home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
linuxmafia.com 2016
/
linuxmafia.com.tar
/
linuxmafia.com
/
kb
/
Hardware
/
sata.add
< prev
next >
Wrap
Text File
|
2008-02-25
|
34KB
|
1,018 lines
Ario Data Networks boards: SATA/SAS OEM supplier. Very expensive.
Ignore for now.
Silicon Image's SiI 4723, the industry's first storage processor to
offer 3Gb/s port speeds with hardware RAID, was announced 2005-08-24.
(Not included in products, yet.) At the same time, they announced
"MSLPhy Serializer/Deserializer (SerDes)" chip technology that will
convert between serial and parallel at bandwidths from 1.5 to 6.0 Gb/s,
and expect products to ship with such chips in early 2006.
Article about eSATA external-connection extension:
http://www.windowsitpro.com/Article/ArticleID/47543/47543.html?Ad=1
PMC-Sierra Inc.: Port multiplexer cards, SATA/SAS disk array
enclosures.
Ignore references to Silicon Image SiI 3726 chip: It's a SATA-II port
multiplier.
Date: Mon, 5 Jun 2006 21:03:24 +0100
From: Ronan <mr_platelet@fastmail.fm>
To: rick@linuxmafia.com
Subject: SATA on Linux: info on VIA VT8251 southbridge chipset
Hi.
As the unfortunate owner of a motherboard containing a VT8251
southbridge, I thought I might send you a note about this chipset,
which you may find useful for your page on SATA support in Linux
(http://linuxmafia.com/faq/Hardware/sata.html#via8251).
It seems that this chipset is not supported in standard Linux. VIA have
created a patch which enables Linux to work with it, but Jeff Garzik has
chosen not to use the patch in its current form. The thread beginning
here has more:
http://marc.theaimsgroup.com/?l=linux-ide&m=113436067230517&w=2
I've checked the current release of the kernel (2.6.16.19), and it still
doesn't seem to have any code for the chipset.
Regards,
Ronan.
Date: Tue, 13 Dec 2005 14:38:05 -0600
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: rick@linuxmafia.com
Subject: sata notes http://linuxmafia.com/faq/Hardware/sata.html#hp
I do lots of playing with SATA drivers and HBAs under Centos 4.2 / Linux
2.6.14 on x86_64. I've settled on the RocketRAID 2220 (uses a vendor
supplied hptmv6 driver), as it delivers 700+ MB/sec off 16 drives
(Seagate 200 or 250s). The Adaptec card (uses aacraid) delivers 460 at
best, with the same drives. The Broadcom bs4852 does OK -- 550MB/sec and
up -- but they slow down as you add more HBAs since the driver has to
get the big kernel lock for each interrupt. Broadcom supports RAIDs
that span cards in software. Highpoint at least has a ribbon cable.
Within a week, I'll be using the LSI PCI-Express stuff, and the
Highpoint 2320 PCI-Express HBAs. I'll let you know how they do.
PS: Nice Web page.
Berkley
Date: Thu, 16 Feb 2006 16:31:16 -0600
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: want to do SAS as well as SATA?
I'm finally getting to Serial Attached SCSI adapters. They are really
too new to have a lot of support yet, but the ones I am familiar with
will talk to standard SATA drives. I tried the Adaptec 4-port 48300
PCI-X SAS controller. No Linux driver unless you use the stock Red Hat
2.6.9-foobar kernel. They didn't want to do that either. This controller
had no issues configuring a RAID of SATA or SAS drives. Dropped right
in, and worked. Bummer, it would only talk 150MB/sec to SATA-2 drives.
It even said so when the BIOS ran. It might get fixed in a later BIOS
rev :-) There was a fire sale on the HBA and two Seagate 15K RPM drives.
$500 for the set including cables. Windows was supported.
I have an LSI-4808E 8-port PCI-Express 8-lane controller coming
(March 10), which my inside source said actually works.
The SAS expander and all SAS is really a 10-GigE switch and some
extra headers. So one port can sustain I/O from 3-4 drives at once.
So an 8-port should be able to handle 8 * 3 * 65MB/Sec or better.
I'll be able to give a performance update then, if you are interested.
Maybe a SATA on SAS entry? I'm very interested in how much I get back
from a real hardware RAID.
The Areca PCI-Express 16-port card struggled to get to 200MB/sec under
Linux. When I waded through the non-English speak of their response, it
turns out that the Intel IOP they used has a _max_ of 400MB/sec regardless
of the number of drives you have. Something about 128K of buffer space
the slow IOP can run XORs on. So, the drives start/stop a lot :-( They
said their next generation product would use a *real* Intel XOR engine
:-)
berkley
Date: Fri, 19 May 2006 07:30:42 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: Adaptec 4805SAS again
So I talked to a "source" at Adaptec, she reports that the 150Gb/sec is
a chip error and they are re-spinning the chip to fix it. New HBAs out
late summer. RAID5 performance - dismal. 330MB/sec RAID5 per 8-port
direct connect. System overhead at 660MB/sec total is still < 20%,
which is much better than the RocketRAIDs. The card uses the Intel
IOP333 @ 500MHz, which in turn has a PCI-X to PCI-E bridge on it. This
slows everything down :-)
As it stands, the adapter will not handle the SAS expanders and give
excellent performance, since the channels are limited to 150Gb/Sec. 300
- 420MB/sec off good drives per HBA is still nothing to sneer at. A mite
slow for my tastes :-) On the plus side, aacraid.ko did its job and
performed OK on the Opteron.
Berkley
--
//E. F. Berkley Shands, MSc//
Date: Wed, 17 May 2006 11:12:23 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: SATA/SAS controller Adaptec 4805SAS
I gave up waiting for the LSI card.
This weeks nightmare is the Adaptec AIC9405W project, called the 4805SAS
PCI-E SAS controller. Two of them, BIOS rev 9117 is required (shipped
with 8832). Yes, it sees SATA drives. No, it runs them at only
150MB/sec, not 300MB/sec for SATA-2 drives. The Linux GUI tools are
worse than sucky. I used Windows to do my builds :-)
Still I am able to get 877MB/sec off 16 drives with the aacraid native
driver in 2.6.16/2.6.17. The aic94xx driver does not like my BIOS on the
cards. Write performance is poor, at ~90MB/sec per 4-drive RAID0.
System overhead for the reads is about 20% vs 60% for the RR2320.
Windows XP-64 runs at about 400MB/sec reads, 50MB/sec per RAID write
rate. I am trying the RAID5 performance now, but it takes > 6 hours to
do a build/format :-)
FYI: the Seagate 7200.10 drives transfer at 88MB/sec on the outside
edges. The 7200.9 (now discontinued) did 58.7MB/sec on the outside
edges. So, my 16-drive RR2320 solution hits 1.3GB/sec for the first TB
of data, then slows to ~900MB/Sec for the next TB.
Berkley
--
//E. F. Berkley Shands, MSc//
t35t0r <t35t0r@gmail.com>:
> All LSI Logic Megaraid 83xx, 84xx, 87xx, and 88xx SAS/SATA cards are
> hardware raid. 82xx are fakeraid. All use megaraid_sas drivers.
>
> http://lsi.com/storage_home/products_home/internal_raid/megaraid_sas/value_line/index.html
> http://lsi.com/storage_home/products_home/internal_raid/megaraid_sas/feature_line/index.html
>
> sata can be used with sff8087 to discrete breakout cables (e.g.
> http://www.ewiz.com/detail.php?name=CBLSFF10M)
83xx:IOP333
84xx:IOP333
87xx, 88xx,
8204 4-port SAS
8208 8-port SAS
MegaRAID SAS 8204ELP, 4 internal, PCI-E
MegaRAID SAS 8204XLP, 4 internal, PCI-X
MegaRAID SAS 8208ELP, 8 internal, PCI-E
MegaRAID SAS 8208XLP, 8 internal, PCI-X
MegaRAID SAS 8308ELP, 8 internal, PCI-E, Intel IOP333
MegaRAID SAS 8300XLP, 8 internal, PCI-X
MegaRAID SAS 8308ELP. 8 internal, PCI-E, Intel IOP333
MegaRAID SAS 8344ELP, 4 internal, 4 external, SAS, Intel IOP333, PCI-E
MegaRAID SAS 8408E, 8 external, PCI-E, Intel IOP333
MegaRAID SAS 84016E, 16 internal, PCI-E, Intel IOP333
MegaRAID SAS 8408E, 8-port RAID Intel IOP333
MegaRAID SAS 8480E, 8 external, Intel IOP333
MegaRAID SAS 8704ELP, 4 internal, LSISAS1078 IOP
MegaRAID SAS 8708ELP, 8 internal, LSISAS1078 IOP
MegaRAID SAS 8888ELP, 8 ports (4 internal/external), LSISAS1078 IOP, PCI-E
Accusys's 64-bit ACS-93000 SAS-SAS RAID controller supports SATA
native command queuing (NCQ) and SCSI tagged command queuing (TCQ), and
has dual PCI-X connections, up to 16 SAS drive channels, a 533MHz RISC
I/O processor, up to 4GB DDR2 cache memory, support for all RAID levels,
online disk group defragmentation and LUN capacity expansion, and
support for Windows Virtual Disk Service (VDS) and Multi-Path I/O
(MPIO).
Applied Micro Circuits Corp. (AMCC) model 9650 SATA
model 9690SA-8I, with eight internal ports
model 9690SA-8E, with eight external ports
model 9690SA-4I4E, with four internal ports and four external ports
All use 3ware 9690SA controller chip, based on the 500S I/O controller
chip from Emulex,
From: Sandro Pedrocchi <sandro.pedrocchi@mitlinks.ch>
To: rick@linuxmafia.com
Date: Mon, 6 Aug 2007 17:38:41 +0200
Subject: TX4300
Hi
Promise does not provide a driver for the TX4300 for kernel 2.6. But the
card works fine with dmraid. Much better than with the promise ftsata2
driver and kernel 2.4.
Greets
Sandro Pedrocchi
Entwicklung
+41 (0)44 444 10 38Fon
+41 (0)44 444 10 45 Fax
sandro.pedrocchi@mitlinks.ch
mitLinks AG . DAS NEUE MEDIEN HAUS
Limmatstrasse 291 . CH-8005 Z#ich . http://www.mitlinks.ch
Date: Thu, 28 Jun 2007 15:55:29 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: rocket raid 2340 / LSi8888ELP / Adaptec 31605
Finally got around to hacking at the RR2340, 16 port Sata-2
controller.
PCI-e 8X. Drop in compatible with the other RR2* series. Disks move
between controllers just fine. 950MB/Sec peak write rate into
16 Seagate 320GB 7200.10's. Read back at 1098MB/Sec. Raid5 writes are
about 25% the raid0 rates, but read back is still fast.
Best read is at 145KB per raid, and 4 raid0s.
Absolutely no tuning needed, and the linux driver patched and compiled
into 2.6.20/2.6.21/2.6.22-rc6 without a peep.
Still a bit slower than the LSI8888ELP, which ROCKS!
The 8888ELP SAS controller drives 16 drives at line rate (1.2GB/Sec).
When used with an AIC external SAS expander, the bottom drops out to
540MB/Sec. Morons. The TSTCOM 16 drive 3U is $1,400 (about 50%) the cost
of the AIC, and comes with standard rails that fit.
This controller shows motherboard limitations (I.E. if the HT link is busy
then performance peaks about 1GB/Sec) but can get LOTS faster
if the HT link is idle. Current BIOS is a must! No kernel tweeks.
Worked out of the box with 2.6.20, 21, 22-rc6, ...
Raid5 speeds are pretty good. 70% raid0 writes.
The Tyan 3992 Motherboard showed limits with 2 8888elp's on board.
If you stress the HT links, you kernel panic, trash file systems,
blue screen (winDoze). Capable of 1.4GB/Sec (until the kernel
dies). This happens within seconds of hitting peak write rates.
Kernel structures get overwritten. Oops! it isn't the driver :-)
Likely the ServerWorks Chipset.
The SuperMicro H8DMi gets 1.2GB/Sec, but does not kernel panic.
But since it has one HT link for all I/O, the real throughput drops
when you do something with that data. You drop to 920MB/Sec if you
touch the data (like send it to another device).
Adaptec 31605 16 port SAS controller. Well, I hope you only wanted one or
two raids per controller. Throughput drops by 50% if you have more
than one raid.
I guess they though everyone wants a single 8TB spindle.
But that does 790MB/Sec (since there is a PCI-e to PCI-X bridge
inside).
that is pretty darn good.
berkley
--
E. F. Berkley Shands, MSc
Date: Wed, 20 Jun 2007 13:35:10 -0700
From: Rick Moen <rick@linuxmafia.com>
To: Pekka Hellen <pekka@hellen.fi>
Subject: Re: Fakeraid
Quoting Pekka Hellen (pekka@hellen.fi):
> Im wondering that what is the situation with Silicon Image 4723, is
> that fakeraid or not? :)
>
> btw great list you have there
> http://linuxmafia.com/faq/Hardware/sata.html#fakeraid
>
> Best regards
> Pekka Hell#
> Finland
> **
The 4723 appears to be genuine hardware RAID (albeit for RAID0 or RAID1
only), and is implemented as a port multiplier and RAID controller
add-on, that attaches to a SATA port, which is then for some reason
referred to as the "EZ-Backup" port, to which you can connect two SATA
drives. Unfortunately, reports I hear suggest that it has very bad
performance.
My apologies for being a bit behind on maintenance of my SATA on Linux
page. I've been away on vacation, and am just now returned.
Date: Wed, 23 May 2007 14:27:12 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: adaptec 31605 16 port sas controller
I was given a new card. 16 SAS lanes, 4 - 4 lane mini SAS connectors, internal
pci-e X 8. Adaptec 31605 256MB cache. Uses a PCI-e to PCI-X internal bridge chip.
I tried 8 seagate 500GB NS series drives in a raid5, 256KB stripe
size.
using AACRAID I wrote at 40MB/Sec into an XFS file system.
using AIC94XX I wrote at 38MB/Sec, and read back at 335MB/Sec.
Not particularly impressive. I'm headed to try the raid0 (2 X 4
drive) next.
berkley
--
E. F. Berkley Shands, MSc
Date: Wed, 23 May 2007 12:43:34 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: seagate gotcha!
hmmm... some crow to eat :-)
seagate ships ALL its current drives with a jumper on them.
That jumper restricts the drive to 1.5Gb/Sec.
Take it off and things run a LOT better.
Now I have to re-run a weeks worth of benchmarks+.
gumble... growl... snap...
berkley
--
E. F. Berkley Shands, MSc
Date: Wed, 16 May 2007 12:19:55 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: high point rr2340 numbers
Using my usual setup, I get dismal results from the rr2340, 16 port
sata controller (pci-e by 8 lanes).
550MB/Sec writes into 16 drives, 1.06GB/Sec reads from the same
drives.
kernel patch into 2.6.20 worked great. the gui functioned well.
The 4-lane connectors are nice, less clutter in the chassis.
The controller reports the seagate 500GB NS series 7200.10 drives
are running at sata-150, not sata-300. No response from highpoint
on that yet.
berkley
--
E. F. Berkley Shands, MSc
Date: Wed, 14 Mar 2007 13:36:59 -0500
From: "Mr. Berkley Shands" <bshands@exegy.com>
To: Rick Moen <rick@linuxmafia.com>
Subject: high point RR2340 16 port sata controller
I obtained and tested the @@2340 16 port PCI-e sata controller from
highpoint.
I used linux-2.6.20 on an x86_64 with 16 seagate 7200.10 320GB
drives.
960 MB/Sec combined write rate with XFS. Not bad.
The driver even installed without complaints. Something must be
wrong!
Sure frees up the slots and rocks the I/O space.
Just something to add to the SATA list.
berkley
--
E. F. Berkley Shands, MSc
Date: Fri, 09 Mar 2007 16:53:50 -0800
From: Tony Alfrey <tonyalfrey@earthlink.net>
To: bofh@linuxmafia.com
Subject: SATA Card
I have spent two days struggling with getting SATA running on a couple
year-old Gigabyte/Athlon MB with PCI slots and have just succeeded with
a SYBA SD-SATA150R. www.syba.com
These people have lots of SATA and RAID card and they support linux.
They use the Silicon Image chipsets and the controller comes with linux
drivers, yet the sata-sil driver that came with SuSE worked. I am not a
real linux geek so you might want to really check these guys out to get
all of the straight information but the card works and the card was sold
at Fry's in the SF Bay area for 30 bucks for the 2-port card.
Tony Alfrey
Date: Sat, 10 Mar 2007 12:49:14 -0800
From: Tony Alfrey <tonyalfrey@earthlink.net>
To: bofh@linuxmafia.com
Subject: more SATA things
The other day I sent you a note about the successful installation of a
SATA card from SYBA. That card works fine, but I see it is only a SATA
card, not a SATAII card so it is limited to 1.5GBits/sec. They make no
PCI cards for SATAII, only PCI-Express cards.
Be that as it may, the card is auto-recognized by a Fedora Core 6
install disk which installs the sata-sil driver automatically.
Tony Alfrey
Date: Sat, 30 Dec 2006 14:26:38 -0800
To: Rick Moen <rick@linuxmafia.com>
From: Tim Utschig <tim@tetro.net>
Subject: Info for SATA page -- Addonics Cardbus
Hi Rick,
If you can use it, I have some info you can add to your excellent SATA
chipsets information page:
http://linuxmafia.com/faq/Hardware/sata.html#addonics-cardbus
I bought an Addonics ADCBSAR5-2E "CardBus eSATA RAID5/JBOD Adapter":
http://addonics.com/products/host_controller/adcbsar5-2e.asp
They even mention that the chipset is "SIL3124" which I saw listed under
"Completely open chipsets" at linux-ata.org:
http://linux-ata.org/driver-status.html#open_chipsets
Here's the lspci -vv output:
03:00.0 RAID bus controller: Silicon Image, Inc. SiI 3124 PCI-X Serial
ATA Controller (rev 01)
Subsystem: Silicon Image, Inc. Unknown device 7124
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping+ SERR- FastB2B-
Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort-<TAbort- <MAbort- >SERR- <PERR-
Latency: 64
Interrupt: pin A routed to IRQ 5
3 Region 0: Memory at 34008000 (64-bit, non-prefetchable) [size=128]
3 Region 2: Memory at 34000000 (64-bit, non-prefetchable) [size=32K]
3 Region 4: I/O ports at 2800 [size=16]
3 [virtual] Expansion ROM at 30000000 [disabled] [size=512K]
Capabilities: [64] Power Management version 2
Flags: PMEClk- DSI+ D1+ D2+ AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
Status: D0 PME-Enable- DSel=0 DScale=1 PME-
Capabilities: [40] PCI-X non-bridge device
Command: DPERE- ERO+ RBC=512 OST=12
Status: Dev=ff:1f.0 64bit+ 133MHz+ SCD- USC- DC=simple DMMRBC=2048 DMOST=12 DMCRS=128 RSCEM- 266MHz- 533MHz-
Capabilities: [54] Message Signalled Interrupts: 64bit+ Queue=0/0 Enable-
Address: 0000000000000000 Data: 0000
I'm running 2.6.19.1 and haven't had any problems with the card yet.
Though I've only been using it for two days. Hotplugging works great
every which-way I could think of testing it.
Let me know if there's any other information about it you could use.
Have a happy new year!
--
- Tim Utschig <tim@tetro.net>
Date: Fri, 20 Oct 2006 00:07:07 -0700
From: Adam Glass <adam@clarity.net>
To: rick@linuxmafia.com
Subject: SiI3112A update
Rick,
I'm writing re: http://linuxmafia.com/faq/Hardware/sata.html.
Sadly, the very popular Silicon Image 3112 chipset does not seem to support
my new Seagate 750GB "7200.10" drive. This is apparently a known issue.
I'm still awaiting a response from Silicon Image regarding what options
exist.
Adam
Date: Tue, 11 Jul 2006 11:56:49 -0400
From: "Oliver Schulze L." <oliver@samera.com.py>
To: rick@linuxmafia.com
Subject: New Sis 965L driver
Hi,
I'm writing about:
http://linuxmafia.com/faq/Hardware/sata.html#sis-965l
There is a new driver for Sis 965L in:
http://www.sis.com/download/agreement.php?id=155927
http://www.sis.com/download/
HTH
Oliver
--
Oliver Schulze L.
<oliver@samera.com.py>
From: Morten Slott Hansen <msh@onliners.dk>
To: rick@linuxmafia.com
Date: Sun, 2 Jul 2006 11:06:11 +0200
Subject: Promise SuperTrak EX8300 8-port SATA-II PCI-X card and
SuperTrak EX8350
+8-port SATA-II PCI Express card
Hi Rick,
I have successfully installed the SuperTrak EX8350 controller on my
gentoo box and as far as I can tell it is real hardware.
System info:
05:0e.0 RAID bus controller: Promise Technology, Inc. Unknown device 8350
Subsystem: Promise Technology, Inc. Unknown device 0374
Flags: bus master, stepping, 66MHz, medium devsel, latency 96, IRQ 17
Memory at d4000000 (32-bit, non-prefetchable) [size=4K]
Expansion ROM at d3000000 [disabled] [size=64K]
Capabilities: [c0] Power Management version 2
Capabilities: [d0] Message Signalled Interrupts: 64bit+ Queue=0/1 Enable-
Capabilities: [e0] PCI-X non-bridge device
Mainbord is: nVidia Corporation CK804
Patches that I used:
http://forums.storagereview.net/index.php?...99entry228899
http://kernel.org/pub/linux/kernel/people/...supertrak.patch
http://kernel.org/pub/linux/kernel/people/...5-support.patch
Date: Mon, 5 Jun 2006 21:03:24 +0100
From: Ronan <mr_platelet@fastmail.fm>
To: rick@linuxmafia.com
Subject: SATA on Linux: info on VIA VT8251 southbridge chipset
Hi.
As the unfortunate owner of a motherboard containing a VT8251
southbridge, I thought I might send you a note about this chipset,
which you may find useful for your page on SATA support in Linux
(http://linuxmafia.com/faq/Hardware/sata.html#via8251).
It seems that this chipset is not supported in standard Linux. VIA have
created a patch which enables Linux to work with it, but Jeff Garzik has
chosen not to use the patch in its current form. The thread beginning
here has more:
http://marc.theaimsgroup.com/?l=linux-ide&m=113436067230517&w=2
I've checked the current release of the kernel (2.6.16.19), and it still
doesn't seem to have any code for the chipset.
Regards,
Ronan.
To: Rick Moen <rick@linuxmafia.com>
Cc: faj@bzz.no
From: Petter Reinholdtsen <pere@hungry.com>
Date: Mon, 05 Jun 2006 11:35:44 +0200
Subject: Re: Info. relevant to
http://developer.skolelinux.no/info/prosjektet/delprosjekt/hw-raid-info.html
[Rick Moen]
> Greetings, O faraway Norwegian cousins!
Greetings, and sorry for the late reply. I've since concluded that I
lack the time to maintain the page, and put it on a wiki for community
maintainence instead.
> As the maintainer of a Web page tracking the Linux support status of
> all known Serial ATA (SATA) chipsets, I've long found your page on
> Linux and HW RAID,
> http://developer.skolelinux.no/info/prosjektet/delprosjekt/hw-raid-info.html
> , useful and informative.
Thank you. I hope it will continue to do so in the future with its
new home <URL:http://wiki.debian.org/LinuxRaidForAdmins>
> It occurs to me that you might not have noticed my page, and that, to
> the extent its information isn't erroneous (which I cannot guarantee),
> you might find it valuable. It is at:
>
> http://linuxmafia.com/faq/Hardware/sata.html
I had not noticed, but now I have added a block at the end of the
document with related pages and added a link to it.
Thank you for the heads up. Please have a look at the wiki page and
insert more relevant information if you got any. :)
Friendly,
--
Petter Reinholdtsen
Date: Wed, 01 Feb 2006 01:51:45 +1300
From: Sigurd Magnusson <sigurd@silverstripe.com>
To: rick@linuxmafia.com
X-Mailer: Apple Mail (2.746.2)
Subject: fakeraid vs true vs software
I recently bought a four disk server with the intention to make use
of RAID1 on two of the drives (containing the OS and important files)
and to take advantage of a hotswap backplane. After reading your page
at http://linuxmafia.com/faq/Hardware/sata.html it seems I have a
dilemma on my hands that you may be able to help solve.
The server will mainly be occupied as a smtp, pop3 and imap server
running serverside spam protection for 20-50 accounts, along with a
less cpu-intense job of samba fileserving to a LAN. Once a day around
4am it would spend around an hour rsycing a few gigs of files from
other servers, to collate a backup onto DVD (this activity is done at
a time when performance to other tasks, like mail, can be sluggish).
Its a SuperMicro 1U Pentium 4 Dualcore 3ghz machine with four 300gb
SATA150 disks; http://supermicro.com/products/system/1U/5015/
SYS-5015P-TR.cfm )
In installing Debian, I struck a few issues so in browsing the
internet I learned (or so I think);
a) The raid is provided by the Intel motherboard- ICH7R - and
although modern, it is "fakeraid".
b) SATA Hotswappability appears in documentation to not be written
yet (in libata?), therefore it seems I need to power off the machine
to juggle disks? (or is it a case of having to manually umount drives
before removal?). I had believed I could replace a failing raided
drive in the blink of the eye, and also would mean that I could
upgrade the size of the two other (non raid) drives without turning
the power off (these drives contain useful files that can be re-
retrieved elsewhere; raid is not needed).
I am trying to piece through my options;
a) Understand the pros and cons of the ICH7R bios/cpu raid vs the
software raid I can set up in the debian install (the latter seeming
to be easier to get going?). Another cross against the ICH raid is
that documentation implies the ICH7R would be slower; are their any
benefits of using ICH7R?
b) Understand the benefits of buying a true hardware RAID controller,
given the performance alone may or may not improve noticeably?
(Improvements to reliability, being able to hotswap, making
installation/maintainence easier are my bigger drawcards). I've
already spent about (NZ dollar) $3500 on the server so to spend
another $500ish on something like a 3Ware 9500S4 (http://www.
3ware.com/products/serial_ata9000.asp) makes me sigh!
Any advice on what RAID option is suggested would be most welcome!
Date: Tue, 22 Nov 2005 14:50:52 +0200 (EET)
From: Jyrki Malinen <jjmaline@cc.hut.fi>
Reply-To: Jyrki.Malinen@hut.fi
To: rick@linuxmafia.com
Subject: About SATA-webpage.
Hello..
Some updates to Promise SATA-II-150 and also new SATA300 information.
http://www.promise.com/support/download/download2_eng.asp?productId=139&category=driver&os=3&go=GO#
Hopefully that link works. (Support -> SATA300 -> Driver -> Red Hat)
2.6 Kernel sources and binaries - I haven't tested yet.
With path Support -> SATA300 -> Driver -> Other; can be found 2.4 kernel
source codes and binaries.
Filenames have both SATA300 and SATA-II-150 mentioned, but I don't know
the truth.
Disclamer: I'm private person and just looking a sata card for my private
home server.
--
Jyrki Malinen / oh2hyt
Date: Thu, 1 Sep 2005 11:30:36 -0400 (EDT)
From: Albert Lash <alby@thirteen.net>
To: rick@linuxmafia.com
Subject: Adaptec 2020SA and AACRAID
Hi Rick,
Thanks for your webpage. I just got a SuperMicro server with a specially
designed card - the Adaptec 2020SA - specifically for the SuperMicro
case.
I am running Gentoo 2.6.12 with the AACRAID driver no problem. You may
want to add this to your page to aid Googlers.
Date: Tue, 9 Aug 2005 15:06:36 -0700
From: Tim Lynch <Tim.Lynch@Reardencommerce.com>
To: rick@linuxmafia.com
Subject: Serial ATA (SATA) chipsets - Linux support status
awesome resource, thanks so much for making this available.
i have an ibm x306 running 2.4 kernel, RHEL3. it has some wacky
frankenstein adaptec ich5 chipset, a SATA RAID. i'm not sure what the
chip is exactly but it uses a binary driver ``aarich'' supplied by
ibm/adaptec:
Adaptec aarich HostRAID driver v4.00.013
Vendor: ADAPTEC Model: AAR-ICHx Version: 4.00.013
Major: 254
Build Date: Feb 19 2004
but looking at /proc/pci, it looks like some onboard intel chip:
Bus 0, device 31, function 0:
ISA bridge: PCI device 8086:25a1 (Intel Corp.) (rev 2).
Bus 0, device 31, function 1:
IDE interface: PCI device 8086:25a2 (Intel Corp.) (rev 2).
IRQ 18.
I/O at 0x1460 [0x146f].
Non-prefetchable 32 bit memory at 0x40000000 [0x400003ff].
Bus 0, device 31, function 2:
RAID bus controller: PCI device 8086:25b0 (Intel Corp.) (rev 2).
IRQ 18.
I/O at 0x1490 [0x1497].
I/O at 0x1484 [0x1487].
I/O at 0x1488 [0x148f].
I/O at 0x1480 [0x1483].
I/O at 0x1470 [0x147f].
Bus 0, device 31, function 3:
SMBus: PCI device 8086:25a4 (Intel Corp.) (rev 2).
IRQ 17.
I/O at 0x1440 [0x145f].
anyways, i think it deserves mention on your page because IBM calls it a
ServeRAID-7e but it's quite different from their others, eg the
ServeRAID-7t.
http://linuxmafia.com/faq/Hardware/sata.html#ibm
Date: Thu, 04 Aug 2005 13:49:31 -0400
From: Rob Hinst <rob@exobitnetworks.com>
To: rick@linuxmafia.com
Organization: Exobit Networks, Inc.
X-Mailer: Evolution 2.2.1.1
Subject: RocketRAID 1820A
Hi,
I noticed you don't have much information about the RocketRAID 1820A
controller card on your listing of SATA chipsets. I just finished
writing an in-depth tutorial on setting up this card on a Debian system.
If you'd like to link to it, you can find it at
http://hinst.net/modules.php?name=News&file=article&sid=1
Thanks.
Date: Mon, 30 May 2005 18:21:16 +0200
From: Sjors Hertstein <sjors@abfresearch.nl>
To: rick@linuxmafia.com
Subject: ICH6R
Dear Rick,
After an intensive search across the net for the LSI Host RAID drivers I
found the following source with a lot of drivers:
ftp://ftp.supermicro.com/driver/Diskimag/LSI/ICH6R/
I got it to work under SuSE 9.2. Probably it will fail to work after the
kernel update with YOU (Yast Online Update). I will keep you up to date
about it. (And how I fixed it).
I also got the source-code from LSI, but could not get it compiled for
the SuSE 9.1 and 9.3 with 2.6 kernel. At least not with the included
instructions. Maybe somebody else can.
Date: Tue, 31 May 2005 18:10:03 +0200
From: Sjors Hertstein <sjors@abfresearch.nl>
To: Rick Moen <rick@linuxmafia.com>
Subject: RE: ICH6R
Quoting Rick Moen (rick@linuxmafia.com):
> I suspect we'll find that there is indeed (as you say) source code,
> but also a core libarary available only in binary form, where
> practically all the functionality is performed
You are right, it's with a core library. It's not the proper way to do
things in Linux, but I like to be on the practical side of things. If it
works, then its fine by me. I think LSI would profit more by making it
open source though.
Quoting Sjors Hertstein (sjors@abfresearch.nl):
> I got it to work under SuSE 9.2. Probably it will fail to work after
> the kernel update with YOU (Yast Online Update). I will keep you up to
> date about it. (And how I fixed it).
Here the steps I took after updating the kernel from 2.6.8-24-smp to
2.6.8-24.14-smp via YOU (do not reboot before running this):
mv /lib/modules/2.6.8-24.14-smp/extra/megaide.ko /lib/modules/2.6.8-24.14-smp/extra/megaide.ko.org
cp /lib/modules/2.6.8-24-smp/kernel/drivers/scsi/megaide.ko /lib/modules/2.6.8-24.14-smp/kernel/drivers/scsi/
depmod -v 2.6.8-24.14-smp
mk_initrd
As you see I re-use the megaide driver of kernel 2.6.8-24-smp which is
not the proper way to do it, but hey it works ! Probably because the
kernel update is only a minor security fix. The available driver in
/lib/modules/2.6.8-24.14-smp/extra/megaide.ko doesn't work for my
LSI/ICH6R chipset (Intel Entry Server Board SE7221BK1-E). So that's why
I replaced it.
Date: Sat, 28 May 2005 06:22:24 +0000
From: Niall Walsh <linux@esatclear.ie>
To: rick@linuxmafia.com
Subject: http://linuxmafia.com/faq/Hardware/sata.html#acera
Hi Rick,
Allegedly the drivers for these cards is held at:
ftp://60.248.88.208/RaidCards/AP_Drivers/Linux/DRIVER/SourceCode/
I understand that it appears to require a kernel rebuild for these
drivers, and the latest drivers are missing the installation
instructions.
It appears to be supported though:
"This is an ARECA maintained driver by Erich Chen. <If you have any
problems, please mail to: erich@areca.com.tw <mailto:erich@areca.com.tw>."
Make of it what you will! One customer at least, not me, will be
emailing them to ask about building the module outside of the kernel.
Niall Walsh
To investigate, when possible: Marvell 88SE61XX
Alleged to be included in these poxy third-party drivers:
http://www.driverstock.com/ASUS-P5WDG2-WS-Professional-driver-download/6-9-10190-81256/index.html
http://dlsvr01.asus.com/pub/ASUS/misc/sata/Marvell6121_V1009_linux.rar
ftp://dlsvr03.asus.com/pub/ASUS/misc/ide/marvell6141/61xx_SATA_linux.zip
http://dlsvr01.asus.com/pub/ASUS/misc/ide/marvell6141/61xx_SATA_linux.zip
ftp://dlsvr02.asus.com/pub/ASUS/misc/ide/marvell6141/61xx_SATA_linux.zip
http://dlsvr04.asus.com/pub/ASUS
http://files.driverstock.com/stor/files/6/pub/ASUS/misc/ide/marvell6141/61xx_SATA_linux.zip
Appears to be an entire Linux 2.6.12.3 kernel source + binary tree,
presumably with a patched version of the "mvsata" driver, though that is
uncertain given the utter absence of documentation.
AHCI-driver support:
http://lists.laptop.org/pipermail/server-devel/2007-July/000070.html
Date: Thu, 7 Feb 2008 15:28:27 +0900
To: bofh@linuxmafia.com
From: Craig Ringer <craig@postnewspapers.com.au>
Subject: FAQ suggestion - generic.all_generic_ide=1
Hi
I noticed in your FAQ that you don't mention one possible SATA
workaround for some chipsets.
Especially for CD installers, it can be necessary to add to the kernel
command line the argument:
generic.all_generic_ide=1
In particular this can be necessary on newer Intel ICH chipsets.
--
Craig Ringer
Reply-To: amah@highpoint-tech.com
From: Allen Mah <amah@highpoint-tech.com>
To: rick@linuxmafia.com
Date: Mon, 25 Feb 2008 16:31:16 -0800
Organization: HighPoint Technologies, Inc.
X-Mailer: Microsoft Office Outlook, Build 11.0.5510
Subject: About HighPoint
Dear Rick, I would like to get you updated on HighPoint's new product
line of RocketRAID 3000 Hardware RAID controllers. The new RocketRAID
3000 Series includes the following products RocketRAID 3522, 3520, 3510,
3320, 3220. The RR3522, 3520 and 3510 have Intel IOP 341 (800MHz)
processor while the RR3320 and RR3220 have Intel IOP 331 (500MHz); every
hardware RAID controller also comes with at least 256MB of DDR-II
memory. We also have the RocketRAID 3120 that uses the Marvell 5182 IOP
All of our RocketRAID 3000 series of controllers are embedded into the
main Linux kernel 2.6.24 and and have been back ported into Red Hat
Enterprise Linux and CentOS.
Here are some links that show our RocketRAID 3000 series included into
the main Linux kernel
http://www.mjmwired.net/kernel/Documentation/scsi/hptiop.txt.
You can find latest in the git.kernel.org.
http://git.kernel.org/?p=linux/kernel/git/jejb/scsi-misc-2.6.git;a=tree;f=dr
ivers/scsi;h=03a18a7adc45ab39658ac83cf90e56e43598ee3e;hb=HEAD
Please update your site to let users know that HighPoint Technologies no
longer has Fake RAID, we truly have a hardware RAID controller.
Thank You
Allen Mah
HighPoint Technologies
www.highpoint-tech.com
408-942-5800 x116