ENEA & EGEE/EGI/IGI: the Gateway Approach
ENEA partecipation in EGEE/EGI/IGI has been focused on the interoperability
between ENEA-GRID
and EGEE/EGI/IGI infrastructure with the target to
provide access to AIX based computational resources.
To overcome the
problem of the unavailability of the EGEE middleware
for the AIX worker nodes a gateway solution has been implemented and it
is used to provide access to 96 AIX WN in the EGEE production
infrastructure. The solution has been successfully tested also for
other UNIX like platforms, as IRIX, Mac OS X, Cray and Altix
systems and it presentely in production also for the CRESCO HPC system (x86_64).
The gateway implementation is based on SPAGO (Shared Proxy Approach for GRID Objects).
This approach is based on the utilization of a proxy worker node that
shares a file system with the CE and the real workers nodes. The GRID
middleware commands are wrapped with a remote execution command (as ssh
or lsrun, in the LSF evironment) so that when the real worker
nodes execute them, the effective execution takes places on the proxy
WN.
In the case of the interoperability between ENEA-GRID and EGEE/EGI/IGI
SPAGO has been implemented making use of the two main component of
ENEA-GRID "middleware", LSF and OpenAFS, but the same
architecture has been also implemented also using other shared
file systems (NFS, GPFS) and resource managers, as PBS.
A full description of the ENEA gateway architecture and its components can be found in:
The utilization of OpenAFS has required the modification of the gssklog
package to add the support of the VOMS features. A small modification
also to the LCMAPS package has also been necessary.
The modifications to gssklog and to the LCMAPS packages are described in:
EGEE Technical Note EGEE-TR-2006-006
"AFS Pool Account Users - GSSKLOG and LCMAPS extension to
support AFS users as EGEE pool account users" Bracco, G;
Giammarino, L; Migliori, S; Sciņ, C. ; available as pdf document
at CERN http://doc.cern.ch//archive/electronic/egee/tr/egee-tr-2006-006.pdf or as local copy