SATA link up 1.5 Gbps

~
I am eyeing this WD SATA harddrive with (maximal) trnasfer rates of 3 Gb/s:
~
Western Digital WD6400AAKS WD Caviar SE16 SATA Internal Hard Drive, 640 GB,
3 Gb/s, 16 MB Cache, 7200 RPM
~
http://www.wdc.com/en/products/products.asp?driveid=394
~
but I don't know if the combination of (processor + IO subsystem + BIOS) I
have (running on 1 GB of RAM)
~
sh-3.1# dmesg | grep Athlon
CPU0: AMD Athlon(tm) 64 Processor 3500+ stepping 02
powernow-k8: Found 1 AMD Athlon(tm) 64 Processor 3500+ processors (version
2.00.00)
~
supports 3 Gbps data transfers
~
Does it?
~
// __ Here more about my underlying hardware
~
sh-3.1# dmesg | grep ata
Loading iSCSI transport class v2.0-724.7sata_sil 0000:00:12.0: version 2.0
ata1: SATA max UDMA/100 cmd 0xF8804080 ctl 0xF880408A bmdma 0xF8804000 irq
17
ata2: SATA max UDMA/100 cmd 0xF88040C0 ctl 0xF88040CA bmdma 0xF8804008 irq
17
scsi0 : sata_sil
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata1.00: ATA-7, max UDMA/133, 156301488 sectors: LBA48 NCQ (depth 0/32)
ata1.00: ata1: dev 0 multi count 16
ata1.00: configured for UDMA/100
scsi1 : sata_sil
ata2: SATA link down (SStatus 0 SControl 300)
~
sh-3.1# dmesg | grep ATA
hda: HL-DT-ST RW/DVD GCC-4482B, ATAPI CD/DVD-ROM drive
hda: ATAPI 48X DVD-ROM CD-R/RW drive, 1536kB Cache
ata1: SATA max UDMA/100 cmd 0xF8804080 ctl 0xF880408A bmdma 0xF8804000 irq
17
ata2: SATA max UDMA/100 cmd 0xF88040C0 ctl 0xF88040CA bmdma 0xF8804008 irq
17
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata1.00: ATA-7, max UDMA/133, 156301488 sectors: LBA48 NCQ (depth 0/32)
ata2: SATA link down (SStatus 0 SControl 300)
scsi 0:0:0:0: Direct-Access ATA ST3808110AS 3.AA PQ: 0 ANSI: 5
~
Thanks
lbrtchx

Previously Albretch Mueller wrote:
~
I am eyeing this WD SATA harddrive with (maximal) trnasfer rates of 3 Gb/s:
~
Western Digital WD6400AAKS WD Caviar SE16 SATA Internal Hard Drive, 640 GB,
3 Gb/s, 16 MB Cache, 7200 RPM
~
http://www.wdc.com/en/products/products.asp?driveid=394
~
but I don't know if the combination of (processor + IO subsystem + BIOS) I
have (running on 1 GB of RAM)
~
sh-3.1# dmesg | grep Athlon
CPU0: AMD Athlon(tm) 64 Processor 3500+ stepping 02
powernow-k8: Found 1 AMD Athlon(tm) 64 Processor 3500+ processors (version
2.00.00)
~
supports 3 Gbps data transfers
~
Does it?


I think it does, but it matters very little. You will not get any
perceptable speed-up over 1.5GB. The one advantage at this time
is better compatibility, and with WDs shoddy SATA protocol
implementation, this may be an issue.

Arno


~
// __ Here more about my underlying hardware
~
sh-3.1# dmesg | grep ata
Loading iSCSI transport class v2.0-724.7sata_sil 0000:00:12.0: version 2.0
ata1: SATA max UDMA/100 cmd 0xF8804080 ctl 0xF880408A bmdma 0xF8804000 irq
17
ata2: SATA max UDMA/100 cmd 0xF88040C0 ctl 0xF88040CA bmdma 0xF8804008 irq
17
scsi0 : sata_sil
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata1.00: ATA-7, max UDMA/133, 156301488 sectors: LBA48 NCQ (depth 0/32)
ata1.00: ata1: dev 0 multi count 16
ata1.00: configured for UDMA/100
scsi1 : sata_sil
ata2: SATA link down (SStatus 0 SControl 300)
~
sh-3.1# dmesg | grep ATA
hda: HL-DT-ST RW/DVD GCC-4482B, ATAPI CD/DVD-ROM drive
hda: ATAPI 48X DVD-ROM CD-R/RW drive, 1536kB Cache
ata1: SATA max UDMA/100 cmd 0xF8804080 ctl 0xF880408A bmdma 0xF8804000 irq
17
ata2: SATA max UDMA/100 cmd 0xF88040C0 ctl 0xF88040CA bmdma 0xF8804008 irq
17
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
ata1.00: ATA-7, max UDMA/133, 156301488 sectors: LBA48 NCQ (depth 0/32)
ata2: SATA link down (SStatus 0 SControl 300)
scsi 0:0:0:0: Direct-Access ATA ST3808110AS 3.AA PQ: 0 ANSI: 5
~
Thanks
lbrtchx

Arno Wagner wrote:
I think it does, but it matters very little. You will not get any
perceptable speed-up over 1.5GB. The one advantage at this time
is better compatibility, and with WDs shoddy SATA protocol
implementation, this may be an issue.

Arno
~
Hmm! Arno, you say "it matters very little. You will not get any
perceptable speed-up over 1.5GB" without saying why ;-)
~
Which combination of (Motherboard + I/O subsystem + BIOS + (?)) will let
you have your cake and eat it too?
~
Also, which manufacturers have more useful SATA, S.M.A.R.T, TLRE, . . .
implementations so that you may better predict when drives are about to
fail and possibly the physical reasons why
~
I am thinking of implementing RAID 5 using a Linux (or BSD) software-based
RAID any good best practices out there?
~
thanx
lbrtchx

Previously Albretch Mueller wrote:
Arno Wagner wrote:
I think it does, but it matters very little. You will not get any
perceptable speed-up over 1.5GB. The one advantage at this time
is better compatibility, and with WDs shoddy SATA protocol
implementation, this may be an issue.

Arno
~
Hmm! Arno, you say "it matters very little. You will not get any
perceptable speed-up over 1.5GB" without saying why ;-)

Why would you get a speed-up? This is the interface speed, not the
drive speed. Drive speed is much lower.

~
Which combination of (Motherboard + I/O subsystem + BIOS + (?)) will let
you have your cake and eat it too?

None.

~
Also, which manufacturers have more useful SATA, S.M.A.R.T, TLRE, . . .
implementations so that you may better predict when drives are about to
fail and possibly the physical reasons why

You need to interpret yourself for that. Basically they are all
usable. Seagate makes pretty bad drives at the moment. WD has
compatibility issues in the Interface. Maxtor still sucks.
Get Samsung or Hitachi.

~ I am thinking of implementing RAID 5 using a Linux (or BSD)
software-based RAID any good best practices out there?

Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Arno

Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Arno

I am sorry, but I am not a hw person. Should I use an attached PCI-Express
card to plug the SATA drives to, even if I a using software-based RAID?

From this "infomercial"

http://www.sci-worx.com/docs/SiI-WP-006-A.pdf

In which they seem to be talking about their external storage which is
connected to a notebook through some PCMCIA card, I read that the highest
data transfer (twice as fast?) can be achieve with SATA drives anyway.
Would you trust these benchmarks? (there is a small chart)

An Internal SATA array should work most faster I guess (maybe wrongly ;-))

Also I have another question for you that generally relates to the prev one

How "possible" (or "unnecessary", . . .) is this?

1) go: http://groups.google.com/group/pgsql.general/topics

2) then search for "HA best pratices with postgreSQL"

(links get corrupted sometimes)

thanks again
lbrtchx

Previously Albretch Mueller wrote:
Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Arno

I am sorry, but I am not a hw person. Should I use an attached
PCI-Express card to plug the SATA drives to, even if I a using
software-based RAID?

In Software RAID all data goes over the sytem bus and it
becomes the main bottleneck. Therefore you want the fastest
bus you can reasonably get.

From this "infomercial"

http://www.sci-worx.com/docs/SiI-WP-006-A.pdf

In which they seem to be talking about their external storage which
is connected to a notebook through some PCMCIA card, I read that the
highest data transfer (twice as fast?) can be achieve with SATA
drives anyway. Would you trust these benchmarks? (there is a small
chart)

USB does limit access speeds to about 25MB/s in practive. SATA
gives you native speed, they are correct about that. THis
works only if the adapter card is connectedd to a fast enough
bus. An express card would typically do.

An Internal SATA array should work most faster I guess (maybe
wrongly ;-))

Also I have another question for you that generally relates to the
prev one

How "possible" (or "unnecessary", . . .) is this?

1) go: http://groups.google.com/group/pgsql.general/topics

2) then search for "HA best pratices with postgreSQL"

(links get corrupted sometimes)

I am not a database expert.

Arno

Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Arno

Also, based on

http://en.wikipedia.org/wiki/List_of_device_bandwidths

you could "theoretically" connect some 20 Serial ATA (SATA-300)
transferring data via the computer bus at 375 MB/s to one PCI Express 2.0
(x16 link) card handling data transfers of 8000 MB/s

So, if mmobos come only with 1 PCI Express slot and SATA ports can be
connected to only one SATA drive, why is it you mostly (only?) see PCI
Express to Serial ATA II controllers with 1 or 2 SATA II ports?

I would love to see a PCI Express 2.0 (x16 link) with 6 or 8 SATA II ports!

thanks
lbrtchx

Hi!

Albretch Mueller wrote:
Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Also, based on

http://en.wikipedia.org/wiki/List_of_device_bandwidths

you could "theoretically" connect some 20 Serial ATA (SATA-300)
transferring data via the computer bus at 375 MB/s to one PCI Express 2.0
(x16 link) card handling data transfers of 8000 MB/s

So, if mmobos come only with 1 PCI Express slot and SATA ports can be
connected to only one SATA drive, why is it you mostly (only?) see PCI
Express to Serial ATA II controllers with 1 or 2 SATA II ports?

I would love to see a PCI Express 2.0 (x16 link) with 6 or 8 SATA II ports!

Why? 2.0 16x?

If you need so many disk drives, first of all, you'll never mount them
in only one desktop (or even server) case. On the other hand, so many
disk drives will be in RAID arrays, not standalone. Also, it is almost
impossible to imagine system who needs to write simultaneously on 20
different drives. In those cases SAN, NAS or DAS systems will be much
preffered solution.

If you need SATA controller with more than two SATA connectors, you have
a plenty of them. Adaptec, AOC (3ware), LSI and some others are
producing those controllers (usually SAS/SATA), and they are working
perfectly.

With best regards,

Iggy

Previously Albretch Mueller wrote:
Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Arno

Also, based on

http://en.wikipedia.org/wiki/List_of_device_bandwidths

you could "theoretically" connect some 20 Serial ATA (SATA-300)
transferring data via the computer bus at 375 MB/s to one PCI Express 2.0
(x16 link) card handling data transfers of 8000 MB/s

So, if mmobos come only with 1 PCI Express slot and SATA ports can be
connected to only one SATA drive, why is it you mostly (only?) see PCI
Express to Serial ATA II controllers with 1 or 2 SATA II ports?

One thing is that these card typically use an 1x PCI-E slot.
The x16 slot is used for the graphics card (but can be used for
other cards, the slots are downward compatible to less lanes).
The controller with the most PCI-E lanes I have seen so far was
24 drive hardware RAID conrtroller from Arcea or 2ware (don't
remember) and it used 8 PCI-E lanes.

I would love to see a PCI Express 2.0 (x16 link) with 6 or 8 SATA II
ports!

Complete overkill! A PCI-E 2.0 lane kann transfer 500MB/s.
The fastest SATA II drives get around 100MB/s. So even for
8 drives, a 2x PCI-E slot would be quite enough. (2x slots
do not exist, but there are 4x slots with only 2 lanes
connected. The standard allows this.)

Arno

Previously Igor Batinic wrote:
Hi!

Albretch Mueller wrote:
Get a controller that is PCI-E attacjed, to acvoid the bus
bottleneck. Can be a board/chipset -integrated one. Make sure
to moditor the array (mdadm) and the SMART status of the disks
(smartd, look for pending secors and reallocated sectors in
particulsr) and make sure error notificatin by both tools work.

Also, based on

http://en.wikipedia.org/wiki/List_of_device_bandwidths

you could "theoretically" connect some 20 Serial ATA (SATA-300)
transferring data via the computer bus at 375 MB/s to one PCI Express 2.0
(x16 link) card handling data transfers of 8000 MB/s

So, if mmobos come only with 1 PCI Express slot and SATA ports can be
connected to only one SATA drive, why is it you mostly (only?) see PCI
Express to Serial ATA II controllers with 1 or 2 SATA II ports?

I would love to see a PCI Express 2.0 (x16 link) with 6 or 8 SATA II ports!

Why? 2.0 16x?

If you need so many disk drives, first of all, you'll never mount them
in only one desktop (or even server) case. On the other hand, so many
disk drives will be in RAID arrays, not standalone. Also, it is almost
impossible to imagine system who needs to write simultaneously on 20
different drives. In those cases SAN, NAS or DAS systems will be much
preffered solution.

Sorry, I have had an * disk and a 4 disk RAID array (software RAID)
in one server. Thet is not the reason PCI-E 16x is inappropriate.
The reason is that it has far, far too much bandwidth for 8 disks.

If you need SATA controller with more than two SATA connectors, you have
a plenty of them. Adaptec, AOC (3ware), LSI and some others are
producing those controllers (usually SAS/SATA), and they are working
perfectly.

True. Just look a bit. Make sure your disks do not saturate
the bus. For example, getting a PCI 4 port SATA controller
would be a bad idea for hardware RAID, as 4 modern disks
can saturate even an 270MB/s 66MHz PIC slot.

Arno