Re: Raw Devices: Increased Performance?

In <4tacun$jid_at_news00.btx.dtag.de>, willyk_at_t-online.de (Willy Klotz) writes:
>You talked about the UNIX kernal, which is optimizing I/O. Is there no
>optimization for raw devices, only for filesystems ?
> <snip> ,,,,
 

>>Result? Filesytem wins by more than 100% over raw device.
>
>This is only true watching accesses for one single disk drive. If a
>system constantly has this type of I/O, then one should consider to
>spread I/Os across several disks.

There is a different between a "raw disk partition" and a "striped SVM managed raw volume".
I tried to make this point a while ago but no one took it on board. I'm not as deep a guru as some of the contributors but I follow the arguments about why f/s is better than raw, but no one seems to have brought SVM and raw volume striping into the picture. It is my belief that the SVM layer does optimise raw device I/O, and with true datafile-level striping, performance is certainly improved over the "dumb" raw-volume scenario. Received on Thu Aug 01 1996 - 00:00:00 CEST