LEE ARONER wrote in a message to MIKE BILOW:
 LA>    Since I AM running HPFS386, this may be why I was starting 
 LA> to consider the wisdom of doing so...
 LA>    The biggest problem seems to be with a 2 gig drive 
 LA> partitioned 1.7 gig in HPFS386 and the balance in FAT. Gets 
 LA> fairly constant errors from chkdsk (ONLY on the HPFS 
 LA> partition), necessitating a boot from floppy to fix... annoying 
 LA> to say the least.
Well, there are probably errors on the FAT partition also, but you don't find 
out about them easily because FAT has no internal consistency checking.  HPFS 
uses doubly linked lists and embedded cyclic redundancy checks to validate a 
number of its internal structures, so CHKDSK or similar utilties can detect 
corruption much more reliably.
If you are asking whether there is any likelihood that having both a FAT and 
HPFS partition makes it any more probable that corruption will occur on the 
HPFS partition, then the answer is no.
Don't overlook the obvious, either.  Any drive which frequently reports 
errors should be examined very carefully, especially with regard to cabling 
and termination.  Write cache comes into play only in terms of maintaining 
integrity of the data after the drive has reported some error condition, so 
you need to find out why these error conditions are being reported in the 
first place.  Assuming you have applied all of the relevant FixPacks, HPFS386 
is pretty solid, but you may have SCSI driver or hardware problems.
MB> You can get Seagate's official utility to manipulate the write cache
MB> enable bit on the appropriate device mode page, APSI-WCE, at:
 LA>    Got it yesterday. 
Let's see if that has any effect.  I think you will need to run that under 
DOS with ASPI support loaded, which can be a pain.  You might be able to run 
it under OS/2, since VASPI support is provided, but I've never tried it.  I 
would be interested to know if that works.
MB> I can save you a lot of aggravation that you will encounter 
MB> when trying to make sense of the Seagate docuemtation: for 
MB> OS/2, turn off the write cache.
 LA>    For Warp workstations, as well as Server?
Yes.  The reasoning applies to any filesystem which can do hotfixing, 
including HPFS and HPFS386.  If you run FAT only, then it makes no 
fference.
 LA>    I'll bet that Win95 would probably do OK with it on?
Yes.  Windows 95 has no support for a filesystem that can do hotfixing.  The 
goal of disabling write cache is to get error reporting that can be tied back 
to the appropriate data buffer and handled within the filesystem logic.  Even 
if Windows 95 is properly notified of a media error with write cache 
disabled, it will simply cough and sputter anyway and throw away the data.  
Windows 95 uses VCACHE, which is basically just a VxD port of SMARTDRV.
 LA>    And I'll bet that NT on the same machine will turn into a 
 LA> *real* pig without it.
No, the performance penalty of disabling hardware write cache is fairly 
minimal, as long as there is efficient cache logic at the filesystem level in 
the operating system.  Most of what hardware write cache does is compensate 
for inefficiencies in the filesystem logic.  NTFS is kind of a pig anyway, 
but it is probably not fair to evaluate it on performance when its stated 
design goal of journaling will have an inherent performance penalty; NTFS is 
a lot slower than HPFS386, but this is because HPFS386 does not do 
ournaling.
 
-- Mike
--- 
---------------