|
|
The MetaNEOS Project
MetaNEOS is a collaboration between computational optimization
researchers and the Condor and Globus metacomputing teams.
Fundamental goals are to
- Use the power of metacomputing platforms to solve very large
optimization problems cheaply.
- Demonstrate that networked resources can be used in scientific and
commercial settings to solve hard problems in optimization,
operations research, supply chain management, without capital
investment in large computational servers.
- Make powerful solve engines available to researchers and
practitioners everywhere by using remote solver interfaces
pioneered in the NEOS Server.
Core MetaNEOS project activities include :
- Designing and Implementing enhanced programming interfaces. Basic APIs
for metacomputing platforms do not provide all the necessary
functionality.
- Discovering algorithms that fit the platforms. Algorithms must be
- Opportunistic: Exploit a processor pool that grows and
shrinks during the computation;
- Asynchronous: The operate efficiently despite heterogeneous
processor and variable interprocessor communication times
- Easily checkpointed, allowing them to be restarted if a
processors disappears without warning.
- Implementing solvers for important problem classes --- linear and
nonlinear integer programming, stochastic programming,
combinatorial optimization, global optimization --- and using them to
solve problem instances of unprecented size and complexity.
- Driving the development of metacomputing infrastructure. Optimization
problems are hard for metacomputing, because it is often extremely
difficult to forecast their computational resource requirement.
|
