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


Hi guys,

This is getting weirder. First, I found that one can disable "fast initialization" in the card's firmware. I did this, let the discs initialize properly (which took about a day), tried again - no improvement.

I replaced the card with my cheap-o 2-port Sil_3112a card, tested writing to the discs independantly, and in kernel based raid, both raid-0 and raid-1. Single disk and raid-0 both gave me between 25mb/s and 30mb/s, raid-1 gave me about 20mb/s.

Then I tested single disc on the LSI card, port by port. Guess what I found? Ports 0, 2 and 4 are dogs, can barely manage 10mb/s. Ports 1, 3, and 5 do around 25mb/s. To confirm this, when I initialized the discs, the ones on ports 0 and 2 (I didn't have any on 4) took about four times as long as the others. Anybody know why this is?

Then I decided, OK, let's make a raid5 (in the card's firmware), with only three discs on ports 1, 3 and 5 (the "fast" ports). Again, I struggle to reach 10mb/s. While this copy is in progress, I see my system load shooting to over 8.0....

So, my theory: This card relies on the megaraid driver to do the heavy lifting. The CPU on the board is the older P4 based Celeron 2ghz (128k cache), and absolute dog in it's own right. Maybe this machine just isn't strong enough for this type of raid?

000:00:07.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID (rev 01) Subsystem: LSI Logic / Symbios Logic MegaRAID SATA 150-6 RAID Controller
     Flags: bus master, 66Mhz, slow devsel, latency 32, IRQ 11
     Memory at ef000000 (32-bit, prefetchable) [size=64K]
     Capabilities: [80] Power Management version 2

Any insights would be appreciated.


Reply to: