Date: Thu, 6 Jul 2000 11:32:47 -0400 (EDT) From: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> To: Stijn Hoop <stijn@win.tue.nl> Cc: current@FreeBSD.ORG Subject: Large disks (was Re: bin/19635: add -c for grand total to df(1)) Message-ID: <200007061532.LAA33286@khavrinen.lcs.mit.edu> In-Reply-To: <20000706131840.B58747@pcwin002.win.tue.nl> References: <16565.962880093@axl.ops.uunet.co.za> <v04220803b58a161b72db@[195.238.1.121]> <20000706131840.B58747@pcwin002.win.tue.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
<<On Thu, 6 Jul 2000 13:18:40 +0200, Stijn Hoop <stijn@win.tue.nl> said: > Maybe this isn't the right list to ask, but stepping into this: > I bought a 30G drive recently, and I was wondering if the 10% 'rule' > for performance is still really needed. I mean, I lose 3 _gigs_ of > storage space, and otherwise the performance detoriates? That > doesn't make sense to me. Yes. The efficiency of the hashing mechanism used to lay out new blocks on the disk depends only on what fraction of the disk is used, not how much space that represents. (On the other hand, it is unlikely that you are significantly stressing the allocator in any meaningful way.) If you're concerned about wasted disk space, there's a lot to be gained by fiddling with the block sizes, bytes per inode, and other layout parameters. Your 30-GB disk probably has a zillion cylinder groups, which is far too many to actually be helpful in disk layout. When creating a large filesystem, it pays to increase the `-c' parameter as high as newfs will permit. -GAWollman To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200007061532.LAA33286>