BaBar computing at RHUL revolves around a SUN machine (babar1.pp.rhul.ac.uk), that is an E250 solaris server composed of two 400 MHz Sparc II processors with 1 GB of memory and 0.5 Tbyte of RAID disks, and a PC farm, composed of two 800 MHz Pentium III double processor front end machines (bfa.pp.rhul.ac.uk and bfb.pp.rhul.ac.uk) and 40 800 MHz Pentium III double processor slave nodes used only for running batch jobs (bf01-bf40). The PC farm has 380 Gbyte of disk available and also shares the 0.5 Tbyte of RAID array with the SUN system.
The system on babar1 and bfa(b) has been configured with BaBar computing in mind. The intention is that it should be optimised for code development, processing of small quantities of data and, with the arrive of the new PC farm, for the production of MC events to be used by the whole BaBar collaboration. In addition the aim has been to make it as easy as possible to manage. With this in mind the decision has been made to NOT make offline releases available locally. Instead they are available via AFS from RAL.
This has definite advantages from the system management point of view, but for it to work three requirements must be met:
If any of these requirements can not be met, then it will be necessary to support releases locally. We'll have to wait and see on that one...
Another wait-and-see issue is how the RAID disks should be partitioned. Separate partitions will be required for the AFS cache, event store and user scratch areas. Currently these have been assigned to directories on the disks in an ad-hoc manner. When the disk requirements are better understood they should be moved into separate partitions.