ROMA
Resource Optimization: Models, Algorithms, and scheduling
The ROMA team aims at designing models, algorithms, and
scheduling strategies to optimize the execution of scientific
applications on High-Performance Computing platforms. More
specifically, ROMA is interested in obtaining the “best” possible
performance from the point of view of the user (e.g., application
execution time) while using ressources as efficiently as possible
(e.g., low energy consumption). The work performed by ROMA ranges
from theoretical studies to the development of software used daily
in the academic and industrial worlds.
The ROMA team is devoted to the study of the fault tolerance, the
energy consumption, and the memory usage of scientific computing
applications executed on clusters and on supercomputers, with a
special interest in direct solvers for sparse linear systems. The work
of the Roma team will be organized along the three following research
themes.
- Resilience. In this theme, we focus on the
efficient execution of applications on failure-prone
platforms. Here, we typically address questions such as: Given a
platform and an application, which fault-tolerance protocols
should be used, when, and with which parameters? Related to this
problematic, is the optimization of the execution of
applications whose behavior is described through some
probability distributions: in both contexts, optimization
problems must be solved in probabilistic settings.
- Multi-criteria scheduling strategies. In
this theme, we focus on the design of scheduling strategies that
finely take into account some platform characteristics beyond
the most classical ones, namely the computing speed of
processors and accelerators, and the communication bandwidth of
network links. In the scope of this theme, when designing
scheduling strategies, we focus either on the energy consumption
of applications or on their memory behavior. All optimization
problems under study are multi-criteria.
- Solvers for sparse linear algebra and related
optimization problems. In this theme, we work on most
aspects of direct multifrontal solvers for linear systems,
usually in the scope of the MUMPS solver that we co-develop. We
also work on combinatorial scientific computing, that is, on the
design of combinatorial algorithms and tools to solve
combinatorial problems, such as those encountered, for instance,
in the preprocessing phases of solvers of sparse linear systems.
In addition, we also work on dense linear algebra: we focus on
the adaptation of factorization kernels to emerging and future
platforms.
The ROMA team is one of the two teams that replaced the former GRAAL team.
Contact: Loris Marchal (Loris.Marchal at ens-lyon.fr)