Re: ext2 large block size slow

Leonard N. Zubkoff (lnz@dandelion.com)
Thu, 24 Apr 1997 12:48:50 -0700

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: Jaromir Koutek: "AMD K5/gcc core dumps [Re: (no subject)]"
Previous message: A Guy called Tyketto: "Odd findings in 2.1.36"
Maybe in reply to: Jeff Garzik: "ext2 large block size slow"

Date: Thu, 24 Apr 1997 12:54:21 -0400 (EDT)
From: jgarzik@pobox.com (Jeff Garzik)

I was experimenting with using non-1024 block sizes in ext2fs --
I tried 2048 and 4096. It seems that using non-1024 block sizes
creates a noticable slowdown. Is using this feature not supported,
or not tested? I was doing this on 2.0.29 and 2.0.30.

It's probably having a mix of 1KB and 4KB that's slowing you down. I use 4KB
block file systems exclusively because my tests show them performing better
than 1KB file systems, and I have more disk space to waste than CPU cycles.

Leonard

Next message: Jaromir Koutek: "AMD K5/gcc core dumps [Re: (no subject)]"
Previous message: A Guy called Tyketto: "Odd findings in 2.1.36"
Maybe in reply to: Jeff Garzik: "ext2 large block size slow"