Netbios alias -max names or chars

[David Collier-Brown]

"Wilson, Mark - MST" wrote:
> In terms of Network traffic I don't think that there will be much difference
> as these apps already have their own netbios name, just spread around a
> number of machines. I want to make sure that we *can* put that many netbios
> aliases on the one samba instance.

	I think you could do a fairly easy experiment
	to get a ball-park estimate of the extra
	overheads, with nothing but a small workstation
	and a sniffer.

	Set up the sniffer (i.e., one such as snoop) so
	it only reports on netbios packets, then start
	the samba server with 1, 10, 100, and 1000
	netbios names. record the time to finish the 
	bookkeeping, and you should get a plot something
	like this:

                                  X|
                                   |
                                   |
                                   |
                                   |
                                   |
                                   |
                         X         |
                                   |
                                   |
                X                  |
             X                     |
             +--+--------+---------+

	Do a few more tests to see where the inflection
	point is, then plan to stay below that.

	Note that you should have several NT and windows
	servers on the net where you do this, or all you'll
	see is samba's own startup overheads.  

	Someone more familiar with netbios negotiations
	than I want to add more advice and caveats?

	This information is also useful to the Samba 
	team, so we'd like to hear your experience here[1]

--dave 
[1. As would I, being a Sun performance engineer. Feel free
    to send me mail.]
-- 
David Collier-Brown,           | Always do right. This will gratify 
Performance & Engineering Team | some people and astonish the rest.
Americas Customer Engineering  |                      -- Mark Twain
(905) 415-2849                 | davecb at canada.sun.com