14 Jul
2020
14 Jul
'20
10:47 a.m.
This may be a bit too new for this list but I thought what the heck - maybe
one of you Compaq/DEC/HP guys would know:
Is there any reason a Smart Array controller can't be used as a simple SCSI
controller? I.E. No array, just using it to drive a tape library? TIA!
-Ali
14 Jul
14 Jul
10:57 a.m.
On 7/14/20 10:47 AM, Ali via cctalk wrote:
This may be a bit too new for this list but I thought
what the heck - maybe
one of you Compaq/DEC/HP guys would know:
Is there any reason a Smart Array controller can't be used as a simple SCSI
controller? I.E. No array, just using it to drive a tape library? TIA!
-Ali
All the Smart Array controllers, to my knowledge, have the ability to
run as a JBOD, so I imagine that would work. The biggest issue I've
found with older SA controllers, is getting them configured (and driver
support for really old ones on modern OSes).
--Jason
11:08 a.m.
All the Smart Array controllers, to my knowledge, have
the ability to
run as a JBOD, so I imagine that would work. The biggest issue I've
found with older SA controllers, is getting them configured (and driver
support for really old ones on modern OSes).
Jason,
Thanks. To be clear there are no HDDs connected to the controller (there is a separate
5304 controller that has all the HDDs on it). I just did not have enough external ports on
the 5304 to drive the tape library so I wanted to repurpose the original 3200 controller.
I would think it would not be an issue but the system keeps complaining on boot that there
are no physical drives (I assume it means HDDs) connected to the 3200.
-Ali
11:32 a.m.
On 7/14/20 12:08 PM, Ali via cctalk wrote:
Thanks. To be clear there are no HDDs connected to the
controller
(there is a separate 5304 controller that has all the HDDs on it). I
just did not have enough external ports on the 5304 to drive the
tape library so I wanted to repurpose the original 3200 controller. I
would think it would not be an issue but the system keeps complaining
on boot that there are no physical drives (I assume it means HDDs)
connected to the 3200.
Can you move one HDD to the SMARTarray (to make it happy) and put the
tape library on the 5304?
My concern is that some RAID cards don't like to present things other
than (logical / virtual) hard drives to the host OS. So I'm not sure if
the tape drive will work connected to it or not.
--
Grant. . . .
unix || die
12:22 p.m.
Can you move one HDD to the SMARTarray (to make it
happy) and put the
tape library on the 5304?
My concern is that some RAID cards don't like to present things other
than (logical / virtual) hard drives to the host OS. So I'm not sure
if
the tape drive will work connected to it or not.
Grant,
I think Dave hit it on the head. The Smart Array controllers apparently only take HDDs and
no other SCSI devices. I guess I am going to have to throw in a Adaptec card after all.
-Ali
3:30 p.m.
On Jul 14, 2020, at 12:22 PM, Ali via cctech wrote:
I think Dave hit it on the head. The Smart Array
controllers apparently only take HDDs and no other SCSI devices.
Not all of them. But it is a model-by-model "feature", and maybe not at all on
any models prior to some generation. I have a tape library (and only a tape library)
attached to a Smart Array P212, but this is a much newer (SAS) host adapter than the 3200.
I vaguely recall having had to do some reading of spec sheets to find out that this one
would work with tape.
ok
bear.
--
until further notice
11:11 a.m.
On 14 Jul 2020, at 18:47, Ali via cctalk <cctalk at
classiccmp.org> wrote:
This may be a bit too new for this list but I thought what the heck - maybe
one of you Compaq/DEC/HP guys would know:
Is there any reason a Smart Array controller can't be used as a simple SCSI
controller? I.E. No array, just using it to drive a tape library? TIA!
Nope. I don?t remember any Smart Array controllers supporting passthrough tapes until much
later on. You just need a Wide Ultra 2 or 3 adapter for libraries, eg 154457-B21. There?s
no mention of tape support in the install manual either.
--
Adrian Graham
Owner of Binary Dinosaurs, the UK's biggest private home computer collection?
t: @binarydinosaurs f: facebook.com/binarydinosaurs
w: www.binarydinosaurs.co.uk
5:04 p.m.
Nope. I don?t remember any Smart Array controllers
supporting
passthrough tapes until much later on. You just need a Wide Ultra 2 or
3 adapter for libraries, eg 154457-B21. There?s no mention of tape
support in the install manual either.
Thanks. I actually needed a dual port beasty so I am going to throw in an Adaptec 39160.
The MSL5026 won't even come close to taxing the Adaptec but it has a Compaq ROM and is
directly supported by SmartStart.
-Ali
11:41 a.m.
Ali,
Its been a long time, but I think the answer is no. A tape is not a disk and
uses a different SCSI command set to a disk. For example you can't rewind a
disk.
I found a SmartArray 3200 manual here:-
http://index-of.es/Tutorials/Proliant%201600/SmartArray3200.pdf
and in the Q&A it says :-
Q. Does the Smart Array 3200 Controller support SCSI tape drives and CD-ROM
drives?
A. No; the Smart Array 3200 Controller only supports Wide-Ultra2
SCSI,Wide-Ultra SCSI-3, Fast-Wide SCSI-2, and Fast SCSI-2 hard drives.
So I would say not.
I know you can define a JBOD but a JBOD is not passed through as a physical
disk. In most RAID controllers you can still carve up a JBOD into multiple
LUNS.
The devices the OS sees are "logical abstractions" of the devices the RAID
controller sees and need to be be mapped to the physical devices in the
controller.
Dave
G4UGM
-----Original Message-----
From: cctalk <cctalk-bounces at classiccmp.org> On Behalf Of Ali via cctalk
Sent: 14 July 2020 18:47
To: 'General Discussion: On-Topic and Off-Topic Posts'
<cctalk at classiccmp.org>
Subject: Compaq Smart Array 3200 Controller as a SCSI Controller
This may be a bit too new for this list but I thought what the heck -
maybe
one of you Compaq/DEC/HP guys would know:
Is there any reason a Smart Array controller can't be used as a simple
SCSI
controller? I.E. No array, just using it to drive a
tape library? TIA!
-Ali
12:22 p.m.
Ali,
Its been a long time, but I think the answer is no. A tape is not a
disk and
uses a different SCSI command set to a disk. For example you can't
rewind a
disk.
I found a SmartArray 3200 manual here:-
http://index-of.es/Tutorials/Proliant%201600/SmartArray3200.pdf
and in the Q&A it says :-
Q. Does the Smart Array 3200 Controller support SCSI tape drives and
CD-ROM
drives?
A. No; the Smart Array 3200 Controller only supports Wide-Ultra2
SCSI,Wide-Ultra SCSI-3, Fast-Wide SCSI-2, and Fast SCSI-2 hard drives.
So I would say not.
Dave,
RTFM - Read The Friggin Manual. :) That pretty much settles it. Thanks for
finding that bit of info. I am going to throw in an Adaptec controller in
there and call it done!
-Ali
1:24 p.m.
-----Original Message-----
From: Ali <cctalk at ibm51xx.net>
Sent: 14 July 2020 20:22
To: 'Dave Wade' <dave.g4ugm at gmail.com>; 'General Discussion:
On-Topic
and Off-Topic Posts' <cctalk at classiccmp.org>
Subject: RE: Compaq Smart Array 3200 Controller as a SCSI Controller
there
Ali,
Its been a long time, but I think the answer is no. A tape is not a
disk and uses a different SCSI command set to a disk. For example you
can't rewind a disk.
I found a SmartArray 3200 manual here:-
http://index-of.es/Tutorials/Proliant%201600/SmartArray3200.pdf
and in the Q&A it says :-
Q. Does the Smart Array 3200 Controller support SCSI tape drives and
CD-ROM drives?
A. No; the Smart Array 3200 Controller only supports Wide-Ultra2
SCSI,Wide-Ultra SCSI-3, Fast-Wide SCSI-2, and Fast SCSI-2 hard drives.
So I would say not.
Dave,
RTFM - Read The Friggin Manual. :) That pretty much settles it. Thanks for
finding that bit of info. I am going to throw in an Adaptec controller in and call it done!
-Ali
Ali,
No probs! It has been a long time since I worked for Compaq, but I remember
having huge issues with mixing tapes and disks and RAID controller
compatibility.
We used to have a big spreadsheet that showed all the compatible
controllers/devices/operating systems, but even that failed us at times.
Oddly, my old Microchannel Cheetah RAID controller in my P390 supports tapes
and CD ROMS....
Dave
2:20 p.m.
On 7/14/20 12:41 PM, Dave Wade via cctalk wrote:
I know you can define a JBOD but a JBOD is not passed
through as a
physical disk.
JBOD / single disk RAID 0 is different than pass through.
I've worked with a few RAID cards that did support actual pass-through mode.
I've not had any problems with things that are truly passed through.
But I've not seen it many times. (Likely an effect of the corpus that
I've worked with.)
--
Grant. . . .
unix || die
15 Jul
15 Jul
1:16 a.m.
On Tue, Jul 14, 2020 at 10:47:11AM -0700, Ali via cctalk wrote:
[...]
Is there any reason a Smart Array controller can't
be used as a simple SCSI
controller? I.E. No array, just using it to drive a tape library? TIA!
In general, hardware RAID controllers cannot be used as ordinary controllers.
There is no good use case for them in 2020, which is why they're all suddenly
quite cheap. Much cheaper in fact than non-RAID controllers, IME, which irks me
as I'm in the market for a plain SAS HBA for use with ZFS.
11:29 a.m.
On 7/15/20 2:16 AM, Peter Corlett via cctalk wrote:
...I'm in the market for a plain SAS HBA for use
with ZFS.
Check and see if you can flash the Initiator Target (IT) firmware onto
the RAID card such that it's no longer an Integrated RAID (IR) device.
Thus giving you the HBA mode that you want.
I've done this on a few RAID chip sets built into Supermicro
motherboards and been quite happily running ZFS on them. }:-)
--
Grant. . . .
unix || die
16 Jul
16 Jul
4:36 a.m.
On Wed, 15 Jul 2020 at 22:09, Ali via cctalk <cctalk at classiccmp.org> wrote:
Happily for me I don't do stuff like build production servers any
more, but my understanding is this:
With modern hardware, it's easier, cheaper and more flexible to build
and manage arrays in software, using modern filesystems such as ZFS,
Btrfs, or MS Storage Spaces on Windows Server.
I was recently documenting the use of Btrfs for this on SUSE Linux
Enterprise Server: the benefits of doing it in software are that you
can dynamically migrate arrays between different RAID levels, add new
drives and resize the array to include them or add them as additional
parity disks on the fly, you can mark individual files as having
different RAID levels (for example, you could place the OS' virtual
memory space in a file on the RAID and tell the filesystem not to
compute parity for it, just stripe it, for better performance). ZFS
and Ceph allow for a mix of high-speed (e.g. SSD, NVMe, even NVDIMM)
storage and low-speed but large rotational storage, and use the faster
storage to cache the slower stuff.
And of course, if your server dies, then the array can be mounted on
any other box with the same OS and you can retrieve data from it --
which is far more difficult if a hardware RAID controller dies, in
which case you might need the same firmware revision etc., and
possibly onboard controller config info.
Ceph now basically lets you build arrays of storage servers, so that
you can, say, have single storage volumes comprising local storage in
different countries, or on different continents, for local access
speed and the software syncs it in the background between zones or
regions. So it's no longer an array of physical disks on one server,
it's an array of servers with disks in them -- and the servers and the
disks may themselves be virtualised.
It gets very complicated but it's also very powerful and flexible.
Dedicated hardware just can't do stuff like this any more.
--
Liam Proven ? Profile: https://about.me/liamproven
Email: lproven at cix.co.uk ? gMail/gTalk/gHangouts: lproven at gmail.com
Twitter/Facebook/LinkedIn/Flickr: lproven ? Skype: liamproven
UK: +44 7939-087884 ? ?R (+ WhatsApp/Telegram/Signal): +420 702 829 053
There is no good use case for them in 2020, which
is why they're all
suddenly
quite cheap.
Peter,
Why do you say that? Not disagreeing per se but just wondering the reasoning
behind it. 
5:09 a.m.
And of course, if your server dies, then the array can
be mounted on
any other box with the same OS and you can retrieve data from it --
which is far more difficult if a hardware RAID controller dies, in
which case you might need the same firmware revision etc., and
possibly onboard controller config info.
Funny story about that: In 1990 I installed a Compaq systempro for
Hechinger's that cost over $100,000. Had a full compliment of memory,
max internal disk on the ATA controller, and two external boxes of disks
with the smart SCSI-ish controller. Massive system running Sybase SQL.
Designed to replace a 24*7 mainframe and expected to be up all the time.
Got a call 2 months later: The system had blown a hole in one of the
disk controllers and was down. Called Compaq, they got someone on a
plane with a spare controller from the west coast and I drove out to
meet them in the middle of the night so we could get the system up by
morning.
That was pretty insane. And pointed out that "mainframe" PC's didn't
have anywhere near the redundancy or support of true mainframes.
C
5:18 a.m.
On Thu, 16 Jul 2020 at 14:09, Chris Zach via cctalk
<cctalk at classiccmp.org> wrote:
Funny story about that: In 1990 I installed a Compaq systempro for
Hechinger's that cost over $100,000.
I just about remember the SystemPro machines. One of my bigger clients
in my first job got one, but they hired a full-time guy to customise
their app for it, and he also became the sysadmin. A very early
example of devops, I suppose. So I never got to play around with it.
:-(
Had a full compliment of memory,
max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
What OS, just out of interest?
and two external boxes of disks
with the smart SCSI-ish controller. Massive system running Sybase SQL.
Designed to replace a 24*7 mainframe and expected to be up all the time.
A single box? Oh dear.
Got a call 2 months later: The system had blown a hole
in one of the
disk controllers and was down. Called Compaq, they got someone on a
plane with a spare controller from the west coast and I drove out to
meet them in the middle of the night so we could get the system up by
morning.
That was pretty insane. And pointed out that "mainframe" PC's didn't
have anywhere near the redundancy or support of true mainframes.
Oh yes indeed.
I've had catastrophic hardware failures, but luckily, none that took
out a RAID controller. I've just heard the horror stories.
I finally left the support business in about 2011, but by then, it was
fairly standard practice to install VMware (the free VMware ESX
hypervisor if the company didn't have a paid vSphere site licence) on
all new boxes, then install the OS in that. Even if it was a dedicated
machine that only ran 1 OS ever. Because that way, if the machine
died, you could restore the backup onto a new, totally different box,
so long as it was running ESX, and it would Just Work? with no driver
or activation issues -- the virtualised hardware was the same.
--
Liam Proven ? Profile: https://about.me/liamproven
Email: lproven at cix.co.uk ? gMail/gTalk/gHangouts: lproven at gmail.com
Twitter/Facebook/LinkedIn/Flickr: lproven ? Skype: liamproven
UK: +44 7939-087884 ? ?R (+ WhatsApp/Telegram/Signal): +420 702 829 053
8:35 a.m.
ATA? That long ago?
Sorry, IDE like. Forgot the terminology. You could put 4 drives on a
controller, then two controllers per unit (EISA was cool).
Possible but unusual in a server, I would have
thought.
What OS, just out of interest?
I think it was SCO Unix.
A single box? Oh dear.
We were so silly back then :-)
I've had catastrophic hardware failures, but
luckily, none that took
out a RAID controller. I've just heard the horror stories.
I remember pulling it. There was a hole where one of the ASIC chips was.
Pretty amazing to be honest, but oh well. Compaq took it back to the
factory for review.
I finally left the support business in about 2011, but
by then, it was
fairly standard practice to install VMware (the free VMware ESX
hypervisor if the company didn't have a paid vSphere site licence) on
all new boxes, then install the OS in that. Even if it was a dedicated
machine that only ran 1 OS ever. Because that way, if the machine
died, you could restore the backup onto a new, totally different box,
so long as it was running ESX, and it would Just Work? with no driver
or activation issues -- the virtualised hardware was the same.
VMWare was *great*. I started using it on a small IBM box, then once I
realized it was like a true mainframe OS we bought a pair of IBM 366's
and a DS4300 SAN. Then upgraded the CPUs on them (4 CPUs). Then a set of
DS4700 SANs (redundant arrays of arrays with 2 controllers per array).
Then got a third so we could always run 2/3 of the cluster as opposed to
50% (for failover, see SystemPro). Then 3850's and 3950's with QPI
memory sharing. God that worked, I was able to get 70-80 servers per box
running away....
Much cheaper to run systems there than in the cloud. But everyone loves
the cloud, so off we go.
C
8:40 a.m.
New subject: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
Had a full
compliment of memory,
max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
What OS, just out of interest?
Target OS was WinNT 3.1 initially and then 4.0. 2K was also supported but the machine
really was not meant for 2k. You could also run OS/2, Novell Netware, Compaq DOS, and
supposedly there was even a version of MS LanMan (the full server OS not the client) for
the Systempro that allowed SMP.
-Ali
8:54 a.m.
New subject: NCR 3550 Digital Library Was Re: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
Now in terms of the most MANLY system I worked on, that would be the
NCR3550 we had at the IEEE Computer Society. When I arrived in 1993 it
had been donated, but was doing nothing with 4 486 CPUs in it and a
weird copy of AT&T unix. I took one look at the 256 bit interleaved
memory architecture the 3 levels of cache with affinity, the infinite
amount of space for disks, and the dual micro-channel busses and fell in
*love*
We talked to NCR, upgraded it to 512mb memory, 8 Pentium Pro/200 CPUs,
and dual Microchannel busses with FDDI and Ethernet interfaces. Loaded
it with disks, installed Windows NT 4.0 on it, and turned it into TALOS,
the main server for the IEEE Computer Society's Digital Library (which I
built).
Partnered with Anderson and Netscape to multi-thread commerce server
(SSL), built an E-account system in Lotus Domino/Notes, and loaded up
all of our SGML with an SGML to HTML converter (Dynaweb) and a custom
tool that could convert Tek math to GIFs on the fly. That process could
take advantage of all 8 CPUs and render complex math articles in real time.
Also did e-commerce for awhile with online credit card processing for
memberships and conferences (SuperComputing/95 was the first conference
to do on-line credit cards, I built that too because I was sick and
tired of keying in the cards myself. Laziness is next to godliness)
It served for years as the CS Digital Library core server with
30,000-40,000 accounts in active use. Man that thing was a truck, I wish
I knew what had happened to it.
And to think, it all started with the computer room ceiling collapsing
Funny story about that - I just setup a Systempro XL at home to play with. It is fully
decked out w/ dual processor 50MHZ 486s (not DX2), 512MB of memory, a 4GB SCSI Boot Drive
and six 2GB SCSI drives in RAID 5. The Compaq systems came standard with what Compaq
called the IDA (Intelligent Drive Array). It was IDE based but did not use standard IDE
drives. I think it could do RAID 0, 1, and 3 (or the equivalents there of). Compaq even
had a few iterations of the controller and cached ones. Interestingly the Systempro XL had
a SCSI 2 controller on the MB mainly used for the tape dive or CD while the base config
came with an IDA 2 controller and could have up to eight drives. In addition you could
install extra IDA controllers for even more drives or to drive external boxes. Or you
could upgrade to a SCSI array - which is what I have running in my Systempro XL.
from all the RS232 cables from the Vax and crushing our
Sun Sparc 20 web
server that kicked off this whole thing.
I should write a book or an article about that. We did so much that was
so... new... and all of that could be forgotten like tears in the rain....
CZ
On 7/16/2020 11:40 AM, Ali via cctalk wrote:
Had a full compliment of memory,
max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
What OS, just out of interest?
Target OS was WinNT 3.1 initially and then 4.0. 2K was also supported but the machine
really was not meant for 2k. You could also run OS/2, Novell Netware, Compaq DOS, and
supposedly there was even a version of MS LanMan (the full server OS not the client) for
the Systempro that allowed SMP.
-Ali
21 Jul
21 Jul
8:50 a.m.
New subject: NCR 3550 Digital Library Was Re: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
Wow, would love to have a machine like that. The ?weird unix? was probably
MP-RAS which was NCR?s SysVr4. NCR was selling massive x86 MCA systems for
Terradata setups in the early ?90s.
On Thu, Jul 16, 2020 at 8:54 AM Chris Zach via cctalk <cctalk at classiccmp.org>
wrote:
Now in terms of the most MANLY system I worked on,
that would be the
NCR3550 we had at the IEEE Computer Society. When I arrived in 1993 it
had been donated, but was doing nothing with 4 486 CPUs in it and a
weird copy of AT&T unix. I took one look at the 256 bit interleaved
memory architecture the 3 levels of cache with affinity, the infinite
amount of space for disks, and the dual micro-channel busses and fell in
*love*
We talked to NCR, upgraded it to 512mb memory, 8 Pentium Pro/200 CPUs,
and dual Microchannel busses with FDDI and Ethernet interfaces. Loaded
it with disks, installed Windows NT 4.0 on it, and turned it into TALOS,
the main server for the IEEE Computer Society's Digital Library (which I
built).
Partnered with Anderson and Netscape to multi-thread commerce server
(SSL), built an E-account system in Lotus Domino/Notes, and loaded up
all of our SGML with an SGML to HTML converter (Dynaweb) and a custom
tool that could convert Tek math to GIFs on the fly. That process could
take advantage of all 8 CPUs and render complex math articles in real time.
Also did e-commerce for awhile with online credit card processing for
memberships and conferences (SuperComputing/95 was the first conference
to do on-line credit cards, I built that too because I was sick and
tired of keying in the cards myself. Laziness is next to godliness)
It served for years as the CS Digital Library core server with
30,000-40,000 accounts in active use. Man that thing was a truck, I wish
I knew what had happened to it.
And to think, it all started with the computer room ceiling collapsing
from all the RS232 cables from the Vax and crushing our Sun Sparc 20 web
server that kicked off this whole thing.
I should write a book or an article about that. We did so much that was
so... new... and all of that could be forgotten like tears in the rain....
CZ
On 7/16/2020 11:40 AM, Ali via cctalk wrote:
Funny story about that - I just setup a Systempro XL at home to play with. It is
fully decked out w/ dual processor 50MHZ 486s (not DX2), 512MB
of memory, a 4GB SCSI Boot Drive and six 2GB SCSI drives in RAID 5. The
Compaq systems came standard with what Compaq called the IDA (Intelligent
Drive Array). It was IDE based but did not use standard IDE drives. I think
it could do RAID 0, 1, and 3 (or the equivalents there of). Compaq even had
a few iterations of the controller and cached ones. Interestingly the
Systempro XL had a SCSI 2 controller on the MB mainly used for the tape
dive or CD while the base config came with an IDA 2 controller and could
have up to eight drives. In addition you could install extra IDA
controllers for even more drives or to drive external boxes. Or you could
upgrade to a SCSI array - which is what I have running in my Systempro XL.
but the
machine really was not meant for 2k. You could also run OS/2,
Novell Netware, Compaq DOS, and supposedly there was even a version of MS
LanMan (the full server OS not the client) for the Systempro that allowed
SMP.
Had a full compliment of memory,
max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
What OS, just out of interest?
Target OS was WinNT 3.1 initially and then 4.0. 2K was also supported
-Ali
10 a.m.
New subject: NCR 3550 Digital Library Was Re: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
That was it: MP-RAS. It was neat, kind of good, but to be honest Windows
NT 3.51 and 4.0 ran very well on it.
Just weighed a literal ton. For all I know it's still in the basement of
their Dupont Center office (now long closed)
C
On 7/21/2020 11:50 AM, Kevin Bowling wrote:
Wow, would love to have a machine like that.? The
?weird unix? was
probably MP-RAS which was NCR?s SysVr4.? NCR was selling massive x86 MCA
systems for Terradata setups in the early ?90s.
On Thu, Jul 16, 2020 at 8:54 AM Chris Zach via cctalk
<cctalk at classiccmp.org <mailto:cctalk at classiccmp.org>> wrote:
Now in terms of the most MANLY system I worked on, that would be the
NCR3550 we had at the IEEE Computer Society. When I arrived in 1993 it
had been donated, but was doing nothing with 4 486 CPUs in it and a
weird copy of AT&T unix. I took one look at the 256 bit interleaved
memory architecture the 3 levels of cache with affinity, the infinite
amount of space for disks, and the dual micro-channel busses and
fell in
*love*
We talked to NCR, upgraded it to 512mb memory, 8 Pentium Pro/200 CPUs,
and dual Microchannel busses with FDDI and Ethernet interfaces. Loaded
it with disks, installed Windows NT 4.0 on it, and turned it into
TALOS,
the main server for the IEEE Computer Society's Digital Library
(which I
built).
Partnered with Anderson and Netscape to multi-thread commerce server
(SSL), built an E-account system in Lotus Domino/Notes, and loaded up
all of our SGML with an SGML to HTML converter (Dynaweb) and a custom
tool that could convert Tek math to GIFs on the fly. That process could
take advantage of all 8 CPUs and render complex math articles in
real time.
Also did e-commerce for awhile with online credit card processing for
memberships and conferences (SuperComputing/95 was the first conference
to do on-line credit cards, I built that too because I was sick and
tired of keying in the cards myself. Laziness is next to godliness)
It served for years as the CS Digital Library core server with
30,000-40,000 accounts in active use. Man that thing was a truck, I
wish
I knew what had happened to it.
And to think, it all started with the computer room ceiling collapsing
from all the RS232 cables from the Vax and crushing our Sun Sparc 20
web
server that kicked off this whole thing.
I should write a book or an article about that. We did so much that was
so... new... and all of that could be forgotten like tears in the
rain....
CZ
On 7/16/2020 11:40 AM, Ali via cctalk wrote:
Funny story about that - I just setup a Systempro XL at home to play with. It
is fully decked out w/ dual processor 50MHZ 486s (not
DX2), 512MB of memory, a 4GB SCSI Boot Drive and six 2GB SCSI drives
in RAID 5. The Compaq systems came standard with what Compaq called
the IDA (Intelligent Drive Array). It was IDE based but did not use
standard IDE drives. I think it could do RAID 0, 1, and 3 (or the
equivalents there of). Compaq even had a few iterations of the
controller and cached ones. Interestingly the Systempro XL had a
SCSI 2 controller on the MB mainly used for the tape dive or CD
while the base config came with an IDA 2 controller and could have
up to eight drives. In addition you could install extra IDA
controllers for even more drives or to drive external boxes. Or you
could upgrade to a SCSI array - which is what I have running in my
Systempro XL.
supported but the
machine really was not meant for 2k. You could
also run OS/2, Novell Netware, Compaq DOS, and supposedly there was
even a version of MS LanMan (the full server OS not the client) for
the Systempro that allowed SMP.
? ?Had a full compliment of memory,
max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
What OS, just out of interest?
Target OS was WinNT 3.1 initially and then 4.0. 2K was also
-Ali
10:13 a.m.
New subject: NCR 3550 Digital Library Was Re: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
I (via Purdue surplus) had some NCR Worldmark 5500's (5 refrigerator sized
cabinets each with two MCA bus mulit-processor Pentium Pro systems, and a
total of 4TB of storage, back when that was about 1000 disks). I still have
the mutli-cpu Pentium box that was the management system for that, but it
hasn't been powered up in a long time.
On Tue, Jul 21, 2020 at 1:01 PM Chris Zach via cctalk <cctalk at classiccmp.org>
wrote:
That was it: MP-RAS. It was neat, kind of good, but to
be honest Windows
NT 3.51 and 4.0 ran very well on it.
Just weighed a literal ton. For all I know it's still in the basement of
their Dupont Center office (now long closed)
C
On 7/21/2020 11:50 AM, Kevin Bowling wrote:
Wow, would love to have a machine like that. The
?weird unix? was
probably MP-RAS which was NCR?s SysVr4. NCR was selling massive x86 MCA
systems for Terradata setups in the early ?90s.
On Thu, Jul 16, 2020 at 8:54 AM Chris Zach via cctalk
<cctalk at classiccmp.org <mailto:cctalk at classiccmp.org>> wrote:
Now in terms of the most MANLY system I worked on, that would be the
NCR3550 we had at the IEEE Computer Society. When I arrived in 1993
it
had been donated, but was doing nothing with
4 486 CPUs in it and a
weird copy of AT&T unix. I took one look at the 256 bit interleaved
memory architecture the 3 levels of cache with affinity, the infinite
amount of space for disks, and the dual micro-channel busses and
fell in
*love*
We talked to NCR, upgraded it to 512mb memory, 8 Pentium Pro/200
CPUs,
and dual Microchannel busses with FDDI and
Ethernet interfaces.
Loaded
it with disks, installed Windows NT 4.0 on
it, and turned it into
TALOS,
the main server for the IEEE Computer Society's Digital Library
(which I
built).
Partnered with Anderson and Netscape to multi-thread commerce server
(SSL), built an E-account system in Lotus Domino/Notes, and loaded up
all of our SGML with an SGML to HTML converter (Dynaweb) and a custom
tool that could convert Tek math to GIFs on the fly. That process
could
take advantage of all 8 CPUs and render
complex math articles in
real time.
Also did e-commerce for awhile with online credit card processing for
memberships and conferences (SuperComputing/95 was the first
conference
to do on-line credit cards, I built that too
because I was sick and
tired of keying in the cards myself. Laziness is next to godliness)
It served for years as the CS Digital Library core server with
30,000-40,000 accounts in active use. Man that thing was a truck, I
wish
I knew what had happened to it.
And to think, it all started with the computer room ceiling
collapsing
from all the RS232 cables from the Vax and
crushing our Sun Sparc 20
web
server that kicked off this whole thing.
I should write a book or an article about that. We did so much that
was
so... new... and all of that could be
forgotten like tears in the
rain....
CZ
On 7/16/2020 11:40 AM, Ali via cctalk wrote:
play with. It
is fully decked out w/ dual processor 50MHZ 486s (not
DX2), 512MB of memory, a 4GB SCSI Boot Drive and six 2GB SCSI drives
in RAID 5. The Compaq systems came standard with what Compaq called
the IDA (Intelligent Drive Array). It was IDE based but did not use
standard IDE drives. I think it could do RAID 0, 1, and 3 (or the
equivalents there of). Compaq even had a few iterations of the
controller and cached ones. Interestingly the Systempro XL had a
SCSI 2 controller on the MB mainly used for the tape dive or CD
while the base config came with an IDA 2 controller and could have
up to eight drives. In addition you could install extra IDA
controllers for even more drives or to drive external boxes. Or you
could upgrade to a SCSI array - which is what I have running in my
Systempro XL.
supported but the
machine really was not meant for 2k. You could
also run OS/2, Novell Netware, Compaq DOS, and supposedly there was
even a version of MS LanMan (the full server OS not the client) for
the Systempro that allowed SMP.
> Had
a full compliment of memory,
> max internal disk on the ATA controller,
ATA? That long ago?
Possible but unusual in a server, I would have thought.
Funny story about that - I just setup a Systempro XL at home to
What OS, just out of interest?
Target OS was WinNT 3.1 initially and then 4.0. 2K was also
-Ali
16 Jul
16 Jul
10:08 a.m.
New subject: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
On 7/16/20 9:40 AM, Ali via cctalk wrote:
there was even a version of MS LanMan (the full server
OS not the client)
Please elaborate on what you mean by "the full server OS".
My understanding is that Microsoft LAN Manager was an /add-on/ product
that could be installed /on/ /top/ /of/ an /existing/ server OS. I've
seen MS LAN Man in associateion with DOS, Windows (3.x / NT), and OS/2.
There is also IBM LAN Manager. I can't articulate the differences
between IBM and Microsoft LAN Manager. Much like I can't articulate the
differences between IBM OS/2 and Microsoft OS/2.
--
Grant. . . .
unix || die
11:07 a.m.
New subject: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
Please elaborate on what you mean by "the full server OS".
MS LanMan was Microsoft's networking OS of choice before NT. The base OS I believe was
based on MS OS/2 1.31.
Wiki has some more info: en.wikipedia.org/wiki/LAN_Manager
I have a copy of it somewhere. Here is a picture of the package:
3.bp.blogspot.com/-KaapV596wY0/WLSlOaIjSJI/AAAAAAAAAdU/VNNQtD-iqjs7ZkvuQGqD…
-Ali
17 Jul
17 Jul
11:38 a.m.
New subject: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
On 7/16/20 12:07 PM, Ali wrote:
MS LanMan was Microsoft's networking OS of choice
before NT. The base
OS I believe was based on MS OS/2 1.31.
Hum.
I'm trying to discern if it was it's own independent OS, or if it was
more a package of a COTS OS (OS/2) and LAN Man package, like Back Office
/ Small Business Server are Windows <version> Server OS and packages.
Wiki has some more info:
en.wikipedia.org/wiki/LAN_Manager
I have a copy of it somewhere. Here is a picture of the package:
3.bp.blogspot.com/-KaapV596wY0/WLSlOaIjSJI/AAAAAAAAAdU/VNNQtD-iqjs7ZkvuQGqD…
Hum.
Now I want to find a copy of it to mess with to see what exactly it is.
I really thought it was a product on top of a COTS OS.
Thank you for enlightening me Ali.
--
Grant. . . .
unix || die
18 Jul
18 Jul
2:31 p.m.
New subject: System Pro WAS RE: Compaq Smart Array 3200 Controller as a SCSI Controller
On Fri, Jul 17, 2020 at 2:38 PM Grant Taylor via cctalk <
cctalk at classiccmp.org> wrote:
On 7/16/20 12:07 PM, Ali wrote:
I believe that "OS/2 LAN Server" was part of "OS/2 1.3 Extended
Edition".
Extended Edition also included
some kind of SNA gateway, iirc, and probably other things I have zero
recollection of. The product had a
notion of "domains" at this point, although I don't know if this was a new
feature introduced with the OS/2
version or if previous products already had it.
-ken
MS LanMan was Microsoft's networking OS of
choice before NT. The base
OS I believe was based on MS OS/2 1.31.
I'm trying to discern if it was it's own independent OS, or if it was
more a package of a COTS OS (OS/2) and LAN Man package, like Back Office
/ Small Business Server are Windows <version> Server OS and packages.
16 Jul
16 Jul
8:20 a.m.
On 7/16/20 5:36 AM, Liam Proven via cctalk wrote:
It gets very complicated but it's also very
powerful and flexible.
Dedicated hardware just can't do stuff like this any more.
I largely agree for data stored on the systems.
However, PCs and compatibles, have long had an issue /reliably/ booting
across multiple drives. Especially with a soft failure on the primary
boot device.
As such, I find it much more convenient to have a hardware RAID
controller for at least the OS boot files. Then use all the fancy
feautes like you mention for data files.
--
Grant. . . .
unix || die
8:52 a.m.
With modern hardware, it's easier, cheaper and
more flexible to build
and manage arrays in software, using modern filesystems such as ZFS,
Btrfs, or MS Storage Spaces on Windows Server.
I have never used a SW RAID solution (except for a RAID 0 on Win2K3 for the boot drive)
and have used HW controllers in my more recent systems (I am particular to the Areca
Controllers - cheap but effective with a good feature mix). What I find problematic with
RAID (specially RAID 6) is that with the larger drives we have in use today build (or more
importantly rebuild/recovery) times are extremely long. Long enough that you could have a
second drive failure during that time based on statistics.
This is an article (for the layman) written in 2010 predicting the lack of usability of
RAID 6 by 2019: www.zdnet.com/article/why-raid-6-stops-working-in-2019/. I found the math
in it interesting and the conclusions pretty true to my experience.
I am wondering if SW RAID is faster in rebuild times by now (using the full power of the
multi-core processors) vs. a dedicated HW controller (even one with dual cores).
-Ali
10:21 a.m.
On 7/16/20 9:52 AM, Ali via cctalk wrote:
I have never used a SW RAID solution (except for a
RAID 0 on Win2K3 for
the boot drive)
Are you sure that was RAID 0 (zero), /striping/? I've never heard of
/software/ RAID 0 (striping) for the /boot/ drive in Windows. I would
expect that to be RAID 1 or something other than the drive with
NTLDR.EXE on it. I also suspect that the drive with %SystemRoot% on it
would need to more conducive to loading driver and software RAID support
files very early in the boot process.
and have used HW controllers in my more recent systems
(I am particular
to the Areca Controllers - cheap but effective with a good feature
mix).
I've completely lost track of hardware RAID controllers. I'm now more
interested in I.T. HBA controllers to use with ZFS based software RAID.
What I find problematic with RAID (specially RAID 6)
is that with
the larger drives we have in use today build (or more importantly
rebuild/recovery) times are extremely long. Long enough that you
could have a second drive failure during that time based on statistics.
That's one of the reasons that ZFS supports three drives worth of
redundancy in addition to the data space. RAID Z1 / Z2 / Z3.
I think we are quickly getting to the point, if not past it, where a
/single/ RAID array can't safely hold the entirety of the necessary
storage. Instead, I see multiple smaller RAID arrays aggregated
together at a higher layer.
I've seen this done by striping / JBODing / LVMing / etc. multiple
discrete RAID arrays together in the OS.
ZFS natively does this by striping (RAID 0) across multiple underlying
RAID sets (of whatever RAID level you want).
This is an article (for the layman) written in 2010
predicting the lack of usability of RAID 6 by 2019:
www.zdnet.com/article/why-raid-6-stops-working-in-2019/. I found
the math in it interesting and the conclusions pretty true to my
experience.
I am wondering if SW RAID is faster in rebuild times by now (using the
full power of the multi-core processors) vs. a dedicated HW controller
(even one with dual cores).
I think that the CPU overhead / computation time is now largely
insignificant. To me, one of the biggest issues is the simple massing
amount of data that needs to be read from and written to multiple
drives. At full interface speed, some drives can take a LONG time to
transfer all the data. What's worse is the sustained I/O speed to
platters of spinning rust being significantly slower than the interface
speed.
This is where some intelligence in the RAID implementation is really
nice. There is very little need to rebuild the yet unused area of a big
RAID array. ZFS shines in this as it only (re)builds the area that has
any data on it. Only have a few hundred GB on that multi TB RAID array
consisteng of multipel 1 TB drives? Fine. Only need to check the few
hundred GB. It's actually quite fast.
--
Grant. . . .
unix || die
10:58 a.m.
Are you sure that was RAID 0 (zero), /striping/?
I've never heard of
/software/ RAID 0 (striping) for the /boot/ drive in Windows. I would
expect that to be RAID 1 or something other than the drive with
NTLDR.EXE on it. I also suspect that the drive with %SystemRoot% on it
would need to more conducive to loading driver and software RAID
support
files very early in the boot process.
Absolutely correct. Proof reading good ;)! It was RAID 1.
That's one of the reasons that ZFS supports three
drives worth of
redundancy in addition to the data space. RAID Z1 / Z2 / Z3.
Interesting. Is there an official RAID level for three drive parity? The Areca controllers
do combined levels (e.g. 60 for two RAID 6 arrays stripped) but I don't think they do
mirroring of parity RAID levels.
I think that the CPU overhead / computation time is
now largely
insignificant. To me, one of the biggest issues is the simple massing
amount of data that needs to be read from and written to multiple
drives. At full interface speed, some drives can take a LONG time to
transfer all the data. What's worse is the sustained I/O speed to
platters of spinning rust being significantly slower than the interface
speed.
True. That is one of the points the article makes too. Basically, you can't get the
data fast enough but that would be inherent in both SW and HW implementations. The only
way to overcome that is to use SSDs I would think.
Only have a few hundred GB on that multi TB RAID
array
consisteng of multipel 1 TB drives? Fine. Only need to check the few
hundred GB. It's actually quite fast.
That is nice. I may have to look at it next time I do a RAID implementation.
-Ali
17 Jul
17 Jul
11:54 a.m.
On 7/16/20 11:58 AM, Ali wrote:
Absolutely correct. Proof reading good ;)! It was RAID
1.
I'm guessing that the cosmic ray flipped the bit in transit to the
mailing list server. :-D
Interesting. Is there an official RAID level for three
drive parity?
I'm not aware of one. But my ignorance thereof does not preclude it
from existing.
True. That is one of the points the article makes too.
Basically,
you can't get the data fast enough but that would be inherent in both
SW and HW implementations. The only way to overcome that is to use
SSDs I would think.
It's the same problem. It's just that the speed limit / goal post has
been moved further out, hopefully in front of the choke point.
That is nice. I may have to look at it next time I do
a RAID
implementation.
;-)
--
Grant. . . .
unix || die
11:58 a.m.
On Jul 17, 2020, at 2:54 PM, Grant Taylor via cctalk
<cctalk at classiccmp.org> wrote:
..
What is "three drive parity"? "Parity" is the description of RAID-5,
and 3 drive RAID-5 is certainly perfectly standard. RAID-1 is not parity, it's
mirroring. Is the question about triple mirroring, i.e., 3 drives all having the same
data on them? That's pretty rare though not unheard of, I've never seen a RAID-x
designation for that.
For high availability, RAID-6 is much more economical (and at this point the standard
choice); triple mirroring is of that class, with the difference that it performs better
for random short writes.
paul
Interesting. Is there an official RAID level for
three drive parity?
I'm not aware of one. But my ignorance thereof does not preclude it from existing.
1:19 p.m.
On 7/17/20 12:58 PM, Paul Koning via cctalk wrote:
What is "three drive parity"?
A poor choice of words in a complex topic.
How about "three drives worth of redundancy" Meaning that your data
will still be accessible if three drives fail.
ZFS has three versions of ZRAID or RAID-Z.
- RAID-Z1 is analogous to RAID-5.
- RAID-Z2 is analogous to RAID-6.
- RAID-Z3 is analogous to ???
I'm not aware of any official definition of a mirror of more than two
drives. I've heard of "3-way" / "4-way" / "n-way"
mirrors.
I think that the industry has settled on RAID-10 / RAID-01 and possibly
RAID-11 / maybe even RAID-00. But that isn't a standard to me.
Further, I see disagreements of what is the strip and what is the mirror
in RAID-10 / RAID-01.
"Parity" is the description of RAID-5, and 3
drive RAID-5 is certainly
perfectly standard. RAID-1 is not parity, it's mirroring.
If you think of it as "redundant drives" or "number of drives that can
fail without destroying data", then yes, RAID-1 does have a numerical
value quite similar to RAID-3 / RAID-5 / RAID-6 / RAID-Z* / etc..
Though nomenclature becomes highly problematic.
Is the question about triple mirroring, i.e., 3 drives
all having
the same data on them?
I was stating that I'm not aware of an official RAID level designation
for ZFS's RAID-Z3.
That's pretty rare though not unheard of, I've
never seen a RAID-x
designation for that.
I've known more than a few people to use n-way mirrors (~RAID-1).
Though I think I've only seen it in software.
For high availability, RAID-6 is much more economical
(and at this
point the standard choice); triple mirroring is of that class, with
the difference that it performs better for random short writes.
Are you comparing RAID-6 to triple (3-way) mirroring? Or something else?
I think that things get really weird and deep in minutia when you start
comparing a 3-way mirror to a 3 drive RAID-6. Same number of drives
(3), and same capacity (1 drive worth), and same fault tolerance (2
drive failures).
One of the other things that I've thus far neglected to mention about
ZFS is it's abilities to take snapshots and then subsequently send &
receive said snapshots* between pools / tape / image files. These are
some things that I think are nigh impossible to do with typical
traditional hardware RAID controllers. Sure, you might be able to do it
with systems that fall into the broad category of a higher end RAID
controller, but that is more typically a SAN controller which is
effectively it's own microcosm.
* You can also send / receive unmounted file systems** in addition to
snapshots of file systems.
** You can do similar with a zDevice, which is a logical block device
created by the ZFS pool.
--
Grant. . . .
unix || die
1:30 p.m.
On 7/17/2020 2:19 PM, Grant Taylor via cctalk wrote:
On 7/17/20 12:58 PM, Paul Koning via cctalk wrote:
Does modern software even let you know if a drive is failing?
Ben.
What is "three drive parity"?
A poor choice of words in a complex topic.
How about "three drives worth of redundancy"? Meaning that your data
will still be accessible if three drives fail.
1:36 p.m.
On 7/17/20 2:30 PM, ben via cctalk wrote:
Does modern software even let you know if a drive is
failing?
I'm not aware of any software RAID solution jumping up and proactively
notifying you that there's a problem /by/ /default/.
But I think the same thing is largely true about hardware RAID
solutions. The only caveat being an alarm that might sound on a
hardware RAID solution. But that assumes that someone is in proximity
to hear it.
All hardware and software RAID solutions that I've worked with do expose
the information. They all also have software that can be used to check
and then take some action to notify people.
--
Grant. . . .
unix || die
1:42 p.m.
On 2020-07-17 5:36 p.m., Grant Taylor via cctalk wrote:
On 7/17/20 2:30 PM, ben via cctalk wrote:
I had a disk that went over a SMART threshold that was part of a mdadm
software mirror and it posted a message every time I rebooted warning of
impending doom.
The systems I support professionally do try to warn the admins, but it
seems that frequently no notice is taken until the level of redundancy
is exceeded at which time the issue becomes more difficult to ignore,
and then begins the waling and gnashing of teeth......
Paul.
Does modern software even let you know if a drive
is failing?
I'm not aware of any software RAID solution jumping up and proactively
notifying you that there's a problem /by/ /default/.
But I think the same thing is largely true about hardware RAID
solutions.? The only caveat being an alarm that might sound on a
hardware RAID solution.? But that assumes that someone is in proximity
to hear it.
All hardware and software RAID solutions that I've worked with do
expose the information.? They all also have software that can be used
to check and then take some action to notify people.
2:04 p.m.
On 7/17/20 2:42 PM, Paul Berger via cctalk wrote:
I had a disk that went over a SMART threshold that was
part of a mdadm
software mirror and it posted a message every time I rebooted warning of
impending doom.
That sounds like the typical S.M.A.R.T. alert from the BIOS. Not
something from the kernel. If it was the kernel, I'd like to know more
about it.
The systems I support professionally do try to warn
the admins, but it
seems that frequently no notice is taken until the level of redundancy
is exceeded at which time the issue becomes more difficult to ignore,
and then begins the waling and gnashing of teeth......
Yep. I suspect many, if not most, of us have been in that teeth
gnashing situation.
--
Grant. . . .
unix || die
3:33 p.m.
Oh yeah. I remember having a DS4300 that had a failed drive and when we
tried to rebuild it corrupted the whole string. Call to IBM ensued where
they had us disable one controller, then the other controller was able
to rebuild the array, then we brought the first controller back in sync.
Tricky....
C
On 7/17/2020 5:04 PM, Grant Taylor via cctalk wrote:
On 7/17/20 2:42 PM, Paul Berger via cctalk wrote:
I had a disk that went over a SMART threshold
that was part of a mdadm
software mirror and it posted a message every time I rebooted warning
of impending doom.
That sounds like the typical S.M.A.R.T. alert from the BIOS.? Not
something from the kernel.? If it was the kernel, I'd like to know more
about it.
The systems I support professionally do try to
warn the admins, but it
seems that frequently no notice is taken until the level of redundancy
is exceeded at which time the issue becomes more difficult to ignore,
and then begins the waling and gnashing of teeth......
Yep.? I suspect many, if not most, of us have been in that teeth
gnashing situation.
3:58 p.m.
On 7/17/20 4:33 PM, Chris Zach via cctalk wrote:
Oh yeah. I remember having a DS4300 that had a failed
drive and when we
tried to rebuild it corrupted the whole string. Call to IBM ensued where
they had us disable one controller, then the other controller was able
to rebuild the array, then we brought the first controller back in sync.
Tricky....
Yep.
If I'm going to be dealing with that type of oddness, I'd rather do it
on my preferred OS of choice than something embedded that I don't have
much experience with.
--
Grant. . . .
unix || die
4:24 p.m.
On 7/17/2020 6:58 PM, Grant Taylor via cctalk wrote:
If I'm going to be dealing with that type of
oddness, I'd rather do it
on my preferred OS of choice than something embedded that I don't have
much experience with.
On the other hand I never lost any data on my IBM DS4300 or 4700 SANs.
The tech support was excellent and we beat the hell out of those things.
Then again I never had a problem at all with Compaq SMART arrays. The
key was to have the SMART software installed and running.
1:39 p.m.
On Jul 17, 2020, at 4:19 PM, Grant Taylor via cctalk
<cctalk at classiccmp.org> wrote:
On 7/17/20 12:58 PM, Paul Koning via cctalk wrote:
Yes. Marketeers like to use the phrase "erasure coding" as a magic token to
indicate great wonderfulness beyond RAID. Actually, it's a math term that describes
the standard RAID systems plus generalizations. I don't know if there is a clear
terminology for the general thing. I tend to call them "k of n" codes, k data
bits, n message bits. Or in the RAID case, k drives worth of data, n total drives. That
makes RAID-1 a 1 of 2 erasure code, the thing we started talking about 1 of 3 erasure
code, RAID-5 is n of n+1 erasure code, and RAID-6 is n of n+2 erasure code. So in the k
of n notation, n-k drives can fail before you lose your data.
Actually, more realistic with today's drive sizes: n-k-1 drives can fail and one drive
can have a hard read error during reconstruction before you lose data. For sufficiently
large drives, the probability of a hard read error somewhere is nearly 1, which is why
RAID-5 is only a good idea for small disk drives.
People have build storage systems with other erasure codes, for example n of n+3. Not
because it's really necessary but as a research experiment; the "Self-star"
system at CMU comes to mind.
What is "three drive parity"?
A poor choice of words in a complex topic.
How about "three drives worth of redundancy" Meaning that your data will still
be accessible if three drives fail.
ZFS has three versions of ZRAID or RAID-Z.
- RAID-Z1 is analogous to RAID-5.
- RAID-Z2 is analogous to RAID-6.
- RAID-Z3 is analogous to ???
I'm not aware of any official definition of a mirror of more than two drives.
I've heard of "3-way" / "4-way" / "n-way" mirrors.
I think that the industry has settled on RAID-10 / RAID-01 and possibly RAID-11 / maybe
even RAID-00. But that isn't a standard to me.
Further, I see disagreements of what is the strip and what is the mirror in RAID-10 /
RAID-01.
"Parity" is the description of RAID-5,
and 3 drive RAID-5 is certainly perfectly standard. RAID-1 is not parity, it's
mirroring.
If you think of it as "redundant drives" or "number of drives that can
fail without destroying data", then yes, RAID-1 does have a numerical value quite
similar to RAID-3 / RAID-5 / RAID-6 / RAID-Z* / etc.. Though nomenclature becomes highly
problematic. Is the
question about triple mirroring, i.e., 3 drives all having the same data on them?
I was stating that I'm not aware of an official RAID level designation for ZFS's
RAID-Z3.
That's pretty rare though not unheard of,
I've never seen a RAID-x designation for that.
I've known more than a few people to use n-way mirrors (~RAID-1). Though I think
I've only seen it in software.
For high availability, RAID-6 is much more
economical (and at this point the standard choice); triple mirroring is of that class,
with the difference that it performs better for random short writes.
Are you comparing RAID-6 to triple (3-way) mirroring? Or something else?
I think that things get really weird and deep in minutia when you start comparing a 3-way
mirror to a 3 drive RAID-6. Same number of drives (3), and same capacity (1 drive worth),
and same fault tolerance (2 drive failures).
2:18 p.m.
On 7/17/20 2:39 PM, Paul Koning via cctalk wrote:
"too many drives failed at the same time"
IMHO, this encompases hard read errors. For what ever reason, the drive
was not able to priovide the data within a reasonable time period.
Be it physical drive failure, hard read error rate, or someone
disconnecting the drive.
--
Grant. . . .
unix || die
16 Jul
16 Jul
10:31 a.m.
On Thu, Jul 16, 2020 at 08:52:16AM -0700, Ali via cctalk wrote:
[...]
This is an article (for the layman) written in 2010
predicting the lack of
usability of RAID 6 by 2019:
www.zdnet.com/article/why-raid-6-stops-working-in-2019/. I found the math in
it interesting and the conclusions pretty true to my experience.
The author screwed up his maths and also made faulty assumptions.
The article states that "SATA drives are commonly specified with an
unrecoverable read error rate (URE) of 10^14. Which means that once every
200,000,000 sectors, the disk will not be able to read a sector." and then "2
hundred million sectors is about 12 terabytes." It seems he is using a sector
size of 64kiB. Standard SATA disks have 4kiB sectors.
"At that point the RAID reconstruction stops". Maybe on his garbage hardware
RAID controller with 64kiB stripes which chokes on a single-bit error in a
stripe because it's too dumb to figure out which disk is lying. ZFS is somewhat
smarter than that.
I am wondering if SW RAID is faster in rebuild times
by now (using the full
power of the multi-core processors) vs. a dedicated HW controller (even one
with dual cores).
Not only is software RAID faster now, but this has been the case for at least
15 years. The necessary calculations are trivially vectorisable and are usually
limited by memory bandwidth. Which is several orders of magnitude faster than a
hard disk.
17 Jul
17 Jul
4:48 a.m.
On Wed, Jul 15, 2020 at 01:09:21PM -0700, Ali via cctalk wrote:
[Hardware RAID controllers]
On the "no good use case" front:
I avoid hardware RAID controllers for a variety of reasons, which mostly boil
down to the use of proprietary firmware and na?ve RAID implementations. These
also apply to many software RAID implementations which blindly copied them.
The biggie is that proprietary RAID means proprietary on-disk formats. If the
controller fails, you need to find a replacement which understands the old
on-disk format. Good luck with that. Related is the generally shoddy nature of
firmware, and it's usually hard-to-impossible to e.g. query the SMART status of
individual disks.
The next-biggie is the RAID Write Hole. A traditional RAID implementation will
rewrite data you might consider to be at rest because it shares a stripe with
newly-written data, and on failure can corrupt said at-rest data. This is a
fundamental problem which hardware RAID controllers try to mitigate by having a
battery backup unit to deal with power failures, and can (potentially) also
work independently of an OS which crashed mid-write, but it doesn't really
solve it. What if your power stays out longer than the battery lasts?
Software RAID which implements traditional RAID cannot even apply this
mitigation and this is one of the reasons it has a bad reputation. The obvious
solution is to not implement tradional RAID, which is where ZFS and similar
copy-on-write journalled filesystem-cum-volume-managers come in.
The last bastion of hardware RAID controllers was if one was using a toy
operating system such as Windows where the software RAID options were woeful or
nonexistent, but it now has Storage Spaces.
On the "suddenly quite cheap" front: Plain SAS controllers based on e.g. the
LSI 9207 or 9211 are north of ?100, whereas MegaRAID controllers based on the
LSI 9260 have plummeted to ?30. The former either supports pass-through mode
out of the box or after reflashing with "IT" firmware, but the latter does not.
ZFS does work atop RAID (of either flavour), but is more robust if it can
manage the raw disks directly. A workaround with hardware RAID cards which
won't do pass-through is to configure them with single-disk RAID0 volumes, but
this is somewhat untidy and still has the problem of proprietary on-disk
formats and general inscrutability.
There is no
good use case for them in 2020, which is why they're all
suddenly quite cheap.
Why do you say that? Not disagreeing per se but just
wondering the reasoning
behind it.
1615
days inactive
1622
days old
45 comments
15 participants
participants (15)
-
abuse＠cabal.org.uk -
bear＠typewritten.org -
bfranchuk＠jetnet.ab.ca -
binarydinosaurs＠gmail.com -
cctalk＠gtaylor.tnetconsulting.net -
cctalk＠ibm51xx.net -
cz＠alembic.crystel.com -
dave.g4ugm＠gmail.com -
jason＠smbfc.net -
kevin.bowling＠kev009.com -
kgober＠gmail.com -
lproven＠gmail.com -
pat＠vax11.net -
paulkoning＠comcast.net -
phb.hfx＠gmail.com