Re: raidframe performance question

[Paul Goyette <paul%whooppee.com@localhost>]

Conclusion:  these disk drives suck!  It's not a raidframe problem....

Details:

I gave up on the first attempt to initialize the parity after it had 
been running for > 30 hours.  I tried to reboot, but unfortunately the 
system paniced (something about a flush-cache not finishing).

After rebooting, I started a

        dd if=/dev/zero bs=32k of=/dev/wd4d

and I had a 'systat vmstat' running in another window.  At first, systat 
reported transfer rate at about 10MByte/sec, but after a minute or so it 
started dropping, more or less linearly.  After about 15 minutes, the 
rate was down to about 800KByte/sec, so I'm assuming that the drive's 
write cache got full.  :)  All the while, the systat output shows that 
the xfer-bytes is always 2K times the xfer-ops, so would indicate that 
something is breaking up the writes into 2K chunks.

Anyway, I started a vi in another window, and suddenly the transfer rate 
jumped back up to 6MByte/sec for a bit and then started decaying again, 
and within 2 or 3 minutes it was back down to 800KByte/sec.

Unfortunately, that trick did not work a second time - I started another 
vi and the disk trasnfer rate continued to decay.  After dd had been 
running for 25 minutes, a ^T gives me

        112788+0 records in
        112787+0 records out
        3695804416 bytes transferred in 1517.492 secs (2435468 bytes/sec)

2.4MByte/sec is pretty poor write performance.  And unrelated to raid.



-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------