hobson@madness.rutgers.edu (Kevin Hobson) (05/13/91)
Sorry to cross post this message but I will like to know
something about netware under 3.##. If this has been ask before sorry
but I have not be reading these groups for the past 2 months.
Previously, when our group in Computer Services install 3.10,
we found out that Novell change the rules on us with Novell network
numbers. Novell requires an internal network number that has nothing
to do with the particular novell network number involved (see picture
below). This internal network number is advertised when I do a "show
novell servers" on a cisco router running novell routing. I "assumed"
incorrectly that it would show me that external network number of the
particular media cisco was connecting the server to. So we had to come
up with a new numbering scheme throughout the university novell
network in order to know what was on what novell network media if we
had redundant networks.
Cisco============================Server-A================================Cisco
^-novell network2(external) ^----novell network 1(internal) ^
|
novell network 3(external)
This past Thursday, a collegue of mine ask for my help
bringing up 3.11 and appletalk services. Again, novell requires an
internal appletalk network number besides the external appletalk
network number. And yet again, the network number being adverstise is
the internal number instead of the media network number. So for every
novell appletalk server that is brought up, 2 appletalk network
numbers are required.
Below is an output from both novell and appletalk from cisco novell
services and CAP package on a SUN. We at the university base our
protocol network numbers on the IP network numbers. I will explain
further below.
Cisco (busch-gw)
-----------------
busch-gw>show novell server
Type Name Net Address Port Hops Interface
4 RUCS-PROTO 80065315.0000.0000.0001::0451 2 Ethernet4
....
107 RUCS-PROTO 80065315.0000.0000.0001::8104 3 Ethernet4
....
busch-gw>show novell interface ethernet 4
Ethernet 4 is up, line protocol is up
...
Novell address is 80065300.aa00.0400.20f4
...
So for novell protocol, a Novell server is on IP network
128.6.83.0 (subnetted Class B network) with host number 15, it's
hex-equivalent external novell network number is 80065300 (based on IP
network number) and internal network numbers 80065315 (based on IP
host number). You will notice that the SAP services are showing up as
network number of 80065315 but cisco interface indicates that it is
network 80065300.
madness$/usr/local/cap/atlook Etalk-83
--------------------------------------
abInit: [ddp: 18.00, 119], [GW: 18.00, 1] starting
Looking for =:=@Etalk-83 ...
1 - RUCS-PROTO:AFPServer@* [Net: 83.1 Node: 1 Skt:253]
2 - busch-gw:ciscoRouter@Etalk-83 [Net: 83.0 Node: 1 Skt:254]
For appletalk protocol, we multiple 256 times the subnet number plus a
number from 0 to 256 to come up with the decimal equivalent appletalk.
So for the above appletalk numbers, these services are on IP network
128.6.83.0. My appletalk sun is on appletalk network 18.00 (call KIP
notation). I query appletalk zonename "Etalk-83" for it services.
RUCS-PROTO is a appletalk novell server advertising AppleShare
services. Noticed that it shows it is on network 83.1. The cisco
router, busch-gw, shows the correct network number (83.0) on that
media. RUCS-PROTO does not have any other network media cards in it.
RUCS-PROTO is sending out correct network routing information on that
particular media of 83.0.
My short term problem is as follows: I, as a network manager,
will have to have an idea that it is on IP network of 83 should
problems happen. It is hard to teach this easily to operators when I
shut down novell (or appletalk) on the network in question and users
start calling them for help (I cannot print to foobar.). What are
other organization during for network number and a global database for
these different protocols? I am thinking of merging the appletalk,
decnet, IP and novell network databases together so I (and others) can
use the IP domain system to find out network numbers and services.
A larger long term problem has to do with the number of
network numbers. There seems to be enough novell network numbers to
allow 2 network numbers to go with a novell network server but I
question the appletalk scheme since you only have 64K (256 network
numbers per media) network number under appletalk phase 2. You would
need only 256 appletalk novell servers times 256 media before you run
out of network numbers. I know this is lot of machines but there are a
lot of departments/organizations at our university. They have separate
needs. Also with the talk about connecting with other university
network services (through protocol tunnelling), this number can easily
become small quickly. Rutgers university has already over 90 AppleTalk
and 40 Novell media networks in one campus (less than 2 years).
Different organizations will use different gateways/routers to get to
these services. Since our group plans on campus/university scale, I
can easily see novell services being used more and more to combine IBM
pc and Macintoshes services at the university. Again, what are other
organizations doing about this?
Thanks for any information you can supply.
--
Kevin Hobson Internet: hobson@rutgers.edu
Rutgers - The State University UUCP: {backbone}!rutgers!hobson
P.O. Box 879, RUCS, Hill Center, Busch BITNET: hobson@{cancer,pisces}.BITNET
Piscataway, N.J. 08855-0879 PHONE: (908) 932-4780brianb@kinetics.com (Brian Bulkowski) (05/21/91)
Hi, Perhaps I can do two things: explain what our thoughs were on having an internal network, and tell you what we are doing about it. I know about the AppleTalk side, and caught the note on the AppleTalk side, thus will only speak about NetWare 3.0 for AppleTalk. The decision to have an internal network was born out of lack of time, and thoughtfulness. If a server has several cards and is running with routing, then there are several nodes involved. There will be one node per media that the router is attached to. There can also be a seperate node for the protocol stack (that thing that printing and files services come from). Given the code we started with, our first release needed to have a node for the protocol stack. If this node was on any of the networks that the router talked to (a REAL network), then partitioning of the net could give unwanted results. For example, if we have two interfaces, with net numbers 1 and 2 on the cable. The protocol stack is on net #1. Suddenly, the cable on the #1 side becomes unplugged. There is a duplicate route between net 1 and net 2. If the router in the box advertises a route to net 1, it could get packets it can't deliver. If the router does not advertise, it will never get packets it could deliver to the stack. Thus, the stack will never get packets, and will drop all connections, even though there is a perfectly good connection between the stack and the outside world. However, if the stack is on an internal network, it will be independant from any physical media, thus be able to communicate even though connections go up and down around it. Even if those connections involve changing configurations: adding and deleting cards, phases, what have you. Thus we thought it was a good feature. Around here we have lots of NetWare for Mac servers. What we do for internal network numbers is use the 4 digit extensions of our phone numbers, and prepend a 3 or 4. I don't see a real problem with running out of net numbers, because the only reason for using a network range is because you can't fit all your nodes (256) on one network number. Thus, a 500 node network would need a cable range of 2. Internal nets, needing only 2 nodes (one for the router, one for the stack) are non-extended and only need a single number. Thus, you could easily accomodate 1000 servers (one net a piece) and the worst case you could have, 1 net per node, would leave you enough room for 63,000 machines. I don't believe there is a single appletalk internet so large, so it isn't a problem yet, and if you are scrounging for the last 1000 net numbers you have a bigger problem, AppleTalk scalability, which is outside of the scope of this humble engineer. I don't understand the argument of 255 servers and 255 media types. I think it had to do with the way you assigned network numbers, thus it seems you need a better way. We do recognise the need of "no internal networks". Thus, in a future release, we are planning on a configuration with no router which will also have no internal net. Thus, one card, one server, no router, no extra routing traffic, no configuration problems. If anyone thinks that there is a good reason to have the router and no internal net, please speak up. Be prepared to counter the argument in paragraph 2. Thanks for raising this question, and if this is the worst problem you have with NetWare for Mac 3.0, I think I'm happy :-). Cheers, brianb In article <May.13.01.22.33.1991.6353@madness.rutgers.edu> hobson@madness.rutgers.edu (Kevin Hobson) writes: > > Previously, when our group in Computer Services install 3.10, >we found out that Novell change the rules on us with Novell network >numbers. Novell requires an internal network number that has nothing >to do with the particular novell network number involved (see picture >below). This internal network number is advertised when I do a "show >novell servers" on a cisco router running novell routing. I "assumed" .....