xfs
[Top] [All Lists]

Re: librmt

To: "Davida, Joe" <Joe_Davida@xxxxxxxxxx>
Subject: Re: librmt
From: Steve Lord <lord@xxxxxxx>
Date: Fri, 29 Sep 2000 15:58:00 -0500
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: Message from "Davida, Joe" <Joe_Davida@maxtor.com> of "Fri, 29 Sep 2000 13:37:13 MDT." <09D1E9BD9C30D311919200A0C9DD5C2C02536FCC@mcaexc01.msj.maxtor.com>
Sender: owner-linux-xfs@xxxxxxxxxxx
>       Thanks Steve,
>       I did a complete fresh checkout and it
>       brought in librmt.
>       
>       I have a question re: best layout of the raid
>       array. For best performance for small files,
>       what should the raid stripe size be? The
>       Xtremeraid2000 allows a minimum raid block
>       size of 8KB per disk in the stripe, max is
>       64KB. Concurrent to this, what should be the
>       XFS block size for best I/O
>       performance on lots of small files? I was thinking
>       that the FileSystem block size should in all
>       cases be identical to stripe block size, to allow
>       as much parallellism of I/O on the drives in the raid
>       as possible.
> 
>       Cheers,
> 
>       Joe
> 

First of all, we are currently only supporting 4K filesystem block size, so
for now you should use this.  I don't really have a hard and fast answer for
what would behave best for small files, since at SGI striped file systems
usually go together with big I/O. But I would try 8K for now.

Steve

p.s. how goes the mkfs issue?



<Prev in Thread] Current Thread [Next in Thread>
  • Re: Librmt, Nathan Scott
    • Re: librmt, Steve Lord <=