Networker

Re: [Networker] Does cloning unmuliplex and make recovery faster?

2007-01-18 09:31:08
Subject: Re: [Networker] Does cloning unmuliplex and make recovery faster?
From: "Landwehr, Jerome" <jlandweh AT HARRIS DOT COM>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Thu, 18 Jan 2007 09:21:44 -0500
My experience with cloning comes from cloning from VTL tapes to physical LTOII 
tapes

My first attempts had the VTL drive parallelism set more than one, resulting in 
interleaved savesets.  Cloning from this was impossibly slow - cloning is a 
single stream operation and to read a saveset, many virtual tapes had to be 
loaded

Upon speaking to the VTL vendor and NetWorker support (at the time Legato) we 
decided that the preferred method was to eliminate as much as possible 
interleaving of savesets on the VTL volumes (mostly unmultiplexed - parallelism 
equals number of available drives, target sessions=1).  This resulted in much 
faster cloning (100x)

Another factor is that each clone operation will use only one destination 
volume, so to increase total throughput, we submit a few nsrclone jobs 
concurrently.  The lack of VTL parallelism also reduces or eliminates 
contention for the same VTL volume during cloning by different clone jobs.

Just to be precise, the methods we use to clone are the following, automatic 
cloning at savegroup completion and command line cloning specifying a set of 
savesets in a file

All my cloned volumes are unmultiplexed but this may be due to the fact I start 
with largely unmultiplexed source volumes

I find it hard to believe that cloning preserves multiplexing, it's my 
understanding that nsrclone goes one by one, SSID to the next SSID cloning each 
singly before moving on to the next

HTH
Jerry

-----Original Message-----
From: EMC NetWorker discussion [mailto:NETWORKER AT LISTSERV.TEMPLE DOT EDU] On 
Behalf Of Curtis Preston
Sent: Wednesday, January 17, 2007 6:16 PM
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Subject: Re: [Networker] Does cloning unmuliplex and make recovery faster?

Damn it, guys.  This question has been bantered about on the list too many 
times (I'm referring to much more than this thread) for us not to have a 
definitive answer on this question.  I'm going to have to do my own testing and 
prove how this works once and for all.  (I saw someone else's test a while 
back, and that's why I believe what I believe, but I can't find that test.)

To answer your question below, multiplexing would be preserved even if you went 
one big tape to 10 smaller tapes 1/10th the size.  I don't see that tape size 
has anything to do with it.  The backup copy would just stop then start again 
on the next volume, still maintaining multiplexing.

As to your comment on the gui clone, that goes against what I've learned.  If 
you clone a volume, I believe it passes the list of SSIDs to nsrclone, and that 
when that happens, multiplexing is maintained.

Like I said, I'm going to have to do my own testing.

---
W. Curtis Preston
Author of O'Reilly's Backup & Recovery and Using SANs and NAS
VP Data Protection
GlassHouse Technologies


-----Original Message-----
From: networker-bounces AT backupcentral DOT com [mailto:networker-bounces AT 
backupcentral DOT com] On Behalf Of Siobhán Ellis
Sent: Wednesday, January 17, 2007 2:36 PM
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Subject: Re: [Networker] Does cloning unmuliplex and make recovery faster?

The answer is not necessarily, as that is not what we have seen in testing.

I used the GUI to clone a volume from like to like, and it de-multiplexed
it. Clone volume does just that. It works out all the savesets, and then
runs that nsrclone command.

Ask yourself this. No two tapes are not exactly the same length. To preserve
the original multiplexing, how would NetWorker achieve that?

Siobhan


On 18/1/07 6:40 AM, "Werth, Dave" <dave.werth AT GARMIN DOT COM> wrote:

> So basically it looks like when multiple SSIDs are specified in a single
> nsrclone command the relative multiplexing between those save sets will be
> preserved.  You can demultiplex it by specifying each SSID in a separate
> nsrclone command.
> 
> Dave
> 
> David Werth
> Garmin AT, Inc
> Salem, Oregon
> dave.werth<at>garmin.com
> 
> -----Original Message-----
> From: Curtis Preston [mailto:cpreston AT GLASSHOUSE DOT COM]
> Sent: Wednesday, January 17, 2007 11:19 AM
> To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
> Subject: Re: [Networker] Does cloning unmuliplex and make recovery faster?
> 
> There are three possible scenarios of cloned data.  Which one of these
> scenarios happens to you will be based on how you clone:
> 1. A group of SSIDs that were mpx'd get completely de-mpx'd when they
> get cloned.
> 2. A group of SSIDs that were mpx'd stay mpx'd exactly the same when
> they're cloned.
> 3. A group of SSIDs that were mpx'd are partially de-mpx'd when they're
> the cloned.  
> 
> Remember that ALL cloning is saveset cloning, whether you're doing
> volume cloning (nsrclone -V <volume>), saveset cloning (nsrclone -S
> <SSID>), or automatic cloning (group level cloning).  In every scenario,
> nsrclone is given a list of one or more SSIDs to clone.  The only thing
> that determines whether or not something is de-multiplexed is how
> nsrclone was given the SSIDs.  Consider the following scenarios:
> 
> 1. Group level cloning.  Nsrclone is given a list of SSIDs from the
> backup that was just created.
> 2. Saveset cloning where nsrclone is given the name of a file with all
> the SSIDs to clone.
> 3. Saveset cloning where nsrclone is given a list of SSIDs on the
> command line.
> 4. Volume cloning where nsrclone is given the name of a volume to clone.
> 5. Saveset cloning where nsrclone is given only one SSID to clone.
> 
> Assume the following for the moment:
> 1. No SSIDs span tapes.
> 2. No other SSIDs are interleaved in with the SSIDs we want to clone.
> (e.g. no other backups were running at the same time from another group
> that we're not going to clone)
> 
> When scenario 1, 2, 3, or 4, nsrclone is given a list of SSIDs, and it
> clones them all, with mpx intact, and the cloned volume(s) will look
> exactly like the original volume(s).  You can verify this by looking at
> the offsets of the savesets on the tape.
> 
> When scenario 5 happens, however, it will de-mpx the tape, as it's only
> cloning one SSID.
> 
> Now, consider these additional scenarios:
> Scenario 1, 2 or 3 above where the list of SSIDs were interleaved with
> other SSIDs.  When this happens, obviously it's going to de-mpx a little
> bit, as you're not cloning all the same data.  But it's not going to
> completely de-mpx the data.  It's just not going to be mpx'd the same
> way the original was.
> 
> Scenario 4 where some SSIDs span that volume and go onto another volume.
> When this happens, the SSIDs that reside entirely on the volume will be
> mpx'd exactly the same.  However, when it starts cloning SSIDs that
> spanned to another volume, where or not they get somewhat de-mpxed will
> be based on whether or not they were mpxed with other SSIDs that are not
> getting cloned.
> 
> 
> ---
> W. Curtis Preston
> Author of O'Reilly's Backup & Recovery and Using SANs and NAS
> VP Data Protection
> GlassHouse Technologies
> 
> 
> -----Original Message-----
> From: networker-bounces AT backupcentral DOT com
> [mailto:networker-bounces AT backupcentral DOT com] On Behalf Of Fazil Saiyed
> Sent: Wednesday, January 17, 2007 5:56 AM
> To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
> Subject: [Networker] Does cloning unmuliplex and make recovery faster?
> 
> There would have been easy way to find that out.
> Just look at original volumes and the savesets on them, then compare
> with 
> cloned tapes, you will have your answer, however, other post have
> indicated that there in no difference in cloning results wheather auto
> or 
> manual that it DOES Demultiplex. ( baring small saveset belew 100 mb).
> Either way, backing up to Disk and staging may be the cleanest and
> fastest 
> way to clone.
> Thanks
> 
> To sign off this list, send email to listserv AT listserv.temple DOT edu and
> type "signoff networker" in the body of the email. Please write to
> networker-request AT listserv.temple DOT edu if you have any problems with 
> this
> list. You can access the archives at
> http://listserv.temple.edu/archives/networker.html or
> via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
> 
> To sign off this list, send email to listserv AT listserv.temple DOT edu and 
> type
> "signoff networker" in the body of the email. Please write to
> networker-request AT listserv.temple DOT edu if you have any problems with 
> this
> list. You can access the archives at
> http://listserv.temple.edu/archives/networker.html or
> via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
> 
> -------------------------
> This e-mail and any attachments may contain confidential material for the sole
> use of the intended recipient.  If you are not the intended recipient, please
> be aware that any disclosure, copying, distribution or use of this e-mail or
> any attachment is prohibited.  If you have received this e-mail in error,
> please contact the sender and delete all copies.
> Thank you for your cooperation
> 
> To sign off this list, send email to listserv AT listserv.temple DOT edu and 
> type
> "signoff networker" in the body of the email. Please write to
> networker-request AT listserv.temple DOT edu if you have any problems with 
> this list.
> You can access the archives at
> http://listserv.temple.edu/archives/networker.html or
> via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
> 


Siobhán

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the body of the email. Please write to 
networker-request AT listserv.temple DOT edu if you have any problems with this 
list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the body of the email. Please write to 
networker-request AT listserv.temple DOT edu if you have any problems with this 
list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the body of the email. Please write to 
networker-request AT listserv.temple DOT edu if you have any problems with this 
list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

<Prev in Thread] Current Thread [Next in Thread>