Mailinglist Archive: opensuse (982 mails)

< Previous Next >
Re: [opensuse] Re: Hard Disk Upgrades
  • From: Greg Freemyer <greg.freemyer@xxxxxxxxx>
  • Date: Thu, 11 Sep 2014 16:34:30 -0400
  • Message-id: <CAGpXXZL7M=9nHg1aaSmbNVUOgfYhW+e2nwp_TZfpUgq2SvA=VA@mail.gmail.com>
On Thu, Sep 11, 2014 at 3:47 PM, Linda Walsh <suse@xxxxxxxxx> wrote:
Greg Freemyer wrote:

You and I have had different experiences. The most obvious is your
statement about Linux seeing a fakeraid as a single disk. With the
fakeraid I've attempted to use, Linux sees each disk and has to be
made to recognize the fakeraid exists and read the raid config out of
the bios, then take the responsibility of managing the disks just as
it does with software raid.

----
Well, this is likely a terminology issue.

In my case, the "fake raid" in the Dell BIOS shows
only 1 disk to linux -- something like 'Dell Virtual Disk [00-xx]'
that act like new physical units (drive C, D, etc..)

It is a solution you get w/o paying extra for a RAID card
but that has *no hardware acceleration*... but the BIOS presents
a RAID0 or RAID1 set as 1 disk for normal interactions.

Note: if I go into /sys, I can find entries corresponding
to each disk under the software controller device -- but I can see
those under a HW-raid device as well.

It is likely that all 'fake raid' solutions are not
created equal. The software-only BIOS-based raids that come with
some BIOS's, may not, in some sense, be 'fake', but may be a
software RAID solution that exists in the BIOS and can only manage
disks in the system (no way to add on external disk(s) unless your
box has multiple eSATA plugs).

The things I liked -- is that it's all handled pre-boot
so @boot OS's think there is one HD there, but unlike HW RAID
solutions, there is no HW to do RAID[5,6,50,60..etc], so it
is limited to whatever commands the HBA can do in parallel on the
disks it manages (usually internally only).


OTOH! -- I've seen HW-accelerated RAIDS where the OS
SAW each local HD, but only accessed the 'group' where it got
benefit of HW-based disk-sum calculations.

That's why I answered the way I did. If the OS's
don't need a separate dev driver and think they are talking to
some generic sata or sas drive (although, perhaps, a bit large) --
even if there is no HW card, I'm not sure I'd call that "fake raid"...
though many may.

I.e. the lines, surprisingly, may be a bit blurry?

I'm not familiar with the on motherboard raid setup your describing.
It seems very close to hardware raid. Also, I've paid good money for
a 3-ware raid controller that could only do Raid-0 and 1. It was
still hardware raid.

Fyi: fake raid works by implementing custom firmware in the bios I/o
interface used during boot. Thus Windows and Linux only see a single
drive during the boot process, but for Linux as soon as the normal
ATA/scsi drivers kick in the bios is bypassed and the kernel has to
manage the raid array itself. dmraid (as opposed to MD raid or
mdraid) is the Linux package that is responsible for reading the
config out of the bios and setting up the Linux kernel to manage the
disks correctly.

===
Are you saying that the BIOS could not provide an virtual 'HW RAID'
interface where the BIOS handled/managed the HD's and doesn't use
the.

As I understand it, the fake raid I'm familiar only provides a psuedo
raid interface via the bios INT13 (
http://en.wikipedia.org/wiki/INT_13H).

INT13 has historically been used during the initial boot process to
access disks. It is a low performance interface, but it is consistent
allows for simplicity in initial boot code.

As I understand it, once the true linux kernel takes over i/o to the
drives, it no longer uses the INT13 interface which means for fake
raid there is no longer any raid functionality provided.

Thus in reality, once the boot process is done, fake raid becomes true
software raid, but without all the flexibility md and mdadm offers.


Due to that, dmraid has to have knowledge of each type of fakeraid
controller and is more likely to have bugs/missing support.

---
There is no dmraid module needed on _these_ SW-BIOS based raid solutions.
Used to be I didn't even include any of the kernel modules for
any of the RAID varieties. That's what I mean by OS independence.


If you need neither dmraid or md / mdadm, then what you have is either
real hardware raid or a very much improved fake raid.


md raid (pure software raid) on the other hand only has to interpret
its own data structures, so there is more flexibility.

---
Before I started including more modules as "options",
I didn't include any linux raid related modules.



You seem to be talking about hardware raid which uses neither dmraid
nor mdraid. I agree hardware raid in general is the best, but it
normally comes with a decent price tag. The cheapest 2-disk hardware
raid 1 controller I've bought was over $100 for just the card.

---
Well, I agree it shares elements w/a HW RAID card, but it has
no HW (thus limited to RAID0 or RAID1 (I'm not sure about RAID10).



I've seen and had linux SW RAID5 (as well as HW), and with the same
power and software interruptions, the SWRAID was more vulnerable to
hard
corruptions. W/SW RAID, you can't have a battery-backed up ram that
will
write to the disk when the OS comes up -- because the memory will be
purged.


To my knowledge, battery-backed up ram is a unique feature of true
hardware raid. Your mentioning it continues to make me wonder if we
are talking about the same thing.

---
There is no need for a battery backed up cache, since
what is written on disk can't be disintegrous -- incomplete, maybe
but if some part of a file is written and the crash happened before
the 2nd DMA did its copy, on recovery, the missing data would be
ignored and used from the copy that did get written and be other copy
would be written after read.



Have you ever used dmraid? (again, that is different than md, mdadm,
etc.)

---
Not that I can remember. I believe it was mdraid.. dmraid
would be like using (not sure how different) lvm to create a stripped
or mirrored volume, no?b

I believe dmraid is primarily a discovery tool. It interogates the
system for fakeraid controllers and it can also look at metadata
blocks on the drives themselves to look for fakeraid signatures. If
it determines fakeraid is in use it does it's best to determine the
appropriate configuration and tells the kernel how to properly
interact with the drives. Here's the list of things dmraid claims to
support:

$ /sbin/dmraid -l
asr : Adaptec HostRAID ASR (0,1,10)
ddf1 : SNIA DDF1 (0,1,4,5,linear)
hpt37x : Highpoint HPT37X (S,0,1,10,01)
hpt45x : Highpoint HPT45X (S,0,1,10)
isw : Intel Software RAID (0,1,5,01)
jmicron : JMicron ATARAID (S,0,1)
lsi : LSI Logic MegaRAID (0,1,10)
nvidia : NVidia RAID (S,0,1,10,5)
pdc : Promise FastTrack (S,0,1,10)
sil : Silicon Image(tm) Medley(tm) (0,1,10)
via : VIA Software RAID (S,0,1,10)
dos : DOS partitions on SW RAIDs

Greg
--
To unsubscribe, e-mail: opensuse+unsubscribe@xxxxxxxxxxxx
To contact the owner, e-mail: opensuse+owner@xxxxxxxxxxxx

< Previous Next >
Follow Ups