After the excellent experience of using Warewulf in a previous job (as post-doctoral fellow), I did not have any second thought to deploy Perceus for my new cluster, which is a diskless cluster with 16 nodes. The Caos NSA distribution was not ready yet at that time so I stayed with Centos 5.2, which to be honest I don't have any inconvenient at all, much the contrary. Even if I found Perceus much simpler to administer (one command to rule them all, well almost :) I was a little confused by the fact that some very basic modules (as the ones that handles the ipadrr assignation and host names) are not activated by default. The very first time I booted the nodes, they were not accessible from the master, and after activating the requested nodes I had to reboot them physically. Some of the scripts (which most of them are invoked through the main command perceus) have some minor imperfections with the access rights for non-root users to the perceus databases. However, the dev-team is simply amazing and answers promptly through the discussion list. The new web portal in perceus 1.5, even if it looks a little too simple since it was just introduced in this release, lets you to change the VNFS capsule for a node or to update the description of the node. Anyways, Perceus simply rocks and I keep giving it praise among my colleagues. A colleague just installed for his own beowulf cluster and is extremely satisfied with his results. »