Subject: Re: Bonnie output
To: None <shanee@rabbit.augusta.de>
From: Leo Weppelman <leo@wau.mis.ah.nl>
List: port-atari
Date: 02/26/1996 12:14:00
> I compiled Bonnie-1.0, a disk-benchmark and get the following result:
>
> -------Sequential Output-------- ---Sequential Input-- =
>
> --Random--
> -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- =
>
> --Seeks---
> Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /se=
> c =
>
> %CPU
> 100 263 91.0 507 31.0 311 39.7 242 89.2 783 48.5 28.=
> 5 =
>
> 16.5
>
> In my TT is a Fujitsu M2263S-512, 0194 with 640MByte, 1658 cyl, 15 head, =
>
> 52 sec 512 bytes/sec. Under TOS with How-Fast II there is a transfererate=
> =
>
> from =
>
> 1.3 MByte. Why it is so slow under NetBSD-1.1?
I think you more or less compare apples with pears when comparing TOS and
NetBSD disk throughput. Some of the difference are caused by:
- Memory protection between processes and proces/kernel. This means
that all I/O will generate an additional copy operation between
kernel and user space.
- Driver structure. The NetBSD driver has to be interrupt driven,
otherwise the system won't be multitasking.
- There are always some other processes that interfere with the
benchmark process when running it under NetBSD.
So the 1.3 Mb you mention for TOS is probably unreachable with NetBSD...
I have no idea if the figures you got are reasonable.
Leo.