RE: disk problems or false alarm??

[Guolin Cheng]

Hi, Alexander,

>  Be aware, that idebus does not mean UDMA speed (like UDMA33 or
UDMA100)
>  or the speed in MB/sec. On actual PC architecture the 32bit PCI bus
> is
>  rated at 33 MHz as well the IDE bus. So setting idebus to anything 
other
>  than the default of 33 is nonsense. 
 Then that means for normal ia-32 machines, the PCI&IDE Bus speed is
mostly 33MHz/s, and so the sustainable IDE disk access speed is always
no more than about 33MB/s?  I tried several types of Linux boxes, the
long-time disk access is about no more than 35MB/s. 

 And there are no other means to accelerate disk access speed? Manuauly
setting IDE disk's access mode (UDMA2/3/4/5/6) with hdparm has no big
difference, as far as from my own experience.

 Any other ideas, besides buy ia-64 hosts? Thanks.

 --Guolin Cheng

disk problems or false alarm??

[Alexander Dalloz]

Am Sa, den 01.05.2004 schrieb Guolin Cheng um 01:31:

>   Then that means for normal ia-32 machines, the PCI&IDE Bus speed
> is
>  mostly 33MHz/s, and so the sustainable IDE disk access speed is always
>  no more than about 33MB/s?  I tried several types of Linux boxes, the
>  long-time disk access is about no more than 35MB/s.  
Which current IDE hard disk do you expect to have more then 35MB/sec to
50MB/sec sustainable speed? Do not mix maximum possible bus speed with
what you can expect from a drive in real. An remind that manufacturers
overstate with specs given for their products.

>   And there are no other means to accelerate disk access speed?
> Manuauly
>  setting IDE disk's access mode (UDMA2/3/4/5/6) with hdparm has no big
>  difference, as far as from my own experience. 
Normally that really makes no real difference. Only for cases where the
disc can deliver data from their cache.

>   Any other ideas, besides buy ia-64 hosts? Thanks. 
Why should IA64 change anything? The disk itself is the limiting factor.

>   Guolin Cheng 
Alexander


-- 
Alexander Dalloz | Enger, Germany | GPG key 1024D/ED695653 1999-07-13
Fedora GNU/Linux Core 1 (Yarrow) on Athlon CPU kernel 2.4.22-1.2188.nptl
Sirendipity 01:40:16 up 4 days, 29 users, load average: 0.50, 0.28, 
                   [ Γνωθι σ'αυτον - gnothi seauton ]
             my life is a planetarium - and you are the stars

disk problems or false alarm??

[William Hooper]

Guolin Cheng said:

>   Then that means for normal ia-32 machines, the PCI&IDE Bus speed
> is
>  mostly 33MHz/s, and so the sustainable IDE disk access speed is always
>  no more than about 33MB/s?  I tried several types of Linux boxes, the
>  long-time disk access is about no more than 35MB/s. 
To very broadly generalize 33MB/s sustainable is a limitation of
physically reading the disk, not the interface from the motherboard to the
disk.  All the ATA66, ATA100, and ATA133 specs are about burst speed
(basically reading from the disk's on board cache).  Alternatives for
faster sustainable speeds would include: faster spinning SCSI disks, RAID,
non-mechanical hard drives.

-- 
William Hooper