Deprecated: Computer Farms

Several computer farms exist that can potentially be used by GlueX collaborators. This page attempts to list the farms and some rough parameters that can be used to gauge their ability. For each farm, there is a contact person who you will need to coordinate with in order to get access. The exception being the JLab farm which can be accessed through the CUE system.

Institution	contact	Nodes	Cores	CPU	Memory	OS	Notes
JLab	Sandy Philpott	240	400	2.66GHz-3.2GHz	500MB-2GB/node	Fedora 8	This is the "Scientific Computing" farm only (there is another HPC farm for lattice calculations.) This is available to anyone with a JLab CUE computer account. However, it is often busy with processing experimental data.
Indiana Univ.	Matt Shepherd	55	110	1.6GHz
Indiana Univ.	Matt Shepherd	768	1536	2.5GHz			This is University-level farm that we can get access to if really needed. From Matt's e-mail: "If we need mass simulation work, we can also try to tap into the university research computing machines (Big Red has 768 dual 2.5 GHz nodes), but these might be best reserved for very large simulation jobs like Pythia background for high-stats analyses."
Univ. of Edinburgh	Dan Watts	1456	1456			Linux	This is a large, high performance farm from which you can buy time. From Dan's e-mail: "We have access to a very large farm here at Edinburgh. We can apply to purchase priority time on the farm or have a current base subscription which schedules the jobs with lower priority (but seems to run fine for the current usage). It is a high-performance cluster of servers (1456 processors) and storage (over 275Tb of disk)."
Glasgow Univ.	Ken Livingston	32 9 26	64 72 52	2MHz Opteron 1.8MHz Opteron 1MHz PIII	1G 16G 0.5G	Fedora 8
Carnegie Mellon Univ.	Curtis Meyer	47	32x8+15*2=286	32 AMD Barcelona, 15 Xeon(older)	1GB/core	RHEL5
Univ. of Connecticut	Richard Jones		360	240 AMD 2GHz 60 AMD 2.4GHz 60 1GHz Xeon	1GB/core	Centos 5	From Richard's e-mail: "This cluster has local users who have top priority, but right now it is really under-utilized. Scheduling is by condor, login on head node only."