|
|
Back to UF - User Diary Index
| RAID Controllers |
by JohnClarke |
2003-11-08 08:09:24 |
Continuation of this.
The SX6000 wasn't obsoleted 3 months ago, it was pretty much obsolete the day it shipped more than 2 years ago--the only reason it had a chance to begin with was that 3ware had announced plans to drop their ATA products. It was a poor performer out the door and as I said before Promise admitted as much.
If the processor on an IDE drive means that a machine has "hardware RAID" then a machine running the RAID that is included in the box with Windows server has "hardware RAID". In that case how do you make a distinction between hardware RAID and software RAID?
The distinction is usually that "hardware RAID" refers to a RAID is managed by an auxiliary processor independent of the main CPU simply feeding it read and write requests, while "software RAID" refers to a RAID that is managed by the main CPU, which does all calculaions necessary to support striping and failover. In that sense the SX150 is "software RAID" in that the RAID management is performed by the main CPU, not a CPU on the board. The "XOR engine" is not a processor anymore than an adder is a processor--it is a processor element but it is not in its own right a Turing machine and thus cannot by the definitions in common use be considered a "processor".
One could argue that it's neither fish nor fowl, with some of the characteristics of hardware RAID and some of the characteristics of software RAID.
I don't know about the S150SX but earlier Promise devices have not always provided "toaster reliability"--they sometimes fail to boot with a dead drive even though the other is fine, and they don't always provide complete isolation between channels, meaning that a drive failure on one channel can corrupt data on the spare on the other channel--I've seen that happen. I contrast that with the _true_ "toaster reliability" of the AMI and Mylex RAID controllers that I've used that can be counted on to get the machine up as long as there is enough of the array left intact for them to be able to come up with a full data set.
You're right that the boot ROM is a convenience but that doesn't make "hardware RAID". There's no reason the ability to configure an array running on the standard IDE controllers built into the motherboard chipsets couldn't be programmed into the boot ROM on the motherboard--I suspect that the only reason that that hasn't been done is that there is no standard for such a RAID.
As for recommending "the best probability in their area" I assume that anybody I'm dealing with online has a credit card and an Internet connection--with those just about anything is available, in Europe, North America, and most of the Pacific Rim anyway. |
|
[ Reply ] |
|
"you do the math" | by polarKitten | 2003-11-10 08:02:00 |
|
Last point first. | by Slamlander | 2004-01-16 06:12:37 |
|
|
[Todays Cartoon Discussion]
[News Index]
|
|
