Try ext4 without journaling
I've found that if I turn off journaling on ext4 on USB flash drives I get a pretty substantial speed boost for writes. Journaling can be disabled with the data=writeback option:
mount -o data=writeback /dev/sdb1 /mnt
It would be interesting to see how a non-journaled ext4 filesystem would fare in these benchmarks.
This doesn't disable the journal, it just allows for data to be written without journaling, and the metadata is still journaled.
Moving data between systems? Not everyone has their LAN set up to sync uid/gids over LDAP or whatever.
Originally Posted by BlackStar
Dbench results for btrfs pulled a funny. Can someone tell me why the speed went up?
I would have liked to have seen ext2 and ntfs
I wish they had specified what options were used, as there are several journalling options for the journalled fs used, and I am too lazy to go dig up the options for the os they used. I'd have like to have known if they used the ssd option for btrfs, also, I think Reiser4 (amazingly enough, people are still working on it) with it's wandering logs would have been a nice addition. UDF would be good too, for those of us who want out of the box portability.
I second to that.
Originally Posted by rehaby
Doesn't ntfs use a journal as well? If so then as its hard to disable the journal on ntfs, it would be a poor choice for a filesystem for a flash drive or an SSD
Originally Posted by Apopas
Last edited by DeepDayze; 11-12-2009 at 10:30 PM.
Why did sqlite suddenly perform so well on ext4? sqlite has sucked hard from 2.6.31 and back on ext4.
Is it because this is 2.6.32? or are barriers turned off? or some other mount option? I'd really really like to know.
I have a Class 10 SDHC card and it is slow as hell on Ubuntu 9.04 ext2.
Problems are cron jobs, the apt-database or access to its cache by Firefox 3.5. I've thought it would easily outperform my slow 1,8" hdd but I was wrong.