We were a big user of HSM on AIX and we decided instead to purchase
more
disk space rather than use HSM. The reason was that we were using HSM on
filesystems which were exported via NFS to other servers.  The problem, like
Richard pointed out, is that if took more than a few minutes to get the file
back, our NFS server would hang.  We tried to resolve this by increasing the
number of nfs daemons running on the system.  This did help but it wasn't
good enough.  In my opinion HSM works well in a controlled environment with
very large files, which applies to all of us, right?!  Good luck


Marco

P.S.  I would like to thank everyone that responded to my question
regarding memory sizing. 1 Gb of memory per server is on order.