[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: 3ware escalade 8006 SATA RAID recommendations; dual boot windows



I merely add more questions and an observation,
and embelish on these further below.
a. Doesn't command queing (NCQ) bring SATA much closer to SCSI performance?
b. Aren't 3Ware adapter cards and most SATA disk drives
   Frankenstein drives with bolted-on translater cards
   that necessarily disallow any command queing?
c. Does moving physical disk drives from one SATA RAID controller
   to another manufacturer's SATA RAID controller fail? 
d. Does Linux software RAID work almost as well as hardware RAID;
   perhaps with fewer difficulties, as when moving a set of drives 
   to a different manufacturer's RAID adapter card?


The article
http://www.tomshardware.com/storage/20040123/index.html
says
  "An NCQ-compatible device, i.e. one that conforms to serial ATA II specifications,
  can accept up to 32 commands and process them in an optimized sequence.
  The forerunner in this respect is Seagate and its Barracuda 7200.7, which already supports NCQ.

Both the following SATA drives handle command queing,
Western Digital Raptor 74GB (not the 36GB version),
Seagate Barracuda 7200.7, 120GB
but none of the large inexpensive drives from Western Digital or Maxtor 
handles NCQ.
Indeed, they are FRANKENSTEIN drives -- ATA drives with 
a board bolted on to convert SATA commands to the ATA disk drive's commands,
disallowing non-ATA commands like NCQ.

I believe 3Ware also uses a FRANKENSTEIN approach for its adapter cards,
so they can't use NCQ command queing.
So usually, both the disk drive and the adapter card are Frankenstein hardware.

I gather the Sil3114 chips and the newer Sil3124 handle NCQ command queing.
You can see some more information at
http://www.tomshardware.com/storage/20040426/backplane_storage-08.html 
http://www.tomshardware.com/business/20040320/cebit2004_1-10.html
where the Promise SATAII 150 SX8 is an 8-channel PCI X slot with tagged and NCQ.

I suspect that if you change SATA control card manufacturers;
eg, from 3Ware to Asus, then your RAID drives cannot immediately
be moved to the new SATA card. 
This is really a question.
When I switched from a BusLogic to an Adaptec SCSI adapter card,
I could no longer use a disk drive formerly attached to my BusLogic card
(although I could have missed a trick).

So, I have wondered about using Linux's software RAID rather than hardware RAID.
That is, use a SATA controller card that treats drives as JBOD (just a bunch of drives)
without using hardware RAID.
I have seen disparate performance reviews,
but some of those reviews showed software RAID performing as good as hardware RAID,
although the CPU must get involved a little more.
This is really another question.



On Thu, Jun 10, 2004 at 12:51:12AM -0500, Nick Lidakis wrote:
> I have been meaning to upgrade my hard disk setup to a RAID array 0 to 
> help increase system responsiveness and reduce system applications start 
> up times. After some research using google, it seems that the 3ware 
> escalade line of RAID SATA cards, specifically the 8006-2LP 2 port RAID 
> card, fit the bill nicely. The cards are supported with an open source 
> Linux driver, is the only true hardware RAID card supported in LINUX (as 
> per the kernel documentation), and is reasonably inexpensive ($139 @ 
> newegg).
> 
> I have a few questions that I coudn't find an answer to using either 
> google or searching through the Debian archives:
> 
> Is anybody using this card on a desktop system, and are they happy with 
> the performance in a RAID 0 array?
> 
> Would I still be able to use LILO or GRUB to dual boot windows off of a 
> IDE drive connected to the motherbaords IDE interface?
> 
> Is there a big performance increase using a SCSI RAID 0 array as opposed 
> to SATA RAID? If yes, which would be a recommended (hardware RAID) SCSI 
> card to use with Debian?

-- 
Jameson C. Burt, NJ9L   Fairfax, Virginia, USA
jameson@coost.com       http://www.coost.com
(202) 690-0380 (work)

Attachment: signature.asc
Description: Digital signature


Reply to: