From owner-freebsd-hackers  Wed Dec 27 16:48:42 1995
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id QAA28482
          for hackers-outgoing; Wed, 27 Dec 1995 16:48:42 -0800 (PST)
Received: from who.cdrom.com (who.cdrom.com [192.216.222.3])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id QAA28473
          for <freebsd-hackers@FreeBSD.org>; Wed, 27 Dec 1995 16:48:38 -0800 (PST)
Received: from mailgate.ericsson.se (mailgate.ericsson.se [130.100.2.2])
          by who.cdrom.com (8.6.12/8.6.11) with ESMTP id QAA13348
          for <freebsd-hackers@FreeBSD.org>; Wed, 27 Dec 1995 16:48:27 -0800
Received: from sa.erisoft.se (epls01.sa.erisoft.se [150.132.128.1]) by mailgate.ericsson.se (8.6.11/1.0) with SMTP id BAA17767; Thu, 28 Dec 1995 01:40:55 +0100
Received: from sws021.sa.erisoft.se by sa.erisoft.se (4.1/SMI-4.1-ERIS0.99)
	id AA28094; Thu, 28 Dec 95 01:40:55 +0100
From: Mattias.Gronlund@sa.erisoft.se (Mattias Gronlund)
Received: by sws021.sa.erisoft.se (5.x/client-1.3)
	id AA04766; Thu, 28 Dec 1995 01:40:21 +0100
Message-Id: <9512280040.AA04766@sws021.sa.erisoft.se>
Subject: Re: iostat and msps
To: se@zpr.uni-koeln.de (Stefan Esser)
Date: Thu, 28 Dec 1995 01:40:20 +0100 (MET)
Cc: freebsd-hackers@FreeBSD.org
In-Reply-To: <199512271546.AA16282@Sysiphos> from "Stefan Esser" at Dec 27, 95 04:46:23 pm
X-Mailer: ELM [version 2.4 PL24]
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-hackers@FreeBSD.org
Precedence: bulk

On Wed, 27 Dec 1995 16:46 Stefan Esser wrote:
> Well, if we agree on some standard format, I'll make the NCR driver
> create the necessary data ...

I think it would be great if we could agree on some standard format.

Here's a proposal of parameters that whould be nice to collect:
a    busy   (clock ticks)
b    read_transfers (number of)
c    write_transfers (number of)
d    read_bytes (number of)
e    write_bytes (number of)
f    queue_length (sum of the queue length when transfer initiated)
g    wait (clock ticks when the wueue length > 1)
h    serve_length (sum of the serve queue length when transfer initiated)
i    fly_interrupts (number of) 
j    brk_interrupts (number of)

k    expected_disconnects (number of)
l    unexpected_disconnects (number of)

Is there any device that can serve more then I request at a time
that would need (h)?

I guess k and l only is relevant for scsi devices but it would
be nice to have them avaible.

If we let the statistics and the number of clockticks from boot (t) 
to be read together by one sysctl call, I think we could calculate at 
least the following intresting information from the collected data:

Average number of bytes read per transfer = d / b
Average number of bytes write per transfer = e / c

Average number of bytes read per clocktick = d / t 
Average number of bytes write per clocktick = e / t

Average number of read transfers per clocktick = b / t
Average number of write transfers per clocktick = c / t

Average number of requests on queue = f / (b+c)
Average number of requests on serve queue = h / (b+c)
Average service time = a / (b+c)

Percent busy = 100 * a / t
Percent wait = 100 * g / t

/Mattias