4 Dec
2000
4 Dec
'00
7:53 a.m.
Hi,
Need help in settling argument...does disk mirroring (via hardware)
predate RAID's mirroring capability? (I'm not interested in
OS software implemented mirroring here.)
Also, are there any hardware mirror methods that one could point
to and say: this is not capable of being called any kind of RAID!
(The argument in question is over the statement that
"all hardware based mirroring is a form of RAID".)
For example, is there any "stealth hardware mirroring" mechanism,
where a drive watches the bus, and obeys writes, but nothing else?
(This would be "drive-based" mirroring, as opposed to "controller-
based mirroring") (Since RAID is controller-based, a drive-based
mirroring mechanism would seem to be clearly not any kind of RAID.)
thanks!
Stan
Stan Sieler sieler(a)allegro.com
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
4 Dec
4 Dec
6:38 a.m.
At 10:53 AM 12/4/2000 -0800, you wrote:
Need help in settling argument...does disk mirroring
(via hardware)
predate RAID's mirroring capability? (I'm not interested in
OS software implemented mirroring here.)
Sounds like a patent suit :-)
Also, are there any hardware mirror methods that one
could point
to and say: this is not capable of being called any kind of RAID!
(The argument in question is over the statement that
"all hardware based mirroring is a form of RAID".)
This is kind of a specious argument. There were volume shadowing (aka
hardware mirroring) products available before people started calling it
"RAID" and in fact the IBM DASD farms could do volume shadowing and it
isn't "RAID" because RAID is "Redundant Array of *Inexpensive*
Disks" and
3380's certainly weren't "inexpensive."
The hoopla, if you can call it that, around RAID was that disks were
getting pretty cheap and you could reorganize them in several ways to
"enhance" certain features (like reliability or performance) that normally
you'd have to pay beau coup bucks for. Nearly everything "raid like" had
been done before by people like IBM and DEC but their disk farms didn't
come cheap.
For example, is there any "stealth hardware
mirroring" mechanism,
where a drive watches the bus, and obeys writes, but nothing else?
Look at some of the CDC dual ported drives. But there is still a controller
on the disk. If you try to say there isn't a controller involved then
you'll be looking at something that is splitting off the write lines to the
platters. (which I wouldn't rule out, just you will have to look for *that*
in the techical manual for so called "high reliability" storage that
perhaps IBM, CDC, Rand, or DEC may have sold to some customers)
(This would be "drive-based" mirroring, as
opposed to "controller-
based mirroring") (Since RAID is controller-based, a drive-based
mirroring mechanism would seem to be clearly not any kind of RAID.)
So, you going to tell us the patent number?
--Chuck
10:19 a.m.
Re:
Sounds like a patent suit :-)
No...just an argument, really! (I'd mention if it was patent
oriented.)
This is kind of a specious argument. There were volume
shadowing (aka
hardware mirroring) products available before people started calling it
"RAID" and in fact the IBM DASD farms could do volume shadowing and it
isn't "RAID" because RAID is "Redundant Array of *Inexpensive*
Disks" and
3380's certainly weren't "inexpensive."
I dug up a copy of the original RAID article: "A Case for Redundant Arrays
of Inexpensive Disks (RAID)" by Patterson, Gibson, and Katz (from June,
1988).
They must have trained to be patent lawyers, because their definition
of RAID is verrrry broad! (To the point where software implemented
mirroring would still be called RAID1.)
Basically, I found that if I tried to forget that RAID was an
acronym with a meaning, I can see how people claim that any and all
disk mirroring is a form of RAID ... even if the mirroring preceded
the name "RAID".
BTW, I found a reprint of the article in "Readings in Computer
Architecture", edited by Hill, Jouppi, and Sohi,
published in 2000 by Morgan Kaufmann. This is a great reference
book. Articles include:
Architecture of the IBM System/360; by Amdahl, Blaauw, & Brooks
Parallel Operation in the Control Data 6600
The Cray-1 Computer System
Cray-1 Computer Technology
Cramming More Components onto Integrated Circuits; by Gordon Moore
wait...maybe there's a web site...
http://www.mkp.com/books_catalog/catalog.asp?ISBN=1-55860-539-8
Chapter 1 - Classic Machines: Technology, Implementation, and Economics
Chapter 2 - Methods
Chapter 3 - Instruction Sets
Chapter 4 - Instruction Level Parallelism (ILP)
Chapter 5 - Dataflow and Multithreading
Chapter 6 - Memory Systems
Chapter 7 - I/O: Storage Systems, Networks, and Graphics
Chapter 8 - Single-Instruction Multiple Data (SIMD) Parallelism
Chapter 9 - Multiprocessors and Multicomputers
Chapter 10 - Recent Implementations and Future Prospects
Highly recommended!
Stan Sieler sieler(a)allegro.com
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
1:24 p.m.
On Mon, 4 Dec 2000, Stan Sieler wrote:
Re:
But actually, don't they have the cart before the horse and it is really
a case of one form of RAID being mirroring?
- don
Sounds like a patent suit :-)
No...just an argument, really! (I'd mention if it was patent
oriented.)
This is kind of a specious argument. There were
volume shadowing (aka
hardware mirroring) products available before people started calling it
"RAID" and in fact the IBM DASD farms could do volume shadowing and it
isn't "RAID" because RAID is "Redundant Array of *Inexpensive*
Disks" and
3380's certainly weren't "inexpensive."
I dug up a copy of the original RAID article: "A Case for Redundant Arrays
of Inexpensive Disks (RAID)" by Patterson, Gibson, and Katz (from June,
1988).
They must have trained to be patent lawyers, because their definition
of RAID is verrrry broad! (To the point where software implemented
mirroring would still be called RAID1.) Basically, I found that if I tried to forget that RAID
was an
acronym with a meaning, I can see how people claim that any and all
disk mirroring is a form of RAID ... even if the mirroring preceded
the name "RAID".
BTW, I found a reprint of the article in "Readings in Computer
Architecture", edited by Hill, Jouppi, and Sohi,
published in 2000 by Morgan Kaufmann. This is a great reference
book. Articles include:
Architecture of the IBM System/360; by Amdahl, Blaauw, & Brooks
Parallel Operation in the Control Data 6600
The Cray-1 Computer System
Cray-1 Computer Technology
Cramming More Components onto Integrated Circuits; by Gordon Moore
wait...maybe there's a web site...
http://www.mkp.com/books_catalog/catalog.asp?ISBN=1-55860-539-8
Chapter 1 - Classic Machines: Technology, Implementation, and Economics
Chapter 2 - Methods
Chapter 3 - Instruction Sets
Chapter 4 - Instruction Level Parallelism (ILP)
Chapter 5 - Dataflow and Multithreading
Chapter 6 - Memory Systems
Chapter 7 - I/O: Storage Systems, Networks, and Graphics
Chapter 8 - Single-Instruction Multiple Data (SIMD) Parallelism
Chapter 9 - Multiprocessors and Multicomputers
Chapter 10 - Recent Implementations and Future Prospects
Highly recommended!
Stan Sieler sieler(a)allegro.com
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
2:07 p.m.
Back in '90. I tried the Novell NETWARE approach to mirroring, but was
unable to make it work. That was due to the fact that the ADAPTEC
controller I was using didn't really work correctly at the alternate address
with the driver they provided for that purpose. I quickly abandoned it, in
favor of the software RAID offered by NT.
It's theoretically quite solid, but I've never used it myself. The NT
(software) approach seems more solid. RAID is always done in software to
greater or lesser extent. The RISC processors commonly used on
"RAID-5-capable" controllers is really the same thing as it really doesn't
matter who runs the software. Unfortunately, under NT4, there were problems
with inter-process communication before SP3. That would make anybody
suspicious. My NT server was emulating the NETWARE server, yet had the
advantage of the software compression offered by NT and the built-in backup
(not the best imaginable, but the best there is <sigh>) AND it emulates the
NETWARE server so I can use it from DOS workstations.
regards,
Dick
----- Original Message -----
From: "Stan Sieler" <ss(a)allegro.com>
To: <classiccmp(a)classiccmp.org>
Sent: Monday, December 04, 2000 11:53 AM
Subject: Question: Disk Mirroring preceding RAID?
Hi,
Need help in settling argument...does disk mirroring (via hardware)
predate RAID's mirroring capability? (I'm not interested in
OS software implemented mirroring here.)
Also, are there any hardware mirror methods that one could point
to and say: this is not capable of being called any kind of RAID!
(The argument in question is over the statement that
"all hardware based mirroring is a form of RAID".)
For example, is there any "stealth hardware mirroring" mechanism,
where a drive watches the bus, and obeys writes, but nothing else?
(This would be "drive-based" mirroring, as opposed to "controller-
based mirroring") (Since RAID is controller-based, a drive-based
mirroring mechanism would seem to be clearly not any kind of RAID.)
thanks!
Stan
Stan Sieler sieler(a)allegro.com
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
10 Dec
10 Dec
2:41 p.m.
----- Original Message -----
From: "Richard Erlacher" <edick(a)idcomm.com>
To: <classiccmp(a)classiccmp.org>
Sent: Tuesday, December 05, 2000 11:37 AM
Subject: Re: Question: Disk Mirroring preceding RAID?
Back in '90. I tried the Novell NETWARE approach
to mirroring, but
was
unable to make it work. That was due to the fact that
the ADAPTEC
controller I was using didn't really work correctly at the alternate
address
with the driver they provided for that purpose.
Oh? Interesting. We have (somewhere) a Netware 3.11 server that is
running
disk mirroring with 2 Adaptec controllers.
The above is on-topic, that 486DX2-66 EISA server is now 10 yrs old hi,
and when it retires (probably next year)
I will be giving it a good home, along with its cabling, tape drive,
UPS, Deskjet 500 and dozen or so diskless workstations, so my kids can
play dos and win3.x based games, on their own personal Novell network,
hehehe.....;^)
Never had a problem with that, or with a later model server and NW4.11
mirroring, also with adaptec's - 2940UW's IIRC.
I quickly abandoned it, in
favor of the software RAID offered by NT.
Yeah, that part of NT seems ok as well. Ran that on some NT 3.51
systems we had here, but we
recovered our sanity and they went.
It's theoretically quite solid, but I've never
used it myself. The NT
(software) approach seems more solid. RAID is always done in software
to
greater or lesser extent. The RISC processors
commonly used on
"RAID-5-capable" controllers is really the same thing as it really
doesn't
matter who runs the software.
Yes, and there is a difference in sophistication. All RAID controllers
were not created equal.
Our current NW4.11 box is an NEC Express 130SA, dual P2-400's.
It has a mainboard embedded Mylex DAC960PJ Raid Controller, and it's
very good indeed.
We have 4 x 8.9Gb drives configured as RAID Level 5, so we have about
24Gb useable.
Drive failure is transparent to the O/S, and we have been able to reboot
the system with a dead drive,
hot swap remove/replace it 5 or 6 times, eventually replace it with
anonther drive, and have the raid controller
rebuild the drive in background, without any significant slowdown, and
no impact on data integrity at all, and no need
to reboot, reconfigure whatever. "It just works." I like it.....:^)
Unfortunately, under NT4, there were problems
with inter-process communication before SP3. That would make anybody
suspicious. My NT server was emulating the NETWARE server, yet had
the
advantage of the software compression offered by NT
and the built-in
backup
(not the best imaginable, but the best there is
<sigh>)
Shudder.
AND it emulates the
NETWARE server so I can use it from DOS workstations.
Is this the NT - Novell Gateway? Tried that, seemed to work ok, but
couldn't really see the point.
Cheers
Geoff Roberts
Computer Systems Manager
Saint Mark's College
Port Pirie,
South Australia
geoffrob(a)stmarks.pp.catholic.edu.au
netcafe(a)tell.net.au
ICQ: 1970476
4 Dec
4 Dec
4:49 p.m.
On Mon, 4 Dec 2000, Stan Sieler wrote:
Hi,
Need help in settling argument...does disk mirroring (via hardware)
predate RAID's mirroring capability? (I'm not interested in
OS software implemented mirroring here.)
The term "disk mirroring" predates the term "RAID 1". The authors of
the RAID paper (students IIRC) just assigned a standard name to it.
Also, are there any hardware mirror methods that one
could point
to and say: this is not capable of being called any kind of RAID!
(The argument in question is over the statement that
"all hardware based mirroring is a form of RAID".)
Hmmm. I think RAID 1 is defined at a high enough level to encompass
all mirror techniques (hardware and software). IE RAID 1 is
maintaining an exact copy of the data on two (or more?) drives.
For example, is there any "stealth hardware
mirroring" mechanism,
where a drive watches the bus, and obeys writes, but nothing else?
(This would be "drive-based" mirroring, as opposed to "controller-
based mirroring") (Since RAID is controller-based, a drive-based
mirroring mechanism would seem to be clearly not any kind of RAID.)
thanks!
Stan
Stan Sieler sieler(a)allegro.com
www.allegro.com/sieler/wanted/index.html www.allegro.com/sieler
7:09 p.m.
Stan Sieler wrote:
Need help in settling argument...does disk mirroring
(via hardware)
predate RAID's mirroring capability? (I'm not interested in
OS software implemented mirroring here.)
At least in actual use.
The Sigma/Webster q-bus SMD and ESDI controllers provided drive mirroring
for PDP-11 and uVax systems, at least a year before I recall any RAID
technology being sold.
8621
days inactive
8627
days old
7 comments
7 participants
participants (7)
-
cmcmanis＠mcmanis.com -
donm＠cts.com -
edick＠idcomm.com -
elmo＠mminternet.com -
geoffrob＠stmarks.pp.catholic.edu.au -
ss＠allegro.com -
vaxman＠uswest.net