[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: WAPBL/cache flush and mfi(4)
On Sat, 25 Aug 2012, Michael van Elst wrote:
bouyer%antioche.eu.org@localhost (Manuel Bouyer) writes:
Sure, but with some PERC controllers it seems there's no option
(at last ine in the firmware's UI) to disable disk caches with SATA drives.
It may be different with SAS drives attached.
I had a case where neither the BIOS nor the Linux command line tool
for the LSI controllers (not a PERC) were able to influence the disk
caches (i.e. they were always turned off). But a "disk managemenent suite"
(delivered as a bootable ISO) was able to change the setting.
And yes, these were SATA drives.
We hae an HP server with an LSI controller and SATA drives. It was
originally running a Linux XEN kernel, and the DOMU was doing lots of disk
writes (updating several thousand RRD files). Performance was quite bad
when the disk writes started. I determined that the RAID1 volume did not
have write cache enabled, and was able to use the Linux lsiutil program to
change that, but still had bad write performance. I determined that the
STAT drives did not have the write cache enabled. I was able to modify
the lsiutil source to add an option to enable the SATA drive cache and
finally got much better write performance (although the DOMU system had
long been moved to a VM and the system was no longer running XEN.).
And then there's the HP cac(4) that had no write caching, and the
ciss(4) that lots of people neglected to add BBU cache to. The older
SmartStart CD for the older servers did have an option to enable the disk
write caching (but not the controller RAID caching), with a warning about
it being dangerous. The raid contollers did seem to perform fairly well
when you could send multiple writes to the controller - which the ciss(4)
driver did not do for some time.
Michael L. Hitch mhitch%montana.edu@localhost
Information Technology Center
Montana State University Bozeman, MT USA
Main Index |
Thread Index |