Don't forget SCSI's ability to multitask which is becoming it's
only reason to use it in a server environment.
I have the same workstation setup as you do, for the same
reason. It was a pretty hard lesson to learn that I could have
done the same thing with IDE and not wasted so much money on
heavy duty SCSI stuff. Sigh.
I just noticed that my email program is responding directly to
you. It is supposed to be responding to the list server. Can
you do me a favor and direct your response to the list server if
I miss it again? I don't know why it chose yours to do this to.
Normally it works.
>>> Cory Petkovsek <[EMAIL PROTECTED]> 12/4/2000 11:56:28
AM >>>
RAID 1 is faster for read than RAID 5, however RAID 5 is still
pretty fast for reading because it doesn't need to calculate so
much, mostly on writes. Hardware raid is vastly superior to
software raid though. And SCSI 3 Ultra160 should be superior to
IDE anything. But apparently it is not.
It looks like the only advantages to a SCSI system is everything
but speed and cost: space, external arrays, fault tollerance.
Yet IDE is faster and cheaper, apparently.
Cory
On Mon, Dec 04, 2000 at 11:27:59AM -0800, Bob Crandell wrote:
> From what I understand about RAID, RAID 1 (mirroring) is
faster
> than RAID 2 because of the math involved. This is true of
> software or hardware RAID.
>
> I don't know if it's true of other OS, but Novell
reads/writes
> from/to the drive that's not busy and syncs during idle times.
> This makes for a system that is visibly faster than a single
> drive box.
>
> Ah yes. I get it now. The old man page ploy. Be prepared
to
> be impressed.
>
> >>> Cory Petkovsek <[EMAIL PROTECTED]> 12/4/2000
10:59:54
> AM >>>
> Ha, your firewall harddrive looks faster than your fileserver.
> You should run it a couple of times so the first time it
> allocates memory and the other times it runs a real
benchmark.
>
> You start tweaking by reading the man page.
>
> quick reference:
> c1/0 32-bit/16-bit mode
> d1/0 dma on/off
>
> ie. hdparm -c1 -d1 /dev/hda
>
> Chris's Ultra66 went from 3.5mb to 27mb/sec. It looks like
his
> single 7200rpm ultra66 drive reads faster than my Ultra160 3
> drive, raid 5 10krpm array. :(
>
> Cory
>
>
> On Mon, Dec 04, 2000 at 10:55:11AM -0800, Bob Crandell wrote:
> > My Firewall where I don't care haw fast the HD is:
> > CSGate:~# hdparm -tT /dev/hda
> >
> > /dev/hda:
> > Timing buffer-cache reads: 128 MB in 2.86 seconds
=44.76
> > MB/sec
> > Timing buffered disk reads: 64 MB in 4.21 seconds =15.20
> > MB/sec
> >
> > My file server (I'm hoping will replace our NW4.11
someday):
> > csmule:~# hdparm -tT /dev/hda
> >
> > /dev/hda:
> > Timing buffer-cache reads: 128 MB in 7.93 seconds =
16.14
> > MB/sec
> > Timing buffered disk reads: 64 MB in 6.18 seconds =
10.36
> > MB/sec
> > Hmm.. suspicious results: probably not enough free memory
for
> a
> > proper test.
> > /dev/hdb:
> > Timing buffer-cache reads: 128 MB in 7.82 seconds =
16.37
> > MB/sec
> > Timing buffered disk reads: 64 MB in 6.65 seconds =
9.62
> > MB/sec
> > Hmm.. suspicious results: probably not enough free memory
for
> a
> > proper test.
> >
> > csmule:~# free
> > total used free shared
buffers
>
> > cached
> > Mem: 160252 156112 4140 32752
118264
>
> > 8412
> > -/+ buffers/cache: 29436 130816
> > Swap: 224896 2152 222744
> >
> > I've heard this stuff is tweak-able. How do I? Where do I
> > start?
> >
> > >>> Cory Petkovsek <[EMAIL PROTECTED]> 12/2/2000
> 9:13:14
> > PM >>>
> > <snip>
> > What kind of rates do you guys get with various hardware?
> > Controllers, Raid, IDE, mdma/udma 33/66/100...
> >
> > Test your read speed under linux:
> > ide only: (tests cache and disk reads, independently)
> > hdparm -tT /dev/hda
> >
> > any drive:
> > time dd if=/dev/sda of=/dev/null bs=1024 count=102400
> > (returns minutes and seconds)
> > bc
> > 100/(minutes*60+seconds)
> >
> > <snip>
> > Cory
