Today's
commercial optical networks provide extremely limited Fault and
Attack Management (FAM) and, although optical networking is one
of the fastest growing areas in networking, even theoretical FAM
research in the optical domain only scratches the surface. We seek
to close this gap by addressing some of the fundamental research
issues and building a simulation toolkit that will incorporate the
products of this research effort. Ultimately, we wish to provide
thorough quantitative answers to questions about the level of resources
needed to support modern FAM.
Our research
in optical FAM broadly lies in three areas: modeling or how to
model faults and attacks; algorithms or how to devise methods
for detection and management of faults or attacks; and protocols
or how to implement in a distributed manner the various steps
involved in the management of a particular fault or attack.
In the area
of modeling, our research will be driven by questions such as:
is optical fault modeling any different from that of non-optical
networks? How do attack models differ from fault models? How are
coordinated attacks best modeled? What level of network operation
can be abstracted away in a fault or attack model? How can a network
measure and detect denial-of-service attacks? For FAM modeling,
we will build high-level models based on lower-level fault and
attack models at the physical level proposed by others.
In the area
of algorithms, we anticipate providing answers to questions such
as: how should routes be assigned to connections in anticipation
of faults and attacks? Does route assignment in anticipation of
a coordinated attack use significantly more resources? Once an
attack occurs, how long does it take for detecting and localizing
it? Can efficient algorithms for periodic fault or attack detection
be designed?
In the area
of protocols, we take up the practical issue of integrating our
algorithms into a standard. In particular, at the time of a fault
or attack, procedures for recovery need to be reliably implemented
as protocols.
Finally, we integrate our contributions in the above three areas
of modeling, algorithms and protocols by incorporating our solutions
into a simulator. The idea is that a user will be able to select
a fault or attack model from a menu of several, will be able to
select a combination of algorithms from a menu of several, and
be able to evaluate the performance by running the simulator.
We will also make is easy for independent third-parties to integrate
their own modules for testing and comparison. To build this simulator,
we will leverage existing simulation technology and visibility
by using the MERLIN optical network simulator developed at the
National Institute of Standards and Technology. Since MERLIN follows
an open architecture so that various modules are plug-ins, we
anticipate that, by adding an "open" FAM section in
MERLIN, other researchers can propose and test algorithms in addition
to our own.
