31 January 2013 in Computer Networks
We investigate the problem of joint routing and link scheduling in Time-Division Multiple Access (TDMA) Wireless Mesh Networks (WMNs) carrying real-time traffic. We propose a framework that always computes a feasible solution (i.e. a set of paths and link activations) if there exists one, by optimally solving a mixed integer-nonlinear problem. Such solution can be computed in minutes or tens thereof for e.g. grids of up to 4 × 4 nodes. We also propose heuristics based on Lagrangian decomposition to compute suboptimal solutions considerably faster and/or for larger WMNs, up to about 50 nodes. We show that the heuristic solutions are near-optimal, and we exploit them to gain insight on the schedulability in WMN, i.e. to investigate the optimal placement of one or more gateways from a delay bound perspective, and to investigate how the schedulability is affected by the transmission range.
1 December 2011 in Management Science
We present the result of research, developed within Nokia Siemens Networks, to solve the downlink sub-frame allocation problem in Mobile WiMAX (IEEE 802.16) technology in its full complexity, while simultaneously fulfilling real-life constraints on processing power and delay. We describe the IEEE 802.16 standard, and introduce two system models. A theoretical analysis of the two-dimensional packing problems originated by such models shows that they are both 𝒩𝒫-hard in the strong sense. From a practical point of view, the processing budget for scheduling in the base station was estimated to be 1 ms on a state-of-the-art PC. Thus, we introduce two highly efficient heuristics that were developed to handle the system practically. A thorough computational analysis of their optimization characteristics and a system-level evaluation in realistic scenarios proved that the algorithms offer significant capacity gain in Mobile WiMAX systems that translate to increased operator revenues. This paper was accepted by Dimitris Bertsimas, optimization.
1 January 2013 in Computer Networks
9 May 2012 in Real-Time Systems
3 October 2013 in IEEE/ACM Transactions on Networking
In IEEE 802.16, the wireless resources are logically partitioned into 5-ms frames, which extend in two dimensions: time and frequency. To break down the complexity of resource allocation at the base station, a split approach has been proposed in the literature, where the tasks of scheduling packets and allocating them into frames are solved in separate and subsequent stages. In this paper, we focus on the allocation task alone, which is addressed in its full complexity, i.e., by considering that data within the frame must be allocated as bursts with rectangular shape, each consisting of a set of indivisible sub-bursts, and that a variable portion of the frame is reserved for in-band signaling. After proving that the resulting allocation problem is NP-hard, we develop an efficient heuristic algorithm, called Recursive Tiles and Stripes (ℜTS), to solve it. ℜTS, in addition to handling a more general problem, is shown to perform better than state-of-the-art solutions via numerical analysis with realistic system parametrization. Furthermore, an extensive evaluation of the interaction between the scheduler and the allocator is carried out in a wide variety of network scenarios .