LinuxQuestions.org
Register a domain and help support LQ
Go Back   LinuxQuestions.org > Forums > Enterprise Linux Forums > Linux - Enterprise
User Name
Password
Linux - Enterprise This forum is for all items relating to using Linux in the Enterprise.

Notices

Reply
 
Search this Thread
Old 06-29-2005, 01:28 PM   #1
dirdej
Member
 
Registered: Jan 2004
Location: Belgium
Distribution: RH 7.2 9.0
Posts: 54

Rep: Reputation: 15
Optimize data flow NFS/LTO-3/SLES9?


I use an HP LTO-3 tape drive to make backups and restores of my data. It is hooked up to a server running SLES9 and I use Bakbone for backup.

Backing up from Server0 to Server1 across the network with the LTO attached runs at about 60MB/s writing speed, however reading from LTO to disk for restore operations is only about 20MB/s.

What can be done to optimize this? Larger buffer sizes? I tried iozone but don't know how to run it in order to test the remote drives. Any suggestions?

On each of the servers I am running 3 Gigabit Ethernet ports bonded in parallel.

Thanks,
 
Old 06-30-2005, 08:37 PM   #2
trickykid
Guru
 
Registered: Jan 2001
Posts: 24,133

Rep: Reputation: 197Reputation: 197
You need to run more checks before assuming its writing at 20MB a second.. Where are you getting this information, from Bakbone software? I'm more familiar with Netbackup Enterprise but in my experience, don't always go by what the backup software tells you its writing at. Also in my experience, its always faster on backup than restore, depending on how you have it backed up, hard drives involved, etc.

You have to imagine, it's also less strain on drives when its reading from the drive rather than writing and restoring to the drive itself. What type of drives are we dealing with here? Is there other loads on these drives? In most scenarios, the faster the backup, the slower the restore. If your in a shop where you depend heavily on a lot of restores, you may want to evaluate how you perform your backups. I've always worked in shops where there were very few restores as the backups were mainly for disaster recovery backups as there were few errors and limitations to users making mistakes in deleting their own data or data on the servers, etc.
 
Old 07-01-2005, 12:27 AM   #3
dirdej
Member
 
Registered: Jan 2004
Location: Belgium
Distribution: RH 7.2 9.0
Posts: 54

Original Poster
Rep: Reputation: 15
I can see the time it takes to backup/restore a job. I am doing digital post production of feature films and each part is about 400 GB, usually 5 or 6 parts per film and maybe two or three versions of each in different stages of completion.

Since each frame (= 1 file) is about 20MB each, I can see the count of files restored and have a pretty good idea of the transfer speed.

My disks are SATA RAID5 with 2TB per server. Each server has triple Gigabit ethernet cards in bonding setup. My second server is running SLES9 slower than SLES8 but this should improve with the upcoming SP2.

Some of my backups run at 5MB/s others at 60+. I want to know why (besides the IP bonding).

Thanks
Dirk
 
Old 07-01-2005, 08:55 AM   #4
trickykid
Guru
 
Registered: Jan 2001
Posts: 24,133

Rep: Reputation: 197Reputation: 197
Quote:
Originally posted by dirdej
Some of my backups run at 5MB/s others at 60+. I want to know why (besides the IP bonding).
You have to consider tape seek, rewind time and so on.. It's the same in Netbackup, it will jump from 2MB/s to 50MB/s and so on.. Restores are always slower than the actual backups.

To find your bottleneck though, you'd have to perform individual tests. Test by doing it on this host and possibly on another test host with and without the bonding, etc. And again, never judge your speed by the backup software as most can use cache compared to what is actually being written to disk, etc.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Utrium 2 LTO Tape: space used/available? JMCraig Linux - Hardware 0 09-05-2005 07:10 PM
Constant Data Flow oberon1879 Linux - Networking 1 03-03-2005 12:23 PM
How to install SLES9 via nfs server? hemings Suse/Novell 0 10-14-2004 03:05 AM
Measuring Data Flow/Rate Bicco Linux - Networking 0 08-21-2003 07:50 AM
NFS lockup on data server Paul_assheton Linux - Networking 4 08-30-2002 07:48 AM


All times are GMT -5. The time now is 12:03 AM.

Main Menu
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
identi.ca: @linuxquestions
Facebook: linuxquestions Google+: linuxquestions
Open Source Consulting | Domain Registration