Warewulf utilized a VNFSD which was basically a modified thttpd server. Perceus does things a bit differently by defaulting to NFS.

Granted many people have scaling issues with NFS on Linux, and while there are ways around many of them I can understand why this may cause question or concern. The main point to remember is that while Linux NFS may not be the best scaling solution, the problem of shared space must be solved for any cluster. Small to mid sized clusters usually use NFS of some form for sharing the users home file system.