PIAF - Parallel PAW for high I/O nTuple analysis
PIAF was designed to allow PAW users easy access to central compute resources (CPU and storage)
The central resources initially consisted of 8 powerful HP workstations in the CERN CC, equipped with RAID disks and an interface to the CC tape vault subsystem.
With this setup, using a clever data caching scheme, remarkable speedups of nTuple queries were achieved.
It had the advantage of providing speedy analyses of large, dedicated data sets.
Latterly, the software has been ported to the IBM SP2 and Meiko CS2.
However, the system is clumsy, hard work to operate and manage, inflexible, and outdated. It will be phased out at CERN over the next year or so ...