> 10K RPM SCSI disks will get (best case) 350 to 400 IOPS. Remember, > the main issue with legacy SCSI is that (SCSI) commands are sent > 8-bits wide at 5Mbits/Sec - for backwards compatibility.
This is true for really old SCSI configurations, but if you're buying a modern disk and controller (U160/U320), commands are packetized and sent at full rate. > You simply can't send enough commands over a SCSI bus to busy > out a modern 10k RPM SCSI drive. If you mean 'small sequential i/o requests', you're probably true. For random i/o requests, you can pretty easily busy it out, and for large requests, the command overhead is negligible. The bigger advantage of SAS is its point-to-point nature which means that when you have multiple disks there's no queueing delay at the bus. So I'd agree with your recommendation to prefer SAS to parallel SCSI :-) but would just like to point out that the command-transfer-speed bottleneck was dealt with many years ago (ca. 1999). This message posted from opensolaris.org _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
