rrw@naucse.UUCP (Robert Wier) (03/13/90)
I recently was running some disk drive benchmarks using disktimer II, V1.0 and got the following: HD20SC Wren Runner Quantum 40 (external) (external) (internal) Reads 97 16 31 Writes 96 16 29 Access 33 5 0 misc info Seagate 225N CDC 94181-15 P40S V9.6 V. 7980 Note that Reads is time for 100 24kb reads in deciseconds, Writes is similar for write times, Access is time for 80 1kb seeks. Misc Info is as reported by SCSI probe V2 Configuration MacIIci, 8 meg, cache off, no appletalk Now, that Quantum access is fast, don't you think? Can anyone make some suggestions as to why the seek times seem to be somewhat off? Thanks! --------------------------------------------------------------------- "Reply" address frequently doesn't work, please use one of the below - Bob Wier Northern Arizona University summer:Ouray, Colorado winter:Flagstaff, Arizona Internet: rrw@naucse.cse.nau.edu | BITNET: WIER@NAUVAX | WB5KXH or ...arizona!naucse!rrw
ephraim@think.com (Ephraim Vishniac) (03/13/90)
In article <1944@naucse.UUCP> rrw@naucse.UUCP (Robert Wier) writes: >I recently was running some disk drive benchmarks using disktimer II, V1.0 >and got the following: > HD20SC Wren Runner Quantum 40 > (external) (external) (internal) >Reads 97 16 31 >Writes 96 16 29 >Access 33 5 0 >misc info Seagate 225N CDC 94181-15 P40S V9.6 > V. 7980 >Note that Reads is time for 100 24kb reads in deciseconds, Writes is >similar for write times, Access is time for 80 1kb seeks. >Now, that Quantum access is fast, don't you think? >Can anyone make some suggestions as to why the seek times seem to be >somewhat off? *************************************************************************** * Don't accept benchmark results unless you know how the benchmark works. * *************************************************************************** DiskTimer II is a very simple test. It obtains the read (write) time by repeatedly reading (writing) the same area of the disk. It obtains the seek time by alternately reading two sectors separated by some fixed distance on the disk. The big problem with these tests is that DiskTimer's trivially and grossly deceived by caching. The control panel cache isn't the problem, since that applies to the file system, and DiskTimer II calls the driver directly. But, any cache implemented by the driver or controller makes DiskTimer's results nearly meaningless. So, a DiskTimer II seek time of "0" means that the driver or controller is caching the two sectors involved in the seek test. BTW, it looks like your HD 20 SC would benefit from formatting with a tighter interleave. Was it originally formatted on the IIci, or on a slower machine? -- Ephraim Vishniac ephraim@think.com ThinkingCorp@applelink.apple.com Thinking Machines Corporation / 245 First Street / Cambridge, MA 02142 One of the flaws in the anarchic bopper society was the ease with which such crazed rumors could spread.
rrw@naucse.UUCP (Robert Wier) (03/14/90)
A followup note to my posting on using DiskTimer II: Several people suggested I give SCSI evaluator a try, and here are the results: Wren Runner Quantum 40 meg HD20SC external internal external access 12 ms 22 ms 70ms read rate 12,600 kbits/s 5,000 kbits/s 1,000 kbits/s supplemental information: Access times are not supposed to be affected by the machine to which the drive is attached. The 12 ms for the Wren Runner agrees nicely with the advertised 10.5 ms. The 70ms for the HD20SC is probably because it's formatted with 3:1 interleave for use on a Plus (or is it just generally a dog?). Read rates for the Runner and Quantum were when connected to a MacIIci, averaged over several runs. The HD20SC is connected to a Mac SE. (No wonder it seems slow in comparison to the ci with the Runner!) Many thanks to those who responded to my message! --------------------------------------------------------------------- "Reply" address frequently doesn't work, please use one of the below - Bob Wier Northern Arizona University summer:Ouray, Colorado winter:Flagstaff, Arizona Internet: rrw@naucse.cse.nau.edu | BITNET: WIER@NAUVAX | WB5KXH or ...arizona!naucse!rrw