Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!pasteur!ames!amdcad!sun!pitstop!sundc!seismo!uunet!auspex!guy
From: guy@auspex.UUCP (Guy Harris)
Newsgroups: comp.bugs.sys5
Subject: Re: 'ps' is too slow
Message-ID: <1155@auspex.UUCP>
Date: 13 Mar 89 19:04:15 GMT
References: <1898@holos0.UUCP> <3569@sugar.uu.net> <11991@swan.ulowell.edu> <1119@auspex.UUCP> <18469@srcsip.UUCP>
Reply-To: guy@auspex.UUCP (Guy Harris)
Distribution: usa
Organization: Auspex Systems, Santa Clara
Lines: 21

>I believe that Sun does the fast ps by cacheing all of the process status
>information in the kernel, rather than having to read /etc/utmp, etc.

You believe something that is not true.  Sun does the fast "ps" by
caching a fair bit of information in "/etc/psdatabase" - amazingly
enough, this is just what the 4.3BSD "ps" does, even down to the name of
the file.  Wow, what a coincidence! :-)

(No, the "kvm" library doesn't do any magic; it's just a library that
encapsulates what used to be independently implemented in several
"kmem"-reader programs - i.e., it opens "/dev/kmem" and grovels through
it in much the same way "ps" used to do.)

More recent System V "ps"s do much the same thing, only they call the
file "/etc/ps_data".

The S5R4 "ps" will probably

	1) use "/proc", since it'll be in S5R4;

	2) use "/etc/ps_data" if necessary.