<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Arial; font-size: 12pt; color: #000000'><div><br></div>All,<div><br></div><div>In the typical cluster, with a single head node, that node provides</div><div>login services, batch job submission services, and often supports</div><div>a shared file space mounted via NFS from the head node to the </div><div>compute nodes. This approach works reasonably well for not-too-large</div><div>cluster systems.</div><div><br></div><div>What is viewed as the best practice (or what are people doing) on</div><div>something like an SGI ICE system with multiple service or head nodes?</div><div>Does one service node generally assume the same role as the</div><div>head node above (serving NFS, logins, and running services like</div><div>PBS pro)? Or ... if NFS is used, is it perhaps served from another</div><div>service node and mounted both on the login node and the compute</div><div>nodes? Read-only? Is it better to support a shared file space via Lustre</div><div>across all the nodes? </div><div><br></div><div>The architecture chosen has implications ... for instance in the</div><div>common case above PBS Pro would be installed on the head</div><div>node, perhaps in the shared space, and its server and scheduler</div><div>would be run by /etc/init.d/pbs off of the shared partition. The </div><div>bin and sbin commands would shared by the compute nodes.</div><div><br></div><div>In a case where the login service node and the shared file</div><div>space, NFS service node are different, PBS installation</div><div>must be done on the NFS service node in the case where</div><div>the share space is mounted read-only and only the commands</div><div>and man pages would be installed on the login node. What</div><div>are the implications for other user applications the one would</div><div>like to install in the share space for use from the login nodes?</div><div>Some might have write requirements into the installation </div><div>directory? Does this indicate that the NFS partition should</div><div>be mounted read-write on the login node, but read-only on </div><div>the compute nodes?</div><div><br></div><div>Comments and suggestions, particularly from those that</div><div>have set things up on SGI ICE cluster systems would be</div><div>much appreciated.</div><div><br></div><div><br></div><div>rbw</div><div><br><br><br></div></div></body></html>