Multiple models and parallel solving with Mosel

File uploaded by Advocate on Aug 8, 2014
Version 1Show Document
  • View in full screen mode

This paper describes several examples of sequential and parallel solving of multiple models with Mosel. Without

being able to give an exhaustive list of possible configurations, the examples showcase different uses of the

Mosel module mmjobs, such as concurrent execution of several instances of a model, the (sequential) embedding

of a submodel into a master, and the implementation of decomposition algorithms (Dantzig-Wolfe and

Benders decomposition).

From a more technical point of view, topics discussed in this paper include model management, synchronization

of concurrent models, and the use of the shared memory IO driver.

We further discuss the difference between multiple models and the more recent functionality of multiple problems

within a single model. Where appropriate (sequential solving) alternative implementations using multiple

problems are described.

When working with several Mosel models the user may choose to distribute them over several instances of

Mosel. We explain the functionality for remote execution of Mosel models as provided by the module mmjobs

and document the corresponding modifications to the examples.