Networker

Re: [Networker] Cloning parallelism

2008-01-24 02:25:19
Subject: Re: [Networker] Cloning parallelism
From: Tim Verbois <Tim.Verbois AT ET.VLAANDEREN DOT BE>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Thu, 24 Jan 2008 08:19:25 +0100
Yaron,

1 drive gives 70 MB/s, the whole VTL has a top speed of 300 MB/s. So when we could read from 2 drives and write it to 1 LTO3 drive, it should increase the speed with at least 30 MB/s for data that is compressable (like file server data, we have a lot of that). For LTO4, that difference would be much bigger, if we could read from 3 or 4 drives, we could increase the speed by a factor of 3 for example.

Tim

Yaron Zabary wrote:
Tim Verbois wrote:
Hello,

Environment:

VTL
L700 (LTO3 + LTO2 drives)
Legato 7.3.2 from SUN
Server + storage node

is it possible to have the parallelism on a cloning set to 2? I changed the setting in the java interface, but it doesn't change a thing.

Why do I want to do that? We clone from VTL to L700 LTO3 tape drives. The speed of an LTO3 drive is 70 MB/s native and can reach 100 MB when compressed. Our VTL can supply data at a speed of 70 MB/s. A client can do a restore at max 50 MB/s. So while we clone at a speed of 70 MB/s compressable data, we lose at least 20 MB/s average speed for cloning. For restores, our tape is to fast, we would be satisfied with an everage restore speed of 40MB/s. The problem we have is that we need the top speed for cloning because of the small time window for cloning (backupping every day and the full weekend, needs cloning at top speeds during all the free time). In a few weeks we start using LTO4, this is going to make the loss of speed even bigger, much bigger. Can this be solved?


I don't see how cloning with parallelism set to 2 will help you here, if indeed your single VTL can only do 70Mb/s. Can your VTL really run at 70Mb/s for each VTL-drive ? As a matter of fact I would expect your performance to drop if you are going to read from two VTL-drive at the same time as these operations tend to interfere each other. It seems that your bottle neck is at the VTL.


--
Tim Verbois
Unix Team
EDS-Telindus
02/553.71.73

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