Announcement Banner
UNSOLVED

pfrassino

updated

20 years ago

P

pfrassino

1 Rookie

•

77 Posts

0

350

January 18th, 2007 09:00

NetWorker adv_file on LINUX

Hi all,
I have to put in place staging on a Linux server. Since I'm not a Linux guru, I've a couple of questions to ask you:

1) I've ext2,ext3,reiserfs chances in order to build filesystems to put my adv_files on. Based on your experience, which kind of filesystem is better to use to have best performances in backup?

2) Unfortunately now I can't figure out how many adv_files I'll build, since is not clear how many pools my policies will be based on. Therefore, the only solution to make a LVM volume available for many adv_file is to create a single filesystem with many directory on it, each of them hosting a disk tape for a different pool. It will be a mess in terms of performances ? I think so; what about you ?
As per my previous life in Digital Unix, the Tru64 advanced file system Advfs let me create several filesystems sharing the same LVM volume, dynamically, with no limitations in size for any filesystems ( potentially each filesystems has the LVM volume size; if one of them grows, the others are limited ...); there is something similar available for Linux ??

Thanks a lot for your opinions and suggestions!!!!!!

Pierpa
  • dk3

    163 Posts

    91

    0

    Posted January 18th, 2007 09:00

    Hi,

    I'd prefer a journalled file system as advanced filetype device as the disk partition hosting the backuped files may be quite large. In case of a reset this would result in a faster file system check during boot. This would let the choice between ext3 and reiser. Ext3 would be my favorite in this case because it is a little less cpu-intensive compared to reiser. Perhaps checking out this page may help you: http://www.debian-administration.org/articles/388

    As an answer to your second question: Of course writing to multiple virtual tapes on the same physical disk at the same time will decrease the writing performance of each stream. I would suggest using a RAID technique offerinc a higher writing performance to solve this problem (eg. RAID 0)