xinetd timings?

	Well, since we are running the dbserver on port 5042 (no xinetd)
and port 5044 (xinetd) you can do a loop like:
	call dbserver on port 5042 -> log the time
	call dbserver on port 5044 -> log the time
Let say you run that for 1 hour during the night, and you look at the
log file. Your point is right, initialise() in called for any request
when running on xinetd, but who cares if it counts for 1% of the total time.
		Laurent

In our previous episode, Rick Riolo said:
> 
> maybe we should do a little timing test?
> 
> my guess is yes, the setup time will take 3 (or N) times as
> long with xinetd vs without, but it will be 2 or 3 (or some small N) 
> times 20ms or some little number, so it won't really matter, 
> even if hit 10 times a second.
> 
> any other guesses?
> 
> laurent: is the below noted difference right?
> (I have no idea what xinetd does...sounds right though).
> if so, we could easily set up a test with just those parts
> (ie strip the initialise() from in.whoisd) and run a loop
> and time the too approaches (divorced from the noise
> of actually doing the lookups and across-network exchanges).
> 
> or we could do a whole bunch of real queries at different
> times to get some nice average that should wash out
> network-induced noise in the timings.
> 
>  - r
> 
> --------
> 
> > From:    Daniel Karrenberg <Daniel.Karrenberg at ripe.net>
> > To:      Laurent Joncheray <lpj at merit.edu>
> > CC:      rr-impl at ripe.net
> 
> > 
> >   > Laurent Joncheray <lpj at merit.edu> writes:
> >   > 	It is an old issue but...
> >   > 	why don't you port whoisd to be called from [x]inetd. So you
> >   > won't need any networking hack. It has been like that on radb.ra.net,
> >   > and it seems to be fine.
> >   > 		Laurent
> >   > PS: lot of advantages: do not need to run as root, IP filtering, logging,
> > 
> > The main issue here is performance. It is 
> > 
> > 	fork()
> > 
> > against
> > 
> > 	fork()
> > 	exec()
> > 	initialise()
> > 
> > For something that potentially gets hit several times a second it makes a
> > difference. Note that our server gets hit more often than yours because
> > we also are the registration database for Europe.
> > 
> > NB: If I remember correctly the code has an option to be run from
> > inetd(8). So you can do it if you like.
> > 
> > Daniel
> 


-- 
Laurent Joncheray,                                  E-Mail: lpj at merit.edu
    Merit Network Inc, 4251 Plymouth Rd, Suite C,   Phone:  +1 (313) 936 2065
    Ann Arbor, MI 48105, USA                        Fax:    +1 (313) 747 3185
"This is the end, Beautiful friend. This is the end, My only friend, the end" JM



-------- Logged at  Thu May 4 15:58:43 MET DST 1995  ---------