It may well be as you said, because Oracle Corporation thinks grouping all
Oracle "processes" into one OS process makes it run faster on NT. But I'd like
to see some kind of official answer from Oracle.
|| I do not have the ability or inclination to make this kind of information
up. I got my information from OWS directly. You are free to do your own
research, of course.
Multithreaded Oracle.exe has its own problem. For example, if one thread breaks, the
entire process hangs or dies.
|| Definitely not true. Cite a source, can you? Do you mean to assert that if a
user connection thread breaks, the entire database system hangs?....No, I didn't think
so....Now, on the other hand, if the PMON thread dies, then the instance goes down, yes.
Just like on Unix, and just exactly the way it should be.
It probably makes sense to bundle all essential background "processes" into one
process because if one of them dies, it's meaningless to have all the others
continue running. But server "processes" running on behalf of user programs as
well as non-essential Oracle background "processes" are also part of oracle.exe,
aren't they? Wouldn't Oracle make a decision to favor stability over performance,
in view of the generally accepted instability of NT?
|| See above. There is the thought inside oracle ( from the note i read 46000.1 and
46053.1 ) that thread mgmt is faster and more stable than process mgmt. As for
your assertion of a "generally accepted instability of NT"...<shrug>...
BTW, NT also supports shared memory and of course context switches but may be
inferior to the counterparts on UNIX (I don't know).
|| NT may well support it. That was not the topic of the post. ORACLE on NT does not need it,
nor does it use it.
I made a mistake in my previous message saying "Oracle on NT runs as one
thread". I meant "one process".
|| Ayup.
