Joshua Baker-LePain wrote:
On Wed, 28 Mar 2007 at 11:54pm, Mark Schoonover wrote
I really didn't mean for this to turn into a tape vs. disk based backups debate -- everyone has their own needs and ways of meeting them. But I have to disagree with the following:
Tape backups require user intervention everyday. There's no way around it, especially if you're testing your backups for reliability.
No, they don't. With a library, you only need to change tapes when you're out of slots (if then, depending on your library, dataset size, retention needs, etc). And any decent backup software should have some sort of verification built in or easily scripted.
If you rely solely on your tape software verification to tell you your tapes are 100%, there will come a day, you'll be in for a surprise. The only 100% sure fire way to determine if your tapes are good is to actually restore from them back to a drive, and open them with the applications that created those files. Only after that kind of testing, can you be sure your tapes are good. I've had many tapes verified from the days of Arcserve, through Brightstor, Arkeia, CTAR and BRU. All of these backup software systems ran, and verified flawlessly, then a disaster struck only to find out my verified tapes actually had problems with them.
Now, the $64K is, how many files do you need to test?? :)
Thanks!
Mark Schoonover *** Winner of the 2008 Best Psychic Award IS Manager American Geotechnical - California, Nevada and Arizona V-> 858.450.4040 F-> 714.685.3909 C-> 858.472.3816
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Thu, Mar 29, 2007 at 04:31:45AM -0000, Mark Schoonover wrote:
If you rely solely on your tape software verification to tell you your tapes are 100%, there will come a day, you'll be in for a surprise. The only 100% sure fire way to determine if your tapes are good is to actually restore from them back to a drive, and open them with the applications that created those files. Only after that kind of testing, can you be sure your tapes are good. I've had many tapes verified from the days of Arcserve, through Brightstor, Arkeia, CTAR and BRU. All of these backup software systems ran, and verified flawlessly, then a disaster struck only to find out my verified tapes actually had problems with them.
And how is that any different than any other media, including disks ?
[]s
- -- Rodrigo Barbosa "Quid quid Latine dictum sit, altum viditur" "Be excellent to each other ..." - Bill & Ted (Wyld Stallyns)
Rodrigo Barbosa wrote:
On Thu, Mar 29, 2007 at 04:31:45AM -0000, Mark Schoonover wrote:
If you rely solely on your tape software verification to tell you your tapes are 100%, there will come a day, you'll be in for a surprise. The only 100% sure fire way to determine if your tapes are good is to actually restore from them back to a drive, and open them with the applications that created those files. Only after that kind of testing, can you be sure your tapes are good. I've had many tapes verified from the days of Arcserve, through Brightstor, Arkeia, CTAR and BRU. All of these backup software systems ran, and verified flawlessly, then a disaster struck only to find out my verified tapes actually had problems with them.
And how is that any different than any other media, including disks ?
An online backup system that uses rsync with the --ignore-times option (as, for example backuppc during full runs), will be reading your existing files frequently and re-copying any mismatches detected with the rsync algorithm. Also, the system may offer the option to archive to some other media at convenient times which gives you another chance at it. A problem I've seen with older tape drives was that the heads would lose alignment so that that tapes would only work in the drive that wrote them. You might verify a tape and send it offsite only to find after a disaster that another drive would not read it. I don't know if newer designs have eliminated this problem or not. If not, you really need a 2nd drive to do the verify/read test.
Les Mikesell wrote:
Rodrigo Barbosa wrote:
On Thu, Mar 29, 2007 at 04:31:45AM -0000, Mark Schoonover wrote:
If you rely solely on your tape software verification to tell you your tapes are 100%, there will come a day, you'll be in for a surprise. The only 100% sure fire way to determine if your tapes are good is to actually restore from them back to a drive, and open them with the applications that created those files. Only after that kind of testing, can you be sure your tapes are good. I've had many tapes verified from the days of Arcserve, through Brightstor, Arkeia, CTAR and BRU. All of these backup software systems ran, and verified flawlessly, then a disaster struck only to find out my verified tapes actually had problems with them.
And how is that any different than any other media, including disks ?
An online backup system that uses rsync with the --ignore-times option (as, for example backuppc during full runs), will be reading your existing files frequently and re-copying any mismatches detected with the rsync algorithm. Also, the system may offer the option to archive to some other media at convenient times which gives you another chance at it. A problem I've seen with older tape drives was that the heads would lose alignment so that that tapes would only work in the drive that wrote them. You might verify a tape and send it offsite only to find after a disaster that another drive would not read it. I don't know if newer designs have eliminated this problem or not. If not, you really need a 2nd drive to do the verify/read test.
You might verify a tape and send it offsite only to find after a
disaster that another drive would not read it
This is a very real problem, tape drives with larger drums and wider tapes will suffer less wear and friction on the tape, and to that end will suffer less from azimuth problems due to age. DLT tapes or LTO tape drives will provide better reliability over 4mm and 8mm formats in this area, However modern tape drives, LTO and DLT included have auto azimuth feature in them to overcome these issue.
The acid test is as wisely spoken above is to do a restore. Online media such as disks, generally don't have replaceable media, which means on sealed disk drives, the head that wrote it is the one that reads it, so in that respect it is very different to removable media such as tapes. I bet everyone here has seen a DVD-RW or CDRW that reads and writes ok in the optical drive that wrote it, only to have it choke in a another drive, this is the same problem and its about wear and tolerences...
Regards
Pete
Peter Farrow wrote:
You might verify a tape and send it offsite only to find after a
disaster that another drive would not read it
This is a very real problem, tape drives with larger drums and wider tapes will suffer less wear and friction on the tape, and to that end will suffer less from azimuth problems due to age. DLT tapes or LTO tape drives will provide better reliability over 4mm and 8mm formats in this area, However modern tape drives, LTO and DLT included have auto azimuth feature in them to overcome these issue.
LTO and DLT are fundamentally different than 4mm/8mm formats...
the 4mm/8mm formats are both 'helical scan', there's a pair of single heads on a drum thats spinning very rapidly at an angle to the tape, as the tape moves past the drum, the head describes a diagonal path on the tape. these heads are typically spinning at 1800 rpm or 30 turns/second. head wear is a huge problem, as is the possibility of extensive media damage from a stray particle due to the high speed of these spinning heads. The tape is quite long and rather thin, much like a video tape (in fact, 8mm formats like AIT are based on 8mm video tape, and 4mm DDS formats are based on R-DAT audio tape). To implement this, the tape path is long and complicated (ever take a VHS VCR apart to clean the rollers? its the same thing only much smaller)
LTO and DLT are linear formats, utilizing 'serpentine' travel. LTO, for example, the 1/2" wide tape moves at a linear 5 or so meters/second or so, the head is stationary, and has several 100 tracks on it. the head is mounted on a servo that follows servo tracks embedded on the tape. the tape has as many as 700 tracks (LTO3). the tape is run past the head for several passes to read or write the whole thing, with one set of ~100 tracks read as the tape leaves the cartridge, then the head repositions and reads the next set of tracks as the tape is rewound into the cartridge, repeat this a few times. The tape is thick, sturdy, and fairly short. The tape path is quite simple.
DAT tapes should be discarded after a dozen or so uses, while LTO tapes are rated at nearly 300 uses, and 5000 load cycles. helical scan heads wear due to the very high linear speed, LTO drives are rated at 250000 hours MTBF at 100% duty cycle. the linear formats are enterprise grade
-
John R Pierce -
Les Mikesell -
Mark Schoonover -
Peter Farrow -
Rodrigo Barbosa