Matthew Dillon writes: > Well I'll be a cat in a hat! IDE DMA actually works on my old > PPro-200 box with a new IDE drive! A solid 8 MBytes/sec out of the > thing. My poor old 4 GB SCSI barracuda could only do 4.3 MB/sec. > Time to replace the barracuda, I think.
What did you use to measure? Bonnie? For kicks time something interesting such as a CVS checkout with both source and destination on the same drive. While the burst thruput is higher am curious what the effect on real world seeks is. 4G Barracuda that only does 4.3 MB/sec? An ST15150[NW]? My narrow 5400 RPM IBM DCAS has been faster than 7200 RPM ST15150-either-N-or-W on SGI, Mac, and FreeBSD hosts. Haven't done the same tests but ST34173's are faster than the ST15150's. HD on my "good" system is an IBM DCHS09. Haven't checked recently but bonnie used to peak at around 8.5 MB/sec. In practical applications such as a CVS checkout or sorting mail thru slocal I notice the practical limit seems to be 80 to 100 tps as reported by "systat -v". And at that rate its often doing only 600k/sec (small transaction blocksize of 4k). Have seen it reach 120 to 140 on occasion. And once it it 150. Caught this just now in a "make release", it just doesn't go much faster than 143 tps: Discs fd0 da0 sa0 sa1 pass0 pass1 pass2 intrn KB/t 0.00 9.49 0.00 0.00 0.00 0.00 0.00 8337 buf tps 0 143 0 0 0 0 0 8461 desiredvnodes MB/s 0.00 1.32 0.00 0.00 0.00 0.00 0.00 18768 numvnodes Recently I enabled optional flags on a 2.2.7 system's wd driver. For lack of any better guess, used 0x80ff. Got a noticable improvement on a Gateway P-133 and the OEM 1G HD but haven't bothered to attempt accurate measure. -- David Kelly N4HHE, dke...@nospam.hiwaay.net ===================================================================== The human mind ordinarily operates at only ten percent of its capacity -- the rest is overhead for the operating system. To Unsubscribe: send mail to majord...@freebsd.org with "unsubscribe freebsd-current" in the body of the message
