ZFS FOR LINUX FINALLY!

Chat about Linux in general
Forum rules
Do not post support questions here. Before you post read the forum rules. Topics in this forum are automatically closed 6 months after creation.
User avatar
linuxviolin
Level 8
Level 8
Posts: 2081
Joined: Tue Feb 27, 2007 6:55 pm
Location: France

Re: ZFS FOR LINUX FINALLY!

Post by linuxviolin »

Only 1 of the reasons Ubuntu 10.04 is a Long-Term Sh** (LTS) release, took from the official Release Notes. About ext4:
Performance regressions with ext4 under certain workloads

[…] In particular, the dpkg package manager is known to run significantly slower on ext4, causing installations using the server or alternate install CD to take on the order of twice as long as before. ext4 does not guarantee atomic renames of new files over existing files in the event of a power failure shortly after the rename, and so dpkg needs to force the contents of the new file out to disk before renaming it in order to avoid leaving corrupt zero-length files after power failures. This operation involves waiting for the disk significantly more than it strictly needs to, and so degrades performance. If fast package management operations are most important to you, then you should use ext3 instead. (570805)
Another thing which can be interesting too:

Image

Write performance penalty for using unaligned partitions, took from Linux on 4KB-sector disks: Practical advice (and redrawn for readability). Tests do not reflect overall performance across file systems; a value of 1.00 means no speed penalty; higher values mean worse performance. — I believe that “alignment” is still a hit-or-miss, as the HDD’s firmware reports fictitious Cylinder-Head-Sector values, but it’s interesting to see that the FS algorithms matter. And I love XFS.
Last edited by linuxviolin on Mon Nov 22, 2010 6:54 am, edited 2 times in total.
K.I.S.S. ===> "Keep It Simple, Stupid"
"Simplicity is the ultimate sophistication." (Leonardo da Vinci)
"Everything should be made as simple as possible, but no simpler." (Albert Einstein)
randomizer

Re: ZFS FOR LINUX FINALLY!

Post by randomizer »

Unless you have a fairly recent HDD (definitely from WD but maybe others) you won't have 4K bytes per sector and you won't have alignment issues. SSDs must be aligned correctly as well, else you lose alot of performance and drive lifespan is reduced.
User avatar
tdockery97
Level 14
Level 14
Posts: 5058
Joined: Sun Jan 10, 2010 8:54 am
Location: Mt. Angel, Oregon

Re: ZFS FOR LINUX FINALLY!

Post by tdockery97 »

OK, I'll concede to XFS and Ext4 being in the top 2. But for what I do on a day to day basis, I can't see changing file systems.
Mint Cinnamon 20.1
User avatar
ibm450
Level 5
Level 5
Posts: 650
Joined: Sun Jan 11, 2009 6:56 am
Location: Hamilton Hill, Western Australia

Re: ZFS FOR LINUX FINALLY!

Post by ibm450 »

so in a nut shell


xfs = best

ext4 = 2nd best

?
?
?
?
?
HP EILITE FOLIO 9470M i7-3667u x 4
GitHub: tolgaerok
Image Image
randomizer

Re: ZFS FOR LINUX FINALLY!

Post by randomizer »

ibm450 wrote:so in a nut shell


xfs = best

ext4 = 2nd best

?
?
?
?
?
Profit!
User avatar
linuxviolin
Level 8
Level 8
Posts: 2081
Joined: Tue Feb 27, 2007 6:55 pm
Location: France

Re: ZFS FOR LINUX FINALLY!

Post by linuxviolin »

randomizer wrote:Unless you have a fairly recent HDD (definitely from WD but maybe others) you won't have 4K bytes per sector and you won't have alignment issues.
Well:
Starting in December 2009, hard disk manufacturers began introducing disks that use 4096-byte sectors rather than the more common 512-byte sectors. Although this change is masked by firmware that breaks the 4096-byte physical sectors into 512-byte logical sectors for the benefit of the operating system, the use of larger physical sectors has implications for disk layout and system performance. This article examines these implications, including benchmark tests illustrating the likely real-world effects on some common Linux® file systems. As disks with 4096-byte sectors become more common throughout 2010 and beyond, strategies for coping with these new disks will become increasingly important.
:wink:
tdockery97 wrote:I'll concede to XFS and Ext4 being in the top 2.
ibm450 wrote:xfs = best

ext4 = 2nd best
For me it's rather XFS and JFS, with the noatime parameter, then the others.... I don't want touch to ext3 no more if the distro I use supports another FS, except perhaps for an external hard drive for good compatibility, and ditto for its "improvement" ext4.
K.I.S.S. ===> "Keep It Simple, Stupid"
"Simplicity is the ultimate sophistication." (Leonardo da Vinci)
"Everything should be made as simple as possible, but no simpler." (Albert Einstein)
User avatar
ibm450
Level 5
Level 5
Posts: 650
Joined: Sun Jan 11, 2009 6:56 am
Location: Hamilton Hill, Western Australia

Re: ZFS FOR LINUX FINALLY!

Post by ibm450 »

For me it's rather XFS and JFS, with the noatime parameter, then the others....
could some one shine some light on where to add this line and how and where?

if possible, full directions on how to acheive this as i would imagine alot of noobies to linux would appreciate the guidance opposed to single line answer...


cheers in advance
HP EILITE FOLIO 9470M i7-3667u x 4
GitHub: tolgaerok
Image Image
User avatar
tdockery97
Level 14
Level 14
Posts: 5058
Joined: Sun Jan 10, 2010 8:54 am
Location: Mt. Angel, Oregon

Re: ZFS FOR LINUX FINALLY!

Post by tdockery97 »

OK you win. I finally found the argument that convinced me here: http://stevesubuntutweaks.blogspot.com/ ... ystem.html. I do save and watch a lot of videos, so now I can see a real advantage for ME to use XFS. Thanks for keeping me thinking and searching. :D

Tom
Mint Cinnamon 20.1
randomizer

Re: ZFS FOR LINUX FINALLY!

Post by randomizer »

linuxviolin wrote:Well:
Starting in December 2009, hard disk manufacturers began introducing disks that use 4096-byte sectors rather than the more common 512-byte sectors. Although this change is masked by firmware that breaks the 4096-byte physical sectors into 512-byte logical sectors for the benefit of the operating system, the use of larger physical sectors has implications for disk layout and system performance. This article examines these implications, including benchmark tests illustrating the likely real-world effects on some common Linux® file systems. As disks with 4096-byte sectors become more common throughout 2010 and beyond, strategies for coping with these new disks will become increasingly important.
:wink:
I consider December 2009 to be fairly recent. That's less than a year old.
ibm450 wrote:could some one shine some light on where to add this line and how and where?
Edit /etc/fstab

It will contain a number of lines, one for each of the mount points that are mounted at boot. Near the end of each you'll probably have "defaults" written. You / partition probably has something else about remounting as well. Simply add "noatime" before or after the existing items separating them by a comma and no space. It should look something like below:

Code: Select all

UUID=somereallylongstringofcharacters         /home               ext4         defaults,noatime            1         1
I'm not sure you even need to include "defaults" if "noatime" is the only other option.
User avatar
linuxviolin
Level 8
Level 8
Posts: 2081
Joined: Tue Feb 27, 2007 6:55 pm
Location: France

Re: ZFS FOR LINUX FINALLY!

Post by linuxviolin »

randomizer wrote:I consider December 2009 to be fairly recent. That's less than a year old.
I didn't say another thing...

randomizer wrote:

Code: Select all

UUID=somereallylongstringofcharacters         /home               ext4         defaults,noatime            1         1
I'm not sure you even need to include "defaults" if "noatime" is the only other option.
I guess ext3, XFS, JFS support noatime but not ext4. From Wikipedia about ext4:
Attributes: extents, noextents, mballoc, nomballoc, delalloc, nodelalloc, data=journal, data=ordered, data=writeback, commit=nrsec, orlov, oldalloc, user_xattr, nouser_xattr, acl, noacl, bsddf, minixdf, bh, nobh, journal_dev
And yes, you have no real need for "defaults" with "noatime", but I don't think that hurts something if you keep it.
K.I.S.S. ===> "Keep It Simple, Stupid"
"Simplicity is the ultimate sophistication." (Leonardo da Vinci)
"Everything should be made as simple as possible, but no simpler." (Albert Einstein)
piratesmack

Re: ZFS FOR LINUX FINALLY!

Post by piratesmack »

hmm... My ext4 partition is mounted with 'noatime'

Code: Select all

/dev/sda3 on /initrd/mnt/dev_save type ext4 (rw,noatime,barrier=1,data=ordered)
Wouldn't mount give a 'bad option' error if it didn't support it?
User avatar
linuxviolin
Level 8
Level 8
Posts: 2081
Joined: Tue Feb 27, 2007 6:55 pm
Location: France

Re: ZFS FOR LINUX FINALLY!

Post by linuxviolin »

piratesmack wrote:Wouldn't mount give a 'bad option' error if it didn't support it?
I guess you are right, ext4 must/should support noatime. Wikipedia has all wrong on that. :twisted:

Btw, you must be careful with noatime in some cases. noatime works good for almost all applications but it breaks Mutt or other applications like it because with noatime you do not update inode access times on the filesystem but Mutt needs this information, needs to know if a file has been read since the last time it was modified. For Mutt, you should only use the relatime option.

EDIT. e.g. At http://www.mjmwired.net/kernel/Document ... s/ext4.txt there are the ext4 options, and no noatime... :roll:
K.I.S.S. ===> "Keep It Simple, Stupid"
"Simplicity is the ultimate sophistication." (Leonardo da Vinci)
"Everything should be made as simple as possible, but no simpler." (Albert Einstein)
randomizer

Re: ZFS FOR LINUX FINALLY!

Post by randomizer »

That's interesting. I've seen nobody else claim that using noatime on ext4 was pointless, and I've seen articles using noatime and ext4 showing benefits. Here's one: http://thunk.org/tytso/blog/2009/03/01/ ... erelatime/

It's quite strange that it's not in the documentation though.
User avatar
linuxviolin
Level 8
Level 8
Posts: 2081
Joined: Tue Feb 27, 2007 6:55 pm
Location: France

Re: ZFS FOR LINUX FINALLY!

Post by linuxviolin »

Well, I have read web sites where they talk about noatime and other where noatime was not in the attributes even though they give many... So what? e.g. in addition to the link I gave above (based on kernel version 2.6.34, page generated on 2010-05-31 16:02 EST), at http://www.kernel.org/doc/Documentation ... s/ext4.txt there is no mention of noatime, neither nodiratime, in the many options...
K.I.S.S. ===> "Keep It Simple, Stupid"
"Simplicity is the ultimate sophistication." (Leonardo da Vinci)
"Everything should be made as simple as possible, but no simpler." (Albert Einstein)
randomizer

Re: ZFS FOR LINUX FINALLY!

Post by randomizer »

There's no mention of relatime either.
User avatar
ibm450
Level 5
Level 5
Posts: 650
Joined: Sun Jan 11, 2009 6:56 am
Location: Hamilton Hill, Western Australia

Re: ZFS FOR LINUX FINALLY!

Post by ibm450 »

theres also no mention on whats the best combo to use in these threads.....could some 1 post or advise their personal tried combo that have noticed a performance hike to the boot up times / operational times with these combos.....
HP EILITE FOLIO 9470M i7-3667u x 4
GitHub: tolgaerok
Image Image
Locked

Return to “Chat about Linux”