optical mesh network). The results demonstrate improved optical throughput and reduced cost of switching and routing when using nonuniform waveband hierarchy.

An alternative approach for the common problem of computing the weights of a general processor sharing (GPS) server is presented. The rate processor sharing (RPS) algorithm of Kumaran and Whiting (see Bell Labs Internal Report, July, 1999), which is based on GPS, is modified in order to remove some of its drawbacks: difficulty in setting its parameters, instability, and a lack of rate adaptivity as the queue state changes. The technique proposed in this paper calculates the basic RPS parameters in an efficient manner so that this scheme take into consideration the time-varying nature of the channel and also be rate adaptive. The main idea combines elements from the latency-rate server theory of Stiliadis and Varma (1995) and the powered earliest deadline first algorithm of Varsou et al. (see Proceedings of IEEE WCNC 2000, Chicago, IL, Sept., 2000), and can be considered as a GPS extension of the latter.

We consider the downlink rate scheduling problem for CDMA networks with multiple users carrying packets with different types of traffic ('voice/audio only', 'data-only' or mixed traffic), each of which has its own QoS requirements that need to be satisfied. At each timeslot we serve the user with the earliest deadline. If his head packet was fully served and there is still power left in the system, we repeat the procedure with the next earliest deadline user, as long as we have power resources. We can thus serve more than one user per timeslot, power permitting. We compare our proposed algorithm to the earliest deadline one, the best scheduling scheme presented by Andrews, Kumaran, Ramanan, Stolyar and Whiting (see Bell Labs Technical Journal, 1999) for similar situations and conclude that we outperform it.