SCIENTIFIC-LINUX-USERS Archives

August 2014

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Yasha Karant <[log in to unmask]>
Reply To:
Yasha Karant <[log in to unmask]>
Date:
Sat, 2 Aug 2014 21:59:58 -0700
Content-Type:
text/plain
Parts/Attachments:
text/plain (53 lines)
On 08/02/2014 07:53 PM, Brent L. Bates wrote:
>       I'm sorry, but the proven, reliable, and fast file system is XFS,
> NOT ext4.  ext4 is the new kid on the block.  XFS has been around for
> probably 20 YEARS, if not longer.  Half of that time also under Linux.
> ext4 hasn't been around nearly that long.  XFS is the tried and true,
> dependable, reliable, resilient, and fast file system.  I've seen it
> survive hardware crashes and flaky disk drives and keep on going.
> I've used it under both 32bit (not huge disk drives) and 64bit Linux
> with no problems.  I would not use any other file system under Linux
> and if I could, I'd use it under other OS's as well.  It is just that
> good, fast, and reliable.
>
Although I am not a strong believer in Wikipedia as an unbiased reliable 
resource; I have found it a good starting point.  The Wikipedia
statements about the limitations of xFS agree with our experience (we 
were an SGI "shop" for a number of years)  -- we have not done the file 
system comparison attributed to Larabel and thus cannot address that 
particular issue from our experience:

http://en.wikipedia.org/wiki/XFS


    Comparison

  * An XFS file system cannot be shrunk.
  * Metadata operations in XFS have historically been slower than with
    other file systems, resulting in, for example, poor performance with
    operations such as deletions of large numbers of files. However, a
    new XFS feature implemented by Dave Chinner and called /delayed
    logging/, available since versoin 2.6.39 of the Linux kernel
    mainline <http://en.wikipedia.org/wiki/Linux_kernel_mainline>, is
    claimed to resolve this;^[21]
    <http://en.wikipedia.org/wiki/XFS#cite_note-21> performance
    benchmarks done by the developer in 2010 revealed performance levels
    to be similar to ext4 <http://en.wikipedia.org/wiki/Ext4> at low
    thread counts, and superior at high thread counts.^[22]
    <http://en.wikipedia.org/wiki/XFS#cite_note-22>
  * In May 2013, M. Larabel published an extensive comparative test of
    filesystems on a single SSD which showed XFS to be not yet as fast
    as ext4.^[23] <http://en.wikipedia.org/wiki/XFS#cite_note-23>

I do note that we do not have an implementation with delayed logging -- 
presumably this is implemented in the current EL7 version.

Assuming that that are no scripts that convert existing extN file system 
manipulation scripts into the equivalent EL XFS scripts, is there a 
"rosetta stone" that shows the equivalences (translations) for common 
operations?  (Commonly performed operations including fsck, mount, 
format, file system growth and shrinkage, etc.)


Yasha Karnat

ATOM RSS1 RSS2