[Simh] SimH Slow VMS I/O

Villy Madsen Villy.Madsen at shaw.ca
Tue Jul 17 13:34:31 EDT 2007


I isn't really slow VMS I/O - it's excessive windows IO - but

One of the things that just twigged was the whole issue of buffering..

There would be no reason for having to write out every single byte of the virtual disk (and the windows task window certainly seems to show that there is a write for every single byte of space on the virtual disk) unless the IO is being buffered.  What really bothers me is that there are not enough bytes READ from anywhere to make up that number - although there are enough page faults to....

So what if

Windows is memory mapping the virtual disk and then making a physical copy of the disk when you do your first write to the disk - 

I know that it doesn't make any sense, but it's the only thing that I can think of that agrees with the evidence that I am seeing.

What I would like to do is to rewrite the IO for the disks to use the CREATEWRITE interfaces.  Then one can try disabling buffering (although one would also have to change the buffer allocation to use FALLOC (io buffers must be disk page aligned if buffering is disabled)).  I started to do it but got carried away and started to change all of the IO.  I am sure that it will never work, so when I get back from vacation I will try that approach but limit it somehow to only do it with the RQ units.

As you should have guessed by how, I am very C challenged so if somewhere more comfortable with C is willing to check this out I would be very happy.  If that plea falls on deaf ears, I will try it out once I get back from vacation at the beginning of August.

Villy

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20070717/a35a595e/attachment-0003.html>


More information about the Simh mailing list