jeff@tc.fluke.COM (Jeff Stearns) (05/26/88)
I have an existing Ethernet with about 100 hosts. I'm considering the
addition of ten or so FastPaths to our Macs and PC's can run TOPS or
AppleShare and be served by our UNIX hosts. It would look something like
this simplified diagram:
UNIX UNIX UNIX UNIX ( ~~ Vitalink bridge ~~~ )
hosts hosts hosts hosts | to distant nets
|| || || || |
|| || || || | [bridges to]====
===========================================================[ other ]====
|| || || [Ethernets ]====
[Fastpath] [Fastpath] [Fastpath]
| | |
| | |
| | |
+Mac +Mac +Mac
| | |
+Mac +Mac +Mac
| | |
+Mac +Mac +Mac
| | |
+Mac +Mac +Mac
Question 1: AppleTalk protocols emit LOTS of broadcast packets at boot
time. Will those broadcasts pass through the FastPath and
pollute our extended Ethernet?
Question 2: If so, can the concept of Zones be applied to control this
in any way? How about configuring each FastPath with a different
network number (or is this a foregone conclusion...)?
Question 3: If the FastPaths were replaced by Cayman GatorBoxes, would the
broadcast situation be changed in any way?
Thanks for any experience and advice!
--
Jeff Stearns
Domain: jeff@tc.fluke.COM
Voice: +1 206 356 5064
If you must: {uw-beaver,microsoft,sun}!fluke!jeff
USPS: John Fluke Mfg. Co. / P.O. Box C9090 / Everett WA 98206sbm@PURDUE.EDU (05/27/88)
As I understand it, FastPaths running KIP only propagate IP broadcasts to the Ethernet. Back when there weren't any AppleTalk hosts on the Ethernet, it didn't make sense to put AppleTalk traffic on the Ethernet. NBPLookup requests on the AppleTalk network, for instance, were answered by the FastPath, which used ARP on the Ethernet side to find hosts that were not on the AppleTalk network. I assume KIP still works this way. Steve Munson sbm@Purdue.EDU sbm@Purdue.CSNET ----------
farallon@well.UUCP (Farallon Computing) (06/02/88)
Jeff Meyer at Fluke asked whether K-boxes pass throught the burst of packets that occur at startup time, thereby clogging up ethernet. The burst of packets comes from the need to do a node address bid. These are 3-byte LAP packets that seek to ensure that no other device has the node address that the newly-booted device intends to use. The node address only has to be unique within the zone (defined by the K-box usually). The Kinetics box does not need to pass on the packets and does not. -- Kurt VanderSluis Voice: (415) 849-2331 Farallon Computing Email: farallon@well.UUCP 2150 Kitteridge AppleLink: D0162 Berkeley CA 94704 FAX: (415) 841-5770
Ravinder.Chandhok@GNOME.CS.CMU.EDU (Rob Chandhok) (06/02/88)
>Jeff Meyer at Fluke asked whether K-boxes pass throught the burst >of packets that occur at startup time, thereby clogging up ethernet. The >burst of packets comes from the need to do a node address bid. These are >3-byte LAP packets that seek to ensure that no other device has the node >address that the newly-booted device intends to use. The node address >only has to be unique within the zone (defined by the K-box usually). >The Kinetics box does not need to pass on the packets and does not. >-- >Kurt VanderSluis Voice: (415) 849-2331 One more time, with feeling. The broadcast burst is LOCAL TO THE CABLE. It has nothing to do with zones, or nets, just node numbers. It is a LAP level braodcast, which does not get forwarded by the gateway (KBox). LAP only knows about node numbers, you need DDP to use a net number. In general, the "node address bid" is a local cable hardware broadcast, which is why there is a problem on Ethernet if you use something like a LanBridge and EtherTalk. Rob
jmg@cernvax.UUCP (jmg) (06/09/88)
In article <849.581266019@GNOME.CS.CMU.EDU> Ravinder.Chandhok@GNOME.CS.CMU.EDU (Rob Chandhok) writes: > In general, >the "node address bid" is a local cable hardware broadcast, which is why >there is a problem on Ethernet if you use something like a LanBridge and >EtherTalk. This tends to imply problems with a LanBridge on very fast back-to-back packets. Since we have some suspicions about such cases, having many such bridges around, as well as Ethertalk, could we have some more details on this problem? (maybe this ought to be in the lan newsgroup! -- _ _ o | __ | jmg@cernvax.uucp | | | | _ / \ _ __ _ __ _| jmg@cernvax.bitnet | | | | |_) /_) | __/_) | (___\ | (_/ | J. M. Gerard, Div. DD, CERN, | | |_|_| \_/\___ \__/ \___| (_|_| \_|_ 1211 Geneva 23, Switzerland
Ravinder.Chandhok@CS.CMU.EDU (Rob Chandhok) (06/14/88)
Hi, I didn't mean to imply that LanBridges could not hanlde back-to-back packets, just that you don't want the flurry of broadcasts propogating through your Ethernet (which will happen with a LanBridge) Rob