Algorithmic aspects of high speed switching
Author(s)Mneimneh, Saadeddine S
Massachusetts Institute of Technology. Dept. of Civil and Environmental Engineering.
MetadataShow full item record
A major drawback of the traditional output queuing technique is that it requires a switch speedup of N, where N is the size of the switch. This dependence on N makes the switch non-scalable at high speeds. Input queuing has been suggested instead. The introduction of input queuing creates the necessity for developing switching algorithms to decide which packets to keep waiting at the input, and which packets to forward across the switch. In this thesis, we address various algorithmic aspects of switching. We prove in this thesis, that many of the practical switching algorithms still require a speedup to achieve even a weak notion of throughput. We propose two switching algorithms that belong to a family to which we refer in this thesis as priority switching. These two algorithms overcome some of the disadvantages in existing priority switching algorithms, such as the excessive amount of state information that needs to be maintained. We also develop a practical algorithm that belongs to a family to which we refer in this thesis as iterative switching. This algorithm achieves high throughput in practice and offers the advantage of not requiring more than one iteration, unlike other existing iterative switching algorithms which require multiple iterations to achieve high throughput. Finally, we address the issue of using switches in parallel to accommodate for the need of speedup. We study two settings of parallel switches, one with standard packet switching, and one with flow scheduling, in which flows cannot be split across multiple switches.
Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Civil and Environmental Engineering, 2002.Includes bibliographical references (p. 145-148).
DepartmentMassachusetts Institute of Technology. Dept. of Civil and Environmental Engineering.
Massachusetts Institute of Technology
Civil and Environmental Engineering.