DEEPFLOW: Alpha/Linux Cluster

DeepFlow is a parallel computer made of inexpensive Alpha workstations connected with an ethernet switch. This distributed memory system allows for very fast processor-to-memory data transfers, as each processor can access its own, dedicated, high speed memory bus and cache memory. At a glance, DeepFlow holds 40MB of distributed cache memory, 17GB of distributed RAM accessible at 28 GB/sec (peak), and is capable of 21 GFlops (peak).

DeepFlow has been build up by Profs. R. Keunings & V. Legat rheology group with a view to enabling kinetic theory models-based numerical simulations of time-dependent, complex viscoelastic flows. It is also accessible to other research groups and runs a wide variety of applications. These include image processing and simulations of turbulent flows. Finally, students may connect to DeepFlow in the framework of their parallel computing course.

DeepFlow is currently managed by the Applied Mechanics Division in collaboration with UCL's High Performance Computing Centre.

• Technical specifications

• Available Software and Libraries

• Login, account setup, and hard disk space

• Writing and compiling your code

• Running your code with the Portable Batch System

• How to contact us

Last update : March 4, 2002
Page Master : A. Leygue
Web Master : E. Mulemangabo