Chapter 3.8.2 - WASPNET

3.8.2 WASPNET

The WASPNET (WAvelength Switched Packet NETwork) project [59] is an
EPSRC (Engineering and Physical Sciences Research Council) funded collaboration
between three British universities: the University of Strathclyde, Essex University,
and Bristol University, supported by a number of industrial institutions.

Figure 3.12 depicts the optical packet switch architecture proposed in
WASPNET. The WASPNET architecture utilizes wavelength selective routers
(such as arrayed waveguide gratings, AWGs) combined with feedback delay
lines. Wavelength routers exhibit low loss and crosstalk and thus superior system
performance; meanwhile, feedback delay lines allow the implementation of multiple
packet priorities [59]. The switch has N input and output ports and includes 4N

FIGURE 3.12 A single plane of the wavelength routing switch with feedback delay lines based on arrayed waveguide grating suggested in WASPNET.

tunable wavelength converters, a 2N × 2N wavelength router, and an N × N wavelength
router. The component count increases linearly with N and not quadratically
(according to N²) as other architectures do. The switch buffer (i.e., the set of
delay lines) is not associated with a particular output and is accessible from all
outputs; for this reason, the architecture is said to implement a shared buffer.

Input packets are initially wavelength converted, so that the first wavelength
router switches them either to the second router, or onto appropriate delay lines
for contention resolution, depending on the delay required for buffering. Half the
first router’s outputs are fed back into its top N inputs after traversing a tunable
wavelength converter and the FDL buffer, which consists of a demultiplexer,
a multiplexer, and a set of delay lines. Each input of this router carries only one
wavelength, as each TOWC that precedes it only allows single wavelength conversion.
Hence the switch controller must ensure that several packets do not leave a
multiplexer in the feedback loop simultaneously; this is achieved by switching
packets to the correct first wavelength router outputs and by storing packets in
longer delays if necessary [59].

Scheduling in the WASPNET switch is performed in two phases; in the first
phase, packets are routed to the correct delay lines to avoid output contention,
and in the second phase, packets are routed to the correct switch outputs. If both
phases are implemented by the same wavelength router, a packet that is preempted
by a higher priority packet upon exiting the FDL buffer can be redirected to a
suitable delay line instead of proceeding directly to the correct output. The architecture
described above can only handle single wavelength input and outputs, but
can be easily modified to handle WDM inputs/outputs. This is performed by
demultiplexing input wavelength channels to separate parallel planes such as the
one depicted in Figure 3.12.

Each packet leaving a plane must be at the correct wavelength to prevent
wavelength contention. To facilitate this, there is an additional wavelength router,
with a tunable wavelength converter on every input. The first wavelength router
sends each packet to that input of the second wavelength router that will carry
it to the correct output, while being at the correct wavelength; hence, multiple
packets may exit on an output of the second wavelength router, each at a
different wavelength. Because each of its inputs can only carry one packet at
once, there is potential for blocking here; in order for the possibility of blocking
to be eliminated, the second wavelength router may be replaced by a many-to-one
space switch [59].

< Previous Excerpt Next Excerpt >

Purchase This Book

TABLE OF CONTENTS

Chapter 3.8.2 - WASPNET

Contact Preferences

This is embarrasing...

Customize Your GlobalSpec Experience

Select Your Free Newsletters

Industry Newsletters

Select Your Free Product Alerts

This is embarrasing...