Subject: Re: [netatalk-admins] Using netatalk in a large, mission-critica
From: Casey Bisson (cbisson@oz.plymouth.edu)
Date: Thu Sep 24 1998 - 12:19:40 EDT
I believe that the user limit is based entirely on the appletalk protocol, and is
lifted when clients access the server via ApplShare over TCP/IP.
I use Netatalk on a FreeBSD 2.2.6 Dell 166MHz box, and have found reliability and
scalability to be very good. Odd problems related to non-static director IDs and
occaisional difficulties with CR<->LF translation are my only difficulties.
Casey Bisson
Information Technology
Plymouth State College
Peter Gutowski wrote:
> Sean A. Snyder wrote:
> > Greetings!
> >
> > I'm writing to find out if anyone on this list is using netatalk in a large
> > environment (larger than 300 users per server) that is 24x7, or nearly 24x7?
> >
> > I'm in a situation where with a product similar to Netatalk where I find
> > myself limited to 127 appletalk connections to my file server. Apparently,
> > this limitation is based upon the 8-bit socket identification number in older
> > versions of the Appletalk stack. According to section 4-5 of Apple's "Inside
> > Appletalk" book (copywrite 1990), sockets 1-127 are reserved for
> > statically-assigned sockets (SAS's), leaving sockets 128-254 for use by our
> > AppleShare clients, hence we get errors when we try to make that 128th
> > connection to the server.
> >
> > Does anyone know if Netatalk is also subject to this 8-bit
> > socket-identification number issue, or has a workaround been found or been
> > created in newer versions of Appletalk? Will using the asun IP patches get me
> > around this limitation if it exists (i.e. unlimited number of IP connections)?
>
> >
>
> Sockets are a node-level construct. As an AppleTalk device speaks to similar
> device it's address is made up of a 16-bit NET number, 8 bit-node number and
> and 8-bit Socket number. The reference you make is to a socket number whereas
> your interest seems to be in node numbers Which i believe is an 8-bit unsigned
> value yielding 254 possible entries (node=0 and node=255 having special meaning.
>
> With >254 nodes sounds like partitioning your network is in order.
This archive was generated by hypermail 2b28 : Sat Dec 18 1999 - 16:33:18 EST