🚀 go-pugleaf

>> 1) IP is 85% of the traffic, Decnet is 10% and Appletalk is 5% which
>>    is what I would have guessed to begin with.
>> 2) For the amount of packets handled by Decnet, a very large percentage
>>    is Hellos.  What can be done to reduce the amount of Hello traffic?

As was mentioned previously, you can control the hello update timer on a
per interface basis. This is especially important for X.25 links.

>> 3) For the amount of traffic that Appletalk and Decnet generate (15%)
>>    it appears to be taking much too much cpu as compared with IP
>>    processing.  Can anyone from cisco give an explanation - or are the
>>    Decnet and AT protocols not as optimized in the cisco as are the IP
>>    protocols?

Traffic doesn't necessarily relate to CPU overhead. How many routing/hello
updates for DECNET and ZIP/RTMP/NBP requests for AppleTalk is the router
having to handle? Flapping interfaces can cause more CPU to be consumed as
routing information is flushed and updated. DECNET and AppleTalk are much
more chatty. RTMP is a 10 second RIP-like protocol. If an AppleTalk network
is learned, the router will send out a ZIP request to obtain a zone name.
If someone advertises a network without properly configuring a zone, all of
the AppleTalk routers will send a ZIP request once per second without much
satisfaction. This is why we leave our debugging commands in the code so
you can determine if this is what is happening on your network.

Greg Satz
cisco