raid SW vs HW , was Okay, this is so not cool..

Matt Clark matt at ymogen.net
Wed Oct 22 05:25:01 CDT 2003


> > Except _all_ the ones with battery backed cache.
> > For read performance it's all much of a muchness probably.
>
> Maybe someone wants to take a look to *best* comparative out there,
> a bit old:
>
> <Symbios_53c876 and Adaptec_7896 SCSI + Linux RAID against ten HW RAID boards>
> http://www.linux-magazin.de/Artikel/ausgabe/2000/08/RAIDController/RAID-Controller.html

I don't speak German, but the MegaRaid BIOS screenshot shows write-through caching enabled, not write-back.  Anyway, it's not worth
arguing about, write-back cache is simply orders of magnitude faster than any other available option (unless of course you're just
doing a single huge sequential write).  You can't do 50MB of random writes in a second any other way (well, OK, a 25 disk RAID0 with
a 4ms seek time might be able to do it!).

> > Not always a toy... http://www.apple.com/xserve/raid/.  Stunningly better
> > value than almost any other external RAID solution on the market.
>
> IDE is cheaper but isn't a _real_ solution, it's only a patch :-)

Huh?  Who cares how the disks are attached, so long as the array speed is sufficient?  I believe the XServe Raid can do 20,000 IOPs.

M






More information about the Linux-PowerEdge mailing list