F This collection contains a virtual disk driver for AXP VMS that allowsF a process to do the storage work for a disk driver. Two host processesE are provided. One of these allows a normal disk file to be the actual F location of stored data (fdhostcry5) with the option of encrypting theB results. The encryption is an XOR with a derived key (derived fromC what is entered, so the entered key isn't used directly) so it runs E pretty fast...only has to do 3 XOR operations on the data at runtime.   A The second is a remote virtual disk over DECnet (see fdremote.doc A and cry2.com as info on these). The remote disk will recover from C temporary network failures by restarting the server end (which runs A where the disk is). Note that the remote server just uses logical ? I/O to read or write blocks to the disk. You can set the DECnet : object up to allow writing or forbid it. That's up to you.  C FQ disks can participate in DEC volume shadowing if you really want A to mess with it...they have the necessary hooks...and can be MSCP A served. Since the fdhostcry5 disks don't have to be on contiguous = disk files, they can be set up on disks too fragmented to use  vddriver with.  B The remote disk can be real handy for remote backup or restore. ItE can be used to "remote" ANY disk (unless the disk does some braindead A operations to act differently with logical i/o than virtual; thus D it won't work with DiskMizer. Use Squash instead...compresses better, and is cheaper and will NOT fail with this.)  ? The timeout for recovering DECnet failures is set by the length > of the retry loop in fdremsub; edit and rebuild if this is not? long enough for you. It works by retrying to start a new server  for a few minutes.  A Executables are provided in case you lack compilers. The programs B DO need privs, and just export basic block I/O. Don't try to mount@ a remote disk r/w in multiple spots...mount it r/o in all places@ but one, since the file activity isn't synchronized. I recommend> you use /nocache on mounts where you'll want to see any remoteB activity quickly from a r/o site. You can put a remote disk on topA of a VD virtual disk to create a r/w disk. Some extension support B exists here to allow limited r/w access from remote sites, so that> backup/record might be supportable, but it is incomplete here.> You should note however that if you DO mount the disk r/w from@ several sites, you CAN do backup/record successfully and withoutE damaging anything (I've tried this a couple times, at least) provided D you mount the volume /norebuild (so volume rebuild can't mess thingsC up) and provided you don't actually do anything else that might try ? to write to the FQ disk remotely. Most deadly are anything that D does create, delete, extend, or truncate, but you MUST be aware that@ no activity should be done. Doing a PRIVATE mount for the use ofD the backup process (i.e., MOUNT/NOREBUILD FQAn: instead of something= like MOUNT/SYS/NOREBUILD) will give a little protection since < random folks other than the backup process will at least seeC the disk as privately mounted & allocated. /NOCACHE is good to have 
 here too.)  C When backing up via this method, remember no locks are known at the > remote site, so implicitly there's an /ignore=interlock on any? backup. Doing back/record is not heavily tested in this mode (I @ have done backup/since=date often though) so be a shade careful.> One thing about this is that at least if record should cause a> file header operation to be lost, a backup of the file exists.> The record pass of Backup does writes to the index file to set> the backup date. The potential problem that can arise is if itB should do its r/w cycle and some other activity that writes to the= header gets in between, like setting a date or writing a file I retrieval pointer in the course of extending the file. The probability of A this is low (could be zero if this happens at 3AM and the disk is 1 idle) but you need to understand what's going on.    Glenn Everhart Everhart@Arisia.GCE.Com 