>From what I know of (checking OpenSolaris source code) the sd_max_xfer_size
>overwrite maxphys value set in /etc/system. 1MB is an "arbitray" value
>selected by ssd plus leadville stack (Solaris fibre Channel Stack code)
>(unfortunate?) I believe most FC HBAs in the market are able to do one DMA
>transfer >>1M, 1M setting in ssd just set a size limit for HBA to do a DMA
>transfer (unfortunate?) That being said, set big value (>1M) for maxphys
>_does_ not help at all. Well it seems reasonable to assume that most
>applications have the IO size less than 1MB, therefore it makes a lot sense to
>set 1M as sd_max_xfer_size. BTW, as a simple test, I can run a raw IO (using
>dd) to tranfer 10MB in one dd, as expected, ssd driver (or Solaris fibre
>Channel stack) truncated one 10MB transfer into multiple 1MB transfers. You
>can see what I mean here. I _donot_ think it is a good design for ssd plus
>leadville stack to set 1MB _and_ overwrite maxphys.
This message posted from opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
