Grid Engine Usage
This information is not being updated!
Grid Engine Overview
The Grid Engine (former Sun Gridengine) is an open source batch-queuing system typically used on a computer farm or high-performance computing (HPC) cluster and is responsible for accepting, scheduling, dispatching, and managing the remote and distributed execution of large numbers of standalone, parallel or interactive user jobs. It also manages and schedules the allocation of distributed resources such as processors, memory, disk space, and software licenses.
The cluster @IMP/IMBA is organised as a grid-like cluster (not a supercomputer cluster!). This means we have loosely coupled nodes in various hardware generations, but all configured the same on the software level. It consists of the following hardware:
|nodename||amount of nodes||cpu(s)||cores/cpu||cpu speed||memory/node||total cores||notes|
|compute-hp-*||3||2||4||3GHz||16GB||24||Special nodes with Intel processors|
|bioinfo-apps||5||4||1||2.2GHz||32GB||20||Old application servers e.g. albert, waibl, ...|
|cla*||7||2||2||2.6GHz||8GB||28||Tim Clausen servers shared with the cluster|
All nodes are from Sun or HP and equipped with AMD64 CPU's except the ones mentioned above.
Additionally we have 3 nodes reserved for the Grid Engine master. Depending on the failsafe operation/takeover it should be running on one of them:
- aragon, rainer, hutter
The current load of the cluster is monitored by the Ganglia software and can be seen on the website: aragon
Notes on Queues at IMP/IMBA
Some compute clusters have queues like highmem.q or lowpriority.q etc.. As we found out, this does not work well for the diversity of the programs used at our institutes.
Therefore we have three types of queues:
- application specific queues for very special software requirements (e.g. Imagic, Xmipp, ...).
- group specific queues where the assigned hardware suits best for the programs typically used by these groups (e.g. stark.q, bioinfo.q, ...).
- a special unlimited nonofficehours.q, which enables nearly all cluster resources for all users without any restriction during night, weekends and public holidays.