-------- Original Message --------
Subject: Re: [Networker] Most efficient way to create duplicate clones?
Date: Tue, 11 May 2004 14:49:44 -0400
From: George Sinclair <george.sinclair AT noaa DOT gov>
Reply-To: george.sinclair AT noaa DOT gov
To: Darren Dunham <ddunham AT taos DOT com>
References: <200405111827.LAA08044 AT redwood.taos DOT com>

Well, both drives in the SDLT tape library are daisy chained with the
picker, and all three devices attach to the host (storage node) via a
single LVD SCSI cable, so everything is on the same SCSI channel via the
host's LVD dual channel Adaptec card.

Drives 1 and 2 in the LTO library are daisy chained with the LTO picker
and use channel A on their host adapter card. Drives 3 and 4 are daisy
chained and use channel B on the same card. This is also a dual channel
Adaptec card. As near as I can tell, the drives connect directly via a
chain.

The primary server and the storage node are not connected and can only
communicate via the network. They share no peripherals, cabling, etc.
Furthermore, the library on the primary server is connected only to the
primary server. So, obviously the network would have to be involved when
cloning data from primary server's library to one of the libraries on
the storagenode, and I think this was the slowdown I was remembering and
was confusing with the speed between the LTO and SDLT which should not
be a major factor.

There could be some speed issues between LTO and SDLT, or vice versa,
but I'm thinking it should not be a major problem and would certainly be
wayyyy faster than when operating between primary server's library and
storage node's libraries since LTO and SDLT libraries are both on the
storage node.

I guess I just have to decide if the simplicity of one clone pool
outweighs the benefits of having the cloned savesets on two media types
for paranoia factor. Hmm ...

George

Darren Dunham wrote:
>
> > Yes. I mean, each library uses its own SCSI card and respective cables
> > to connect to the storage node server. We have two libraries connected
> > to this storage node server: the Storagetek LTO and the ATL P1000 SDLT
> > library. We also have an older ATL P1000 DLT7000 drive library that's
> > attached to the server. Perhaps I was thinking of the slow times
> > involved in cloning tapes from the server's library over to the LTO or
> > SDLT? Clearly, in this case, you're dealing with a tremendous drive
> > mismatch in terms of read/write speed, not to mention the network for
> > transferring the data. Guess the network shouldn't be involved when
> > cloning between the SDLT and LTO??? I mean, communication between the
> > server and storage node would use network but data should not be going
> > over network, right?
>
> You say "each library uses it's own SCSI card and respective cables to
> connect to the storage node server".  I don't really care about the
> libraries themselves, I care about the drives.  How do the drives
> connect?  Do they connect directly in a chain to the host, or are they
> connected into some sort of aggregator inside the library?
>
> As long as the drives themselves are directly connected to a host scsi
> controller, the fact that the librares are also connected is
> irrelevant.
>
> > I know I've cloned tapes before from the server's library over to either
> > of the other two libraries, and the times were much, much slower than
> > when going from the LTO to the SDLT or vice versa. Does that sound like
> > the behavior you'd expect in that case? I guess the network would be the
> > main culprit in that case and maybe not so much the drive mismatch?
>
> Is there a network involved?  Without knowing the topology of the drive
> connections, it's impossible to guess.
>
> > We have 2 SDLT drives and 4 LTO drives. My plan before, assuming the
> > savesets on the tapes did not span each other, was something like:
> >
> > clone operation 1: SDLT drive 1 => SDLT drive 2
> > clone operation 2: LTO drive 1 => LTO drive 2
> > clone operation 3: LTO drive 3 => LTO drive 4
>
> I like it.  Because the drives will operate at nearly the same speed for
> the clones.
>
> > Now:
> >
> > clone operation 1: SDLT drive 1 => LTO drive 1
> > clone operation 2: SDLT drive 2 => LTO drive 2
> > clone operation 3: LTO drive 3 => LTO drive 4
> >
> > but I guess it really doesn't matter how you slice it, I'm not loosing
> > or gaining any more drive use by switching back to just one clone pool.
>
> Hmm.  I don't know if you'd have significant speed issues with mixing
> like that or not.
>
> > On the other hand, if I decided to simply clone the clone volumes, then
> > that would make it easy because I wouldn't have to figure out which
> > ssids to clone, since I would want to clone all of them. Having two
> > different clone pools might be nice because then I could clone anything
> > on an LTO to a SDLT and vice versa so that every saveset has both an
> > SDLT clone and an LTO clone to satisfy the "media diversity" factor.
> > Might feel safer having two copies on separate media types?
>
> Good point.
>
> --
> Darren Dunham                                           ddunham AT taos DOT 
> com
> Senior Technical Consultant         TAOS            http://www.taos.com/
> Got some Dr Pepper?                           San Francisco, CA bay area
>          < This line left intentionally blank to confuse you. >

--
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.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=