Any way to speed up multi-dump loads?

[d***@gmail.com]

I'm having some performance problems with loading multiple dumps from
tape.

The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.

This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.

Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.

There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).

What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

[ThanksButNo]

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

I would suggest that if you need Rapid Recovery Response, acquire
extra hard drive(s) and save your backup dumps to that instead of to
tape. Use the tapes for off-site backup.

[d***@gmail.com]

Post by ThanksButNo

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

I would suggest that if you need Rapid Recovery Response,

What's that?

Post by ThanksButNo
acquire extra hard drive(s) and save your backup dumps to that instead of to
tape. Use the tapes for off-site backup.

We already do dumps to disk as well. The tape backups are for
transfers between disparate sites. Transferring dumps over the network
isn't an option.

[ThanksButNo]

Post by d***@gmail.com

Post by ThanksButNo

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

I would suggest that if you need Rapid Recovery Response,

What's that?

Just made it up coz it has a nice alliteration.

Post by d***@gmail.com

Post by ThanksButNo
acquire extra hard drive(s) and save your backup dumps to that instead of to
tape. Use the tapes for off-site backup.

We already do dumps to disk as well. The tape backups are for
transfers between disparate sites. Transferring dumps over the network
isn't an option.

Acquire removable hard drives. Something you can hot-swap.

[d***@gmail.com]

Post by ThanksButNo

Post by d***@gmail.com

Post by ThanksButNo

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

I would suggest that if you need Rapid Recovery Response,

What's that?

Just made it up coz it has a nice alliteration.

Right.

Post by ThanksButNo

Post by d***@gmail.com

Post by ThanksButNo
acquire extra hard drive(s) and save your backup dumps to that instead of to
tape. Use the tapes for off-site backup.

We already do dumps to disk as well. The tape backups are for
transfers between disparate sites. Transferring dumps over the network
isn't an option.

Acquire removable hard drives. Something you can hot-swap.

Know of any 1-2TB ones?

[ThanksButNo]

Post by d***@gmail.com

Post by ThanksButNo

Post by d***@gmail.com

Post by ThanksButNo

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

I would suggest that if you need Rapid Recovery Response,

What's that?

Just made it up coz it has a nice alliteration.

Right.

Post by ThanksButNo

Post by d***@gmail.com

Post by ThanksButNo
acquire extra hard drive(s) and save your backup dumps to that instead of to
tape. Use the tapes for off-site backup.

We already do dumps to disk as well. The tape backups are for
transfers between disparate sites. Transferring dumps over the network
isn't an option.

Acquire removable hard drives. Something you can hot-swap.

Know of any 1-2TB ones?

Even compressed? DANG!

I did a search, and hot-swappable terabyte drives are available in
clusters of 500Gb units held together in some sort of RAID situation.
Otherwise, it seems clear I can be of no further use with your
problem. SORRY!

[d***@gmail.com]

I'm not getting any response from Sybase people on Sybase forums so
I'm reposting this to Usenet.
I've since tried the "mt fsf 1" mentioned below but the tape still
rewinds. Either Backup Server
always rewinds regardless or I'm not positioning the next dump file
correctly.

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?

[Keith]

Post by d***@gmail.com
I'm not getting any response from Sybase people on Sybase forums so
I'm reposting this to Usenet.
I've since tried the "mt fsf 1" mentioned below but the tape still
rewinds. Either Backup Server
always rewinds regardless or I'm not positioning the next dump file
correctly.

Post by d***@gmail.com
I'm having some performance problems with loading multiple dumps from
tape.
The dumps are large and we have 28 databases on tape. Towards the end
of the loads, its takes about 20 minutes to find the next dump. I'm
assuming backup server is rewinding the tape and seeking even though
the dump is the next one in sequence.
This is with Backup Server 12.5.x on HP-UX 11.x with Ultrium 800GB
tapes.
Is there any way to speed this process up? I reduced the time from 5
hours for full restores to 4 hours by placing the smaller databases at
the start of the tape rather than the larger ones. However, there are
times when we may need to restore just a large database so the 20
minute wait time is a little onerous.
There doesn't seem to be an option to tell backup server not to rewind
(even with a non-rewinding tape device).
What if I issued an "mt fsf" command in between the loads? Would it
check the next tape block or rewind regardless?- Hide quoted text -

- Show quoted text -

It's been a while since I backed anything up to tape but back as I
recall there were typically two Unix / Linux devices for a single
physical connection to a tape drive / jukebox / whatever.... One
automatically rewinds, the other does not. E.g.: /dev/rmt0n where N
is for no-rewind,
and /dev/rmt0.

What device are you backing up to?

Keith

[d***@gmail.com]

Post by Keith
It's been a while since I backed anything up to tape but back as I
recall there were typically two Unix / Linux devices for a single
physical connection to a tape drive / jukebox / whatever.... One
automatically rewinds, the other does not. E.g.: /dev/rmt0n where N
is for no-rewind,
and /dev/rmt0.

That's correct.

Post by Keith
What device are you backing up to?

The non-rewinding device. The What's New doc for ASE 12.5.x states
that backups were sped up by not rewinding a non-rewinding device
during multiple dumps (i.e. dump at the next position).

I doesn't say anything about loads and it appears that Backup Server
rewinds the tape device regardless.

I guess the load database command needs an option to tell it to check
the current/next file on tape before rewinding.

Anyone know how to ask Sybase to make an enhancement?

Dean

[b***@sybase.com]

Post by d***@gmail.com
Anyone know how to ask Sybase to make an enhancement?
Dean

www.isug.org

[Mikhail Teterin]

Either Backup Server always rewinds regardless or I'm not positioning the
next dump file correctly.

Sybase will always rewind a regular dump after dumping to modify its very
beginning -- only a few bytes are changed there, but it does it.

This does NOT happen, when you dump to a plugin such as "compress::..." or
my "pipe::..." (http://libpipe.com/)

You can work-around by either doing modicum of compession (compressing with
level 1 will, likely, be a net speed gain -- a little more CPU time vs. a
LOT of tape I/O):

dump database foo to 'compess::1::/dev/rmt0'

or buy my pipe-plugin (http://libpipe.com/) and do something like:

dump database foo to 'pipe::cat > /dev/nrmt0'

Yours,

-mi