[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RaiserFS PPC status

"David N. Welton" wrote:
> Andrew Sharp <andy@netfall.com> writes:
> > That would make sense, since they do all the Linux XFS developemnt
> > on ia32 and ia64 boxes!  At least they did when I worked in that
> > group.  But XFS is a very mature piece of code in general compared
> > to ReiserFS, which is much newer piece of code.  Maybe one day I'll
> > have a chance to do a performance analysis between the two.  They
> > have different internal designs.
> http://bulma.lug.net/body.phtml?nIdNoticia=626
> It looks kind of simplistic, but I've also heard anectdotal evidence
> that XFS is quite fast (and performs faster while being more stable
> than Reiserfs).

Yeah, anecdotal evidence is worth about what you pay for it ~:^). 
But this is a tad interesting, but obviously flawed.  One rule of
benchmarking is that if your results don't match what common sense
tells you, then you need to figure out what is going wrong.  Ext2fs
shouldn't be faster than reiserfs, especially if it is in one test
but not in another.  The results should also be much more
reproduceable.  The fact that copying a multi-megabyte file takes 20
seconds the first time and 60 seconds the next time, and 30 seconds
the next time, etc., that's not giving you decent data.  Statistics
guys have a word for that, but I don't know what it is off hand.

They didn't specify important details, like size of file system,
processor(s), memory size (very important).  That makes it quite
difficult to determine circumstancial info.  XFS has been in
development for maybe ten years or so, so be prepared for it to be
faster and better scaling in MP land.  But be similarly prepared for
it to bloat your kernel  ~:^)  If you have 4GB mem and 2 TB disk,
you probably don't mind the bloat at all. 


Reply to: