Date: Fri, 13 Apr 2007 00:42:11 +0800 From: "Tzu-Hua Wang" <biogary@gmail.com> To: "Ivan Voras" <ivoras@fer.hr> Cc: freebsd-hardware@freebsd.org Subject: Re: the performance of SAS is not better than SCSI in FreeBSD, WHY? Message-ID: <224a745a0704120942i19c42927o9826b0343938d093@mail.gmail.com> In-Reply-To: <evgqk5$du4$1@sea.gmane.org> References: <224a745a0704092137h67f5f463w6b355535a6b8415f@mail.gmail.com> <evgqk5$du4$1@sea.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi: When using 'dmesg' command, it shows that da0 at ciss0 bus 0 target 0 lun 0 da0: <COMPAQ RAID 0 VOLUME OK> Fixed Direct Access SCSI-0 device da0: 135.168MB/s transfers da0: 419946MB (860051248 512 byte sectors: 255H 32S/T 65535C) Oh, my goodness. My SAS HD is only the SCSI-0 device. I think this is the major reason. Why? Can you give me a hand? Problem of configuration on the RAID Card? I am sorry, I am really new to the RAID system. Thanks 2007/4/11, Ivan Voras <ivoras@fer.hr>: > > Tzu-Hua Wang wrote: > > > Generally, the performace of SAS in FreeBSD is far worse than SCSI in > > FreeBSD. > > Is it normal? How can I do to improve the performance of SAS HD in > FreeBSD? > > There's a huge number of possible reasons - unless you're testing the > same controller (I think there exist SAS+SCSI controllers, I might be > wrong), you can't conclude that the SAS interface is less supported. > > For what it's worth, I've got approximately the same numbers on DL380 > G5, but I've attributed them to 2.5" small form factor drives. > > 2x 2.5" 10k RPM, CISS hw RAID1 2x 3.5", 7.5k RPM, soft RAID1 > Full stroke 3.863 ms 5.942 ms > Half stroke 3.929 ms 6.186 ms > Quarter stroke 3.998 ms 5.700 ms > Short forward 3.889 ms 4.060 ms > Short backward 4.715 ms 4.748 ms > Seqential outer 1.011 ms 0.167 ms > Sequential inner 1.013 ms 0.177 ms > Read transfer, outside 43.6 MB/s 61.2 MB/s > Read transfer, middle 38.8 MB/s 54.8 MB/s > Read transfer, inside 32.8 MB/s 34.9 MB/s > > The 3.5" drives were SATA in a different machine (Dell's). It might be > that the CISS driver is unreasonably slow, which is kind of possible > given the suspicious lower bound of about 1 ms in seeks. Maybe something > is throttling the I/O rate? > > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?224a745a0704120942i19c42927o9826b0343938d093>