The CPD Itanium2 cluster - hardware information

Description of the cluster

The configuration of the Itanium2 cluster is shown below. It consists of 32 compute nodes connected by a Myrinet and a fast ethernet network. All nodes run Rocks 4.2.1/CentOS 4.3. All compute nodes are accessed from the front end machine (cpd.arc.wm.edu). An Apple Xserve-RAID array is directly attached to the fron end and contains the home directories and 2 globally accessible scratch spaces of about 1 TB each.

Diagram of the CPD Itanium2 cluster

Technical details

The following table lists the details of each machine in the cluster:

Table 1: Front end and compute node details.
#	Node name	architecture	processor	memory	disk	Note
1	cpd.arc.wm.edu	HP Itanium rx2600	Itanium2 / 1.3 GHz	4 GB	(see below)	No Myrinet
2	it01,it02	HP Itanium rx2600	dual Itanium2 / 1.3 GHz	4 GB	20 GB	No Myrinet
30	it03 - it32	HP Itanium rx2620	dual Itanium2 / 1.3 GHz	6 GB	20 GB

Table 2 shows the various filesystems that users may access within the cluster. The /home, /scr1 and /scr2 filespaces are all mounted via NFS over the fast ethernet network. Currently a 10GB quota is set for all users /home directory space. /scr1 and /scr2 are to be used as global scratch space. The global scratch space is to be used for temporarily storing large files. Neither /scr1 or /scr2 is backed up and purging is not done automatically. Users with large amounts of data (>100 GB) will be made to reduce their usage when the scratch space starts to get too full (> ~70%).

NO JOBS SHOULD WRITE TO /home,/scr1 or /scr2 DURING EXECUTION You must always use the /lscr filesystem for writing for ALL jobs on this cluster. There are NO exceptions

Table 2: Filesystems on the Itanium cluster
name	size	notes
/home/$USER	480 GB	users get a 10 GB quota
/scr1/$USER	900 GB	Global NFS scratch 1
/scr2/$USER	900 GB	Global NFS scratch 2
/lscr	23 GB	Local scratch on compute node