DLT has the potential for very fast retrieval. DLT stores data in a "serpentine" pattern. What that means is that the tape passes over the heads while data is written to two of the 128 tracks. When it reaches the end of the tape, it switches to two other tracks and reverses the tape. The tape passes over the heads 64 times before filling up the cartridge (DLT7000s write to 4 tracks at a time). On a retrieval, the heads can be positioned at the specific tracks that contain the desired file, eliminating the need to serially read the entire tape. Clearly, the STORE/RESTORE software for the HP3000 does not take advantage of this capability. Hopefully, as more customers move to DLT, HP and the other backup software vendors will enhance their software to use DLT "intelligently". >>> "Gale, Dave" <[log in to unmask]> 06/17/98 05:45pm >>> No waste of bandwidth here. The truth be known, I have my own attitudes twards backups. Some of my experiences are: NEVER place printers or other dust emitting equipment in the same room with your DAT drive. This caused us to replace the mechs about every 6 months. Despite the speed hype of DLTs, it takes us over 3 hours to restore files (even small ones) from a strip set. They deffinetly hold more data but access speed to specific data is sllooww. And I who heartedly endorse using VSTORE as part of tape validation. Just my opinion, with some experiences from the school of hard knocks. ---------- From: Stigers, Greg ~ AND To: HP3000-L Subject: Request for comment on backup procedures Date: Wednesday, June 17, 1998 12:26PM Sorry for the waste of bandwidth, but I am in a win-lose situation, where the loser is liking to have their competancy questioned. Also, I imagine that in this context, our proposals raise questions of liability if these changes are made for us, but not for others. I am interested in just criticisms of the following, both any inadequacies in what I propose, and in the current procedures as described. I have requested that our data center verify our weekly full backups, which are performed on Sundays. They are questioning this, and suggesting that if we question DAT integrity, we move to a technology that we do trust. I am also told that no one else regularly verifies backups along these lines. The details of the request are to split the backup across our three available DAT / DDS drives, along the lines of how we have split the daily backups, in the interest of backup up and adding verification in a smaller window of down time. I have specifically requested that we not 'strip' the backup in parallel across the three tapes, because in that scenario, if one tape ever has problems, then the entire backup (with the possible exception of files so small as to fit on one strip on one tape) is unusable. This contrasts with three seperate backups that backup all files; then, if one tape goes bad, only that tape is lost. I have asked that the VSTORE version of the STORE software be used against the weekly full backups to confirm that what is on the tape is what is on disc. Also, I have asked that due to well-known issues with DAT calibration, the tapes be rotated among the three drives before the VSTORE, so that we are reasonably sure that a tape made on one drive is readable on another. This does leave us suceptible to drift over time, where an older tape can no longer be read due to drives drifting out of calibration. Perhaps the rotation should alternate direction (A->B, B->C, C->A one time, then A<-B, B<-C, C<-A the next), to reduce the likelihood of this. While moving to DLT is probably well advised, doing so would almost certainly take longer than our data center requesting that their tape library provide three tapes for full backup and implementing a procedural change. In any case, we would merely want variations of these procedures using a more robust technology.