Overview
RAH (Reda, Azem, Hamwi) is a parallel filesystem that runs on a hetergeneous cluster of PC computers
(Beowulf cluster). It runs under the Linux kernel, but can be easily ported to run on other operating systems.
This allows one to use RAH in a cluster of networked workstation running different operating systems,
a common environment that is usually found in university labs.
What does the term parallel filesystem imply?
A parallel filesystem should satisfy the following conditions
- The filesystem should make use of multiple storage units
- Multiple physical paths to the storage units must exist
- The parallel filesystem should be optimized for high speed concurrent read/write access
Parallel filesystems (such as RAH) should not be confused with distributed filesystems (for example, NFS). While
distributed filesystem often distribute file on several hardware nodes, they are usually intended for general use,
and often, not optimized for high speed concurrent access.
How does RAH acheive its system and architecture independence?
It uses a system indpendent middleware and encoding. Specifically, it uses the PVM (Parallel Virtual Machine)
library, a common library used for scientific computing and parallel applications.
Publications
- Khairi Reda, Maher Azem, Yazan Al-Hamwi: Design and Implementation of the RAH Parallel File System, July, 2004.
Download (Arabic only).
Last update: August 15, 2006