Sebastian Smith wrote:
You didn't happen to ask him why he chose a RAID 1 over four disks did you?
I did, actually, but he could not justify the decision at my request.
The only answer I got out of him was 'for redundancy'. My solution was
three-fold: 1) Use RAID 5 with a hot-spare, 2) Get a real disk
controller (UHCI does _not_ count), and 3) RAID is _not_ a disaster
recovery solution, it is for disaster prevention. In short, 'go away'.
> ...but when the new disks arrive (NCQ enabled)
In my experience, they don't come with NCQ enabled by default. There is
a vendor-specific tool (usually) to enable the feature. Often, however,
it seems to _lower_ drive thoroughput - in my experience alone, of
course. ;)
The nForce 4 Pro chipset uses the PCIe bus to provide 4 channels of SATA.
I know... my workstation is on an Asus A8N-SLI Premium. Same chipset. I
loathe the SiI controller, though.
Do you have any write/read benchmarks to show off?
I got the 8 port, ARC-1220. The customer payed $650 for it. Nothing
scientific, but I do have some general numbers to offer. I used my
workstation as a testbed until I got the new server board for my client
- it was the only PCI-E I had on hand... so I wasn't _too_ out of line.
;) On the above mentioned board and an A64 3200+, 1GB PC3200 RAM
(2-2-2-5), and 3 Seagate SATA disks I happened to have lying about I got
between 140-150 MB/s sustained and nearly 650MB/s burst on a RAID5. Note
that I had to enable 'SLI Mode' on the board, which gave the Areca 8
lanes of PCI-E goodness. This was also coupled with an approx. CPU usage
of 1.5% during my little 'test'.
---
Brandon
_______________________________________________
RLUG mailing list
[email protected]
http://lists.rlug.org/mailman/listinfo/rlug
