<<On Mon, 19 Sep 2005 18:31:54 +0200, Frank Bagehorn <[EMAIL PROTECTED]> said:
> How about two more proposals: > - I don't know how it would really perform, but you could create a > fileserver that uses an underlying database. > - Or you create a virtual layer for the existing fileservers, so that you > would have a well-defined interface to whatever lies below. (NAMEI or > INODE or DB or whatever comes to peoples mind.) I recommended (in private mail) using one of the existing "inode filesystem" implementations (effectively *fs without the naming layer) and extended attributes to create a more portable implementation of an inode fileserver. On systems which don't have an inode filesystem, you could fake it using getfh() and fhopen() with an index stored in a database. (The idea here, in both cases, is to avoid the overhead associated with the naming layer in the filesystem.) -GAWollman _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
