Bacula-users

Re: [Bacula-users] Bacula LTO-5

2017-06-01 10:03:02
Subject: Re: [Bacula-users] Bacula LTO-5
From: Steven Hammond <bcf1689 AT gmail DOT com>
To: Kern Sibbald <kern AT sibbald DOT com>, bacula-users AT lists.sourceforge DOT net
Date: Thu, 1 Jun 2017 09:02:16 -0500
Oh, should I do hardware compression or software compression? Reason I ask, I tried just hardware but didn't seem to get much out of the tape (1.7TB). However, with client side compression, my file server was compressed nearly 80+%. I assume the LTO-5 is set to hardware compression by default (it appears to be enabled when I run tapeinfo). I don't mind the extra time needed for client side compression, but I don't want to slow down the tape drive trying to compress data that is already compressed.

Steven Hammond


On 6/1/2017 7:04 AM, Kern Sibbald wrote:
On 31/05/2017 21:43, Steven Hammond wrote:
I've been reading through some of the articles about settings for my HP LTO-5 drive. I have a question concerning FIXED vs VARIABLE block size.

1. Is it safe (Bacula 7.XX) to set the block size to something other than 64K?
Yes. However, once you have written a tape with a given block size, you cannot changed the block size unless you re-initialized the tape.

2. Does increasing the block size increase the throughput? (2 x 3.6Ghz PROCS - 24 cores, 32GB RAM, 6x600GB 15K RPM SAS, external SAS / HP LTO-5)

Yes. If you are running lots of jobs, it can significantly increase performance.

3. Are there any down sides to increasing the block size? (e.g., existing tapes created)
Yes. The optimal size, IMO, is about 512K. Making the block size much bigger can increase the chances of a tape error. That said there are Bacula users that use 1MB or even large without problems.

On my system I use 256K because it performs very well, and I am not interested to squeeze every tiny bit of performance from the system.


SD configuration looks like:

Device {
  Name = LTO-5
  Media Type = LTO-5
  Archive Device = /dev/nst0
  AutomaticMount = Yes
  AlwaysOpen = Yes
  RemovableMedia = Yes
  RandomAccess = No
  AutoChanger = No
  Maximum Spool Size = 500G
  Maximum Job Spool Size = 100G
  Maximum Concurrent Jobs = 5
  Spool Directory = /var/spool/bacula
}

Despooling speeds avg 60+ MB/s. I also have a quad nic bonded (bond-mode 6 with MTU 9000). I did run the btape test and speed. It's results were much faster.

block size = 64K
btape -c /etc/bacula/bacula-sd.conf /dev/nst0
speed file_size=5 nb_file=8
Zero data / Raw Avg: 280 MB/s
Random data / Raw Avg: 111 MB/s
Zero / bacula Avg: 168 MB/s
Random / bacula Avg: 108 MB/s

O/S: Ubuntu 17.04 (so we could get a later version of bacula without compiling from source)

I'm not overly concerned about the speed (everything backs up over night fine) but I was hoping to get over 100 MB/s seeing that the LTO-5 theoretically can go up to 144 MB/s.
Any tips would be greatly appreciated.

You will need multiple concurrent jobs to see high speed writes. You also need very fast spooling disks.

Best regards,
Kern

Steven Hammond

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users




------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

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

ADSM.ORG Privacy and Data Security by https://kimlaw.us