Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 3 Apr 1996 12:40:22 +0930 (CST)
From:      Michael Smith <msmith@atrad.adelaide.edu.au>
To:        bde@zeta.org.au (Bruce Evans)
Cc:        luigi@labinfo.iet.unipi.it, msmith@atrad.adelaide.edu.au, davidg@Root.COM, dutchman@spase.nl, freebsd-hackers@freebsd.org
Subject:   Re: HDD cpu usage (IDE vs. SCSI).
Message-ID:  <199604030310.MAA17901@genesis.atrad.adelaide.edu.au>
In-Reply-To: <199604021147.VAA11954@godzilla.zeta.org.au> from "Bruce Evans" at Apr 2, 96 09:47:18 pm

next in thread | previous in thread | raw e-mail | index | archive | help
Bruce Evans stands accused of saying:
> 
> >No, you're not understanding.  For a given CPU, IDE will _always_ use more
> >CPU time than SCSI.  Period.  
> 
> Really?  Please give numbers for a PIO mode 4 IDE controller vs an ST01
> SCSI controller :-).  Please give numbers for your choice of controllers
> vs my choices of applications an i/o access patterns.  I'll choose a
> memory intensive application that stalls the CPU waiting for the SCSI
> controller.  I'll arrange the i/o so that memory caching is defeated
> at strategic places.

Grrr, pedant 8)  I'll show you a pile of happy users who've said 
"wow, that's so much faster", and "gee, I was going to replace the CPU"
when finally talked into using SCSI for their production systems.

> >If you have lots of free CPU, then IDE is fine, but if you feel that your 
> >CPU has better things to do with its time than copy data to and from
> >your disk, then SCSI is the only solution that makes sense.
> 
> I think lots of free CPU is the usual case.  E.g., right now on freebsd.org:
> 
>     3:40AM  up 19 days, 11 mins, 16 users, load averages: 0.41, 0.32, 0.31

These are averages, and have little or no bearing on the 'feel' of a 
system, particularly during 'peaky' loads (starting an X application is a 
good example 8).

> I would prefer lower latency to lower overhead in most cases.  IDE disks
> have natural advantages in this area (no complicated SCSI protocol to
> interpreted by the slow i/o processor on the controller).

Presuming you only have one application making requests in a linear fashion,
that's fine.  Tagged queueing and disconnect rapidly improves things 
once you start to get busy though.

You're arguing out of character, which is confusing.  Stop it 8)

> Bruce

-- 
]] Mike Smith, Software Engineer        msmith@atrad.adelaide.edu.au    [[
]] Genesis Software                     genesis@atrad.adelaide.edu.au   [[
]] High-speed data acquisition and      (GSM mobile) 0411-222-496       [[
]] realtime instrument control          (ph/fax)  +61-8-267-3039        [[
]] Collector of old Unix hardware.      "Where are your PEZ?" The Tick  [[



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604030310.MAA17901>