Re: [OT?] Re: [gentoo-user] ext3: 10% non-contiguous

2005-11-03 Thread Peter Gordon
Richard Fish said:
> Based on what the developers presented at the 2005 OLS, delayed
> allocation, and an extents-based format (ext4?) are coming:
>
> http://www.linuxsymposium.org/2005/linuxsymposium_procv1.pdf
That looks very intriguing. :-D

Thanks for your thorough explanations, Richard!

--Peter
-- 
gentoo-user@gentoo.org mailing list



Re: [OT?] Re: [gentoo-user] ext3: 10% non-contiguous

2005-11-02 Thread Richard Fish

Peter Gordon wrote:


On Mon, 2005-10-31 at 20:34 -0800, Joshua Schmidlkofer wrote:
 


For a more sustainable situation, switch to XFS [It involved a
backup/format/restore by whatever means you want]  In any case, xfs
has a tool called 'xfs_fsr'  Which means 'file system reorganizer'.
It does defragmentation, and balances some other stuff too.   I run it
weekly on my production servers, and nightly on most of my
workstations.
   



Thus why it is broken by design in my view. A good filesystem should not
need to be defragmented. All filesystems will become fragmented over
time, but a filesystem which is well-behaved should take minimal, if
any, performance loss from it. 
 



The performance impact of fragmentation is unavoidable.  It is a simple 
physical fact that seeking the heads on a disk drive and waiting for the 
platter to spin around to a fragment is a time-consuming operation, 
taking between 6ms (on very fast SCSI hard drives) and 20ms (on very 
slow laptop hard drives).  The average 7200rpm desktop drive has an 
access time of around 12ms.  At that speed, anything smaller than about 
one megabyte can actually take longer to access than it takes to read.  
The only way to avoid the performance impact of fragmentation is to not 
fragment to begin with, which is what most linux filesystems try to do, 
and where Microsoft fails miserably.


The way xfs defends against fragmentation is a bit smarter than ext2/3 
at this point, because it delays the allocation step until the data is 
actually written to disk.  This means that it can wait until the last 
possible moment before deciding where on the disk the file (or fragment 
of a file) should live, and thus should create fewer fragments overall.  
This is assuming that whatever is creating or appending to the file 
doesn't flush after every write of course...


Ext2/3 tackles fragmentation by pre-allocating blocks when a file is 
growing.  It is effective, but would be even more effective with delayed 
allocation.  Note that this pre-allocation is not preserved when the 
file is closed, it is only active for files that are open and growing.


Based on what the developers presented at the 2005 OLS, delayed 
allocation, and an extents-based format (ext4?) are coming:


http://www.linuxsymposium.org/2005/linuxsymposium_procv1.pdf

-Richard

--
gentoo-user@gentoo.org mailing list



Re: [OT?] Re: [gentoo-user] ext3: 10% non-contiguous

2005-10-31 Thread Dale
Peter Gordon wrote:

>
>For what it's worth, I've never had a *single* problem with Ext3, and
>I've been using it with various distributions since I first started
>playing with GNU/Linux a few weeks after Fedora Core 1 was released.
>
>--Peter
>  
>
I use reiserfs and have had no problems either.  I still remember how
bad windoze used to be though.  We would run defrag programs all night
they were so bad.  Even my brother's Win XP with NTFS gets really bad
after a while.  All he does is play games and check email.

I'm going to go see if I can find that thread in the forums. 

I'll be back.  LOL

Dale

-- 
gentoo-user@gentoo.org mailing list



[OT?] Re: [gentoo-user] ext3: 10% non-contiguous

2005-10-31 Thread Peter Gordon
On Mon, 2005-10-31 at 20:34 -0800, Joshua Schmidlkofer wrote:
> For a more sustainable situation, switch to XFS [It involved a
> backup/format/restore by whatever means you want]  In any case, xfs
> has a tool called 'xfs_fsr'  Which means 'file system reorganizer'.
> It does defragmentation, and balances some other stuff too.   I run it
> weekly on my production servers, and nightly on most of my
> workstations.

Thus why it is broken by design in my view. A good filesystem should not
need to be defragmented. All filesystems will become fragmented over
time, but a filesystem which is well-behaved should take minimal, if
any, performance loss from it. 

For what it's worth, I've never had a *single* problem with Ext3, and
I've been using it with various distributions since I first started
playing with GNU/Linux a few weeks after Fedora Core 1 was released.

--Peter


signature.asc
Description: This is a digitally signed message part