The demand of data traffic bandwidth is ever increasing. Optical networking technologies hold the promise of meeting this ever increasing demand. In the 1990's, we saw tremendous progress in DWDM (Dense Wavelength Division Multiplexing) technologies through which more than 100 wavelength channels can be multiplexed onto one single strand of fiber to provide an aggregate data rate of hundreds of gigabits per second. Not stopping at point-to-point DWDM sytems, scientists have kept coming up with wavelength switching and control technologies for all-optical networks where bandwidth allocation can be done dynamically, flexibly, and delay-free. Note that on top of optical technologies, control technologies and intelligent routing are required to realize this dream....[ Read more ]

The demand of data traffic bandwidth is ever increasing. Optical networking technologies hold the promise of meeting this ever increasing demand. In the 1990's, we saw tremendous progress in DWDM (Dense Wavelength Division Multiplexing) technologies through which more than 100 wavelength channels can be multiplexed onto one single strand of fiber to provide an aggregate data rate of hundreds of gigabits per second. Not stopping at point-to-point DWDM sytems, scientists have kept coming up with wavelength switching and control technologies for all-optical networks where bandwidth allocation can be done dynamically, flexibly, and delay-free. Note that on top of optical technologies, control technologies and intelligent routing are required to realize this dream.

In a DWDM network, the data rate of an individual wavelength channel is typically in the gigabit-per-second range. In this thesis, we consider an optical network in which an optical channel can be subdivided in time into time slots. We call such a network a Photonic Circuit Switching (PCS) network. PCS is analogous to Optical TDM as has been suggested by others to be one of the feasible network proposal for the future. As opposed to optical burst switching, PCS does not require header processing of each "opitcal" packet. Rather, time slots are switched based on their positions within a Time Division Multiplexed (TDM) frame as in conventional TDM circuit switching in the electrical domain. However, there is one major problem for PCS - there is no RAM-like buffer available in the optical domain, and FDL (Fiber Delay Line) seems to be the only choice for dealing with buffering in the near future. FDL is expected to be far more expensive when compared to RAM, implying that it must be used highly efficiently.

Therefore, in this thesis, we propose and analyze a number of buffer-efficient Routing and Time-slot assignments (RTA) algorithms for determining routes and assigning time-slots for photonic circuits. Our aim is to minimize the buffer usage in the whole network. In this thesis, two categories of RTA algorithms are investigated, namely, the coupled scheme and the decoupled scheme. The coupled scheme determines path and time-slots simulataneously, while the decoupled scheme chooses the path first, and then afterwards assigns the time-slot. Performances of different schemes are evaluated and compared.

An efficient connection management scheme for setting up and tearing down connections is essential for the effective operation of a PCS network. In this thesis, we investigate how the GMPLS architecture with RSVP-TE as the connection management protocol can be adopted for connection management of the PCS network. In a dynamic network where the network status changes frequently, modifications to the RSVP-TE are needed. In addition, extensions and modifications required to improve the buffer efficiency will be proposed in this thesis.