Date: Fri, 12 Apr 1996 05:57:47 +1000 From: Bruce Evans <bde@zeta.org.au> To: luigi@labinfo.iet.unipi.it, toor@dyson.iquest.net Cc: bde@zeta.org.au, dutchman@spase.nl, freebsd-hackers@FreeBSD.org, se@zpr.uni-koeln.de Subject: Re: HDD cpu usage (IDE vs. SCSI). Message-ID: <199604111957.FAA04906@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>> NCR-SCSI driver ASUS SC200 W/ST12400 2.1GB Hawk
>> Command overhead is 1370 usec (time_4096 = 2093, time_8192 = 2815)
>> transfer speed is 5.6707e+06 bytes/sec
>> Command overhead is 839 usec (time_4096 = 2093, time_8192 = 3347)
>> transfer speed is 3.26577e+06 bytes/sec
>First I assume you measure the overhead as
> t4K - (t8K - t4K)
>but the numbers for the NCR probably suggest that data ought to be
>averaged over a larger number of experiments.
Ignore experiments where the transfer speed isn't within a few percent
of the maximum possible transfer speed. The transfer speed is only
printed so that you can do this non-automatically. For SCSI controllers
it is usually 10MB/sec (is that 10*100*100 or 10*1024*1024?). For IDE/
EIDE controllers it is unknown in general.
>Second, I guess you are measuring the SCSI/IDE/EIDE bus transfer speed,
>rather than the disk transfer speed, isn't it ?
No, we are attempting to measure command overhead.
>Also, how comes that a 2X CDROM has such a low speed ?
Perhaps the drive doesn't buffer it.
>> SCSI 2.2X CDROM
>> Command overhead is 5655 usec (time_4096 = 9033, time_8192 = 12410)
>> transfer speed is 1.21281e+06 bytes/sec
>> Command overhead is 5670 usec (time_4096 = 9046, time_8192 = 12422)
>> transfer speed is 1.21324e+06 bytes/sec
It's 14 times faster than mine :-]:
MATSHITA CR-533
Command overhead is 131637 usec (time_4096 = 130286, time_8192 = 128936)
transfer speed is -3.03353e+06 bytes/sec
^^ garbage
and 2 times faster than my Zip drive:
IOMEGA ZIP 100
Command overhead is 20410 usec (time_4096 = 20410, time_8192 = 20410)
transfer speed is 3.9767e+10 bytes/sec
^^^^ garbage
These tests took so long that the command overhead estimates are probably
accurate despite the garbage transfer speed estimates. They show that it
takes little or negative extra time to read more data. The speeds for
dd'ing the Zip drive for 40 seconds are consistent with the estimated
command overhead:
block size speed 1000000.0/command_overhead*block_size
----- ------ -------
512 24800 25085 <- small frags
1024 49000 50171 <- default frags
2048 97000 100342
4096 185000 200645
8192 313000 401371 <- default single blocks
16384 492000 802743
32768 708000 1605487 <- cmd overhead stops dominating
65536 858000 3210975 <- clustered blocks
>Third, what do you mean by "Standard IDE" vs EIDE ? Isn't it rather an
>ISA vs. VLB/PCI comparison ? I have tried both a WDC540 and a WDC1.6GB,
>and the 540 gets a maximum speed (with iozone or bonnie) of some
>2.2MB/s at most, no matter how fast is the interface or the system.
>Instead, the WDC 1.6GB (as you also experienced) is rated at 5.5MB/s on
>the same system).
>> ...
>> EIDE W/WDC 1.6GB
>> Command overhead is 217 usec (time_4096 = 513, time_8192 = 809)
>> transfer speed is 1.38444e+07 bytes/sec
>> Command overhead is 196 usec (time_4096 = 502, time_8192 = 808)
>> transfer speed is 1.33987e+07 bytes/sec
The transfer speed reported by the disk should depeend mainly on the
PIO mode. Fast PIO modes are more common on PCI buses. It looks like
the above is for a PIO speed of 15 or 16MB/sec.
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604111957.FAA04906>