[Top] [All Lists]

Re: Comparing XFS with ext3 and ReiserFS

To: Eric Sandeen <sandeen@xxxxxxx>
Subject: Re: Comparing XFS with ext3 and ReiserFS
From: Juha Saarinen <juha@xxxxxxxxxxxx>
Date: Mon, 7 May 2001 20:34:34 +1200 (NZST)
Cc: GCS <gcs@xxxxxxxxxxxxxxxx>, "linux-xfs@xxxxxxxxxxx" <linux-xfs@xxxxxxxxxxx>
In-reply-to: <3AF5DC85.FACD3C1@xxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Sun, 6 May 2001, Eric Sandeen wrote:

> It's NOT that every created inode wastes space.  If you have some
> application that, on average, uses 1 million inodes, and sometimes
> spikes to 1.1 million inodes, then you might sometimes have 100,000
> unused, but still allocated, inodes on the system.
> In other words, the "high water mark" of allocated inodes is never
> reduced.
> But that's not so bad - create a fresh ext2 system, and you start off
> with allocated, but 100% unused inodes right from the start.  And if you
> didn't create it with _enough_ inodes, then your app will fail when it
> runs out.  So you make the ext2 filesystem with, maybe, 1.5x what you
> anticipate.  Which seems like a bigger waste...
> So with XFS, you only waste space if you dramatically change the inode
> usage downward.  If you use a filesystem for a squid cache, then delete
> all those files and make one big vidcap file, then yes, you'd be wasting
> space.  But that sort of scenario isn't too likely.

Just out of curiosity, how does NTFS handle such a scenario?



PGP fingerprint:
B7E1 CC52 5FCA 9756 B502  10C8 4CD8 B066 12F3 9544

<Prev in Thread] Current Thread [Next in Thread>