With our LTO drives in an StorageTek L700 library, the clone tapes write
around 20 MB per second. This allows us to clone around 80 GB or more
per hour.
Also, if you have a tight window, scripts can be used to begin cloning
full tapes before the actual backup is complete.
Bob
-----Original Message-----
From: George Sinclair [mailto:George.Sinclair AT noaa DOT gov]
Sent: Thursday, April 17, 2003 10:12 AM
To: Legato NetWorker discussion; Buehler, Bob
Subject: Re: [Networker] Merits of cloning versus dual backups?
You make an excellent point! I guess I can see that cloning would be
reasonably efficient if you were dealing with incrementals or small
savesets, but when we run fulls, we're talking about a lot of savesets,
some rather large, for a total of maybe 200 savesets or more across as
many as 6 tapes. Now, how on earth would I have time to wait for
NetWorker to read through each of those savesets and clone each one?
Granted, you're not dealing with the network, so you've eliminated that
bottleneck, but still, I mean in a 4 drive box, Networker could only
clone savesets from two different tapes at a time, which would still
leave two more to go. In our other library, it could only clone savesets
from one tape at a time since that library only has two drives, which
would leave one tape to go.
If I consider the amount of time it would take to only back up all the
savesets that run on any one given night during a level full -- we space
our fulls out across a week -- I shutter to think how much longer it
would take if my parallelism was set to 2 on the devices on one library
and 1 on the other. They'd be running all night and all day.
Multi-plexing is the only way we could ever hope to finish our backups
in a reasonable time (our device parallelism is set to 4 on all
devices), but we still space the fulls out across a week. That's kind of
what I see the clone operation as: a backup with a parallelism of 1. How
can you possibly get all those full savesets cloned in an 8 hour window?
I just don't see the speed being fast enough to finish it in time when
it can only do one at a time.
Is there some way to clone more than one saveset on the same tape at a
time? I don't see how since NetWorker can only read one at a time from
the same tape.
George
"Buehler, Bob" wrote:
>
> One advantage of cloning is that you are not using the resources of
the
> backed up machine. You would be with a dual backup.
>
> Bob
>
> -----Original Message-----
> From: George Sinclair [mailto:George.Sinclair AT NOAA DOT GOV]
> Sent: Thursday, April 17, 2003 8:59 AM
> To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
> Subject: [Networker] Merits of cloning versus dual backups?
>
> I've been messing around with cloning, and I would like to hear some
> feedback, criticism and advice about its merits as opposed to just
> creating a second backup tape. Let's suppose I have to create a
one-time
> backup of some static data, and I want to have two tapes in case I
loose
> one or one goes bad. Obviously, I could set up cloning to have the
> savesets cloned to another tape as I back them up to the first.
> Alternatively, I could just maintain two separate backup tapes, maybe
> each being a member of a different pool like full and full2 where
> anything on a tape in the full2 pool is a second backup of its
> counterpart on the full pool. In thinking about this, I just don't see
> how cloning would be the preferred method for the following reasons:
>
> 1. Cloning is slow because it must wait until the backups are finished
> on the source tape before it can begin cloning these savesets to the
> clone volume. Also, it cannot multi-plex the savesets to the clone
> volume. Instead, it must write them one at a time. That could be along
> haul!!!
>
> On the other hand, if you just make two backups, there is nothing to
> preclude you from having the data multi-plexed to the second tape just
> like the first; so, it's way faster and you're not tying up that
second
> drive anywhere near as long as cloning would.
>
> 2. Maybe I'm wrong about this one, but I thought I read that if you
> remove the savesets on the clone you wipe out the counterpart saveset
on
> the original since they both share the same SSID. Vice versa, too.
>
> This is a moot point with two backup volumes as the same savesets
would
> have different SSIDs, making them independent of each other. You've
> never have to worry about some accidentally setting off a chain
> deletion.
>
> 3. For a one-time backup, cloning would have the advantage of not
> listing the same saveset under the saveset recover window more than
> once, whereas making a separate backup tape would cause the same
saveset
> to be listed twice -- could cause confusion -- but given the fact that
> we're talking about static data, it would be obvious that one was a
copy
> since they'd both have the same size and 'details' would show the
> volumes named something meaningful like:
>
> full.001
> full2.001
>
> So you'd know that the one on say full2.001 was most likely a
duplicate
> backup.
>
> I guess the only real advantage to cloning that I see is that it
affords
> you the ability to create a second volume in the event that the
savesets
> on the original volume can no longer be re-created because they no
> longer exist on disk. So, you can't just go back, for example, and
> re-back them up a second time. That might not be an option.
Furthermore,
> even if you could go back and re-back them up, they might have long
> since changed, leaving the second backups of these savesets possibly
way
> different than the first. I guess it all depends on how long after the
> fact you decided to create that second backup tape, but assuming we're
> talking about static data, I really just don't see why cloning would
be
> preferred. I think making a second backup tape in close time proximity
> to the first would be the fastest and most expeditious way to go --
you
> have multi-plexing working for you for goodness sakes -- not to
mention
> the fact that you don't have to mess with the original tape while
you're
> doing it.
>
> How do you guys deal with the slowness?! I mean if a bunch of savesets
> are being backed up to tape, and then NetWorker has to come along
later
> that night, or whenever, and clone those, it's gonna grind through
each
> one one at a time, right? I mean, it can't very well clone more than
one
> at a time since it can only read one at a time. Of course, you could
> clone the entire volume, but that goes one at a time, too, and it
would
> make no sense if the original source tape wasn't full yet. I just
don't
> see why anyone would not instead opt to just create dual backups to
> separate pools as they go. Can someone convince me otherwise, please?
> I'd love to hear that I'm wrong cause I think I might be, and I don't
> want to be ignorant longer than necessary.
>
> Sorry to be so long winded.
>
> Thanks.
>
> George
> George.Sinclair AT noaa DOT gov
>
> --
> Note: To sign off this list, send a "signoff networker" command via
> email
> to listserv AT listmail.temple DOT edu or visit the list's Web site at
> http://listmail.temple.edu/archives/networker.html where you can
> also view and post messages to the list.
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
>
> --
> Note: To sign off this list, send a "signoff networker" command via
email
> to listserv AT listmail.temple DOT edu or visit the list's Web site at
> http://listmail.temple.edu/archives/networker.html where you can
> also view and post messages to the list.
> =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
