Partitioning cubes by time might be your quickest option plus it could increase query performance. The Feb 2004 SQLMAG issue has a great article on partitioning. I would avoid HOLAP if possible unless the relational database is on the same server as Analysis Services or you have a super fast relational database.
I found the below passage at http://support.microsoft.com/defaul...-US;249698
"In situations in which a large data file prevents use of the archive utility for backing up OLAP Services databases, another backup utility, such as Windows NT Backup, should be used to make backups of all files in the OLAP Services Data folder. "
I haven't tried this but it seems logical and it wouldn't require you to reprocess the cubes.
Good luck and let us know what happens