Accessing SCSI-Devices >2TB

[Glenn Dawson]

At 09:36 PM 6/13/2005, Scott Long wrote:

>>It is a SCSI RAID controller, but the interface to the OS is a block
>>interface, not a SCSI interface (*).  It isn't routed through the SCSI
>>layer, since that wouldn't really help it that much.
>>(*) There is a SCSI passthrough interface for some aac controllers that
>>can be enabled via the aacp driver.
>>Ken
>
>And the aacp device causes many more problems than it solves, so I'll 
>likely be disabling it soon.  It only provides simple userland SCSI access 
>to cdroms and tape drives, anyways.

This is a little OT for this thread, but I'm currently trying to solve a 
problem which might be related to the aac driver, so I'll toss this out 
there and see if anyone has any ideas for me.

It trying to make the transition from 4.x to 5.x I've noticed that 
performance of the md(4) (in 5.x) driver seems to be much worse than the 
vn(4) driver under 4.x.

Here's the scenario:

A pair of mirrored disks on an adaptec 2120S using the aac(4) driver is 
being used to house files that are the backing store for a number of 
virtual disks.  Each file is typically about 4GB.

Under 4.11 using the vn(4) driver I see throughput of about 17MB/sec when 
writing to one of the virtual disks (using dd).

Under 5.4 using the md(4) driver I only see throughput of about 7MB/sec 
when writing to the virtual disk.

Tests were run using the exact same hardware.


>Scott
>_______________________________________________
>freebsd-current at freebsd.org mailing list
>http://lists.freebsd.org/mailman/listinfo/freebsd-current
>To unsubscribe, send any mail to "freebsd-current-unsubscribe at freebsd.org"