With the upgrade to Sl6.4 we have opted not to upgrade the gridengine based beowulf cluster and this service is now decommissioned. The hardware is now fairly elderly and ECDF/Eddie provides an equivalent service with better resources for high thoughput computing. The recent splintering of the gridengine software with Oracle bringing the standard release back under a closed source license has also made any upgrade path considerably more risky.
The school has maintained a gridengine based beowulf cluster since early 2002, initially as a collection of desktops(dell gx240s) and workstations (Dell WS530s and latterly as a number of rackmounted servers (DELL PE1425s) connected with dedicated ethernet and myrinet switches. We have used gridengine since very early on with the cluster originally being developed using PBS just as MJR were bought by Veridian and PBS was brought back into a closed source license. The supporting filesystem was originally on NFS exported from a single server and has grown into a 4.6Tb GPFS filesystem served off of a cluster of 5 nodes.
The hardware is also used on the Schools hadoop cluster and we’re currently looking into moving this on to newer and more powerful nodes at ECDF at which point the nodes will be disposed off and we’ll make a not insignificant dent in the Schools electricity consumption.