[HECnet] Multinet circuit costs

[Johnny Billquist]

Why? What is the percieved problem?

For MIM atleast, the changes in the routing tables were expected, and benifical. From MIMs point of view, the network worked before, and works now. However, it now works better, as response times for areas far away have improved.

What is the problem at LEGATO? I haven't heard anything, but if there is a problem there, we should try and figure out why. The fact that there are multiple paths should not be a problem. Yes, there is still broadcast packets on all links, just as before. Nothing have changed from that point of view. But in some situations, actual traffic will now take different paths than before. In almost all situations, this should be when there are multiple hops to get to the destination. And traffic will now favor multinet links a bit more.

But unless someone really place a much higher cost on their ethernet circuit, it will not be unreasonably favoritism of multinet links.

	Johnny

On 2012-01-05 08.48, hvlems at zonnet.nl wrote:
DECnet supports multiple paths between hosts (e.g. Ethernet and CI) and circuit cost is a way to favour one path over another.
In situations where a bridge and a Multinet TCP tunnel operate in parallel that still leads to multiple paths. Even if the TCP tunnel has the lowest cost the ethernet path will send out multicast messages. I think that is the reason LEGATO has issues (cannot copy files to area 44) and possibly why MIM now sees a different world.
The map shows several parallel paths. I'd suggest   in stead of modifying cost it is better to have either the bridge program or a TCP tunnel per site for a couple of days and see what happens.
Hans
------Origineel bericht------
Van: Johnny Billquist
Afzender: owner-hecnet at Update.UU.SE
Aan: hecnet at Update.UU.SE
Beantwoorden: hecnet at Update.UU.SE
Onderwerp: Re: [HECnet] Multinet circuit costs
Verzonden: 5 januari 2012 00:48

On 2012-01-04 23.31, Steve Davidson wrote:
Have we decided on what the circuit costs should be set to for the
Multinet circuits? I just did:
NCP TELL MIM SHOW KNOW AREAS
and discovered that MIM's view of the world has changed. From what I can
tell it is due to changes is circuit costs. I have made no changes at my
end. I was waiting for consensus before making any changes.
-Steve

I don't know if a general consensus was reached, but I agree with the
idea expressed. When you have both the bridge, and Multinet DECnet
tunneled, make the costs for the Multinet circuits lower.

Yes, the changes made by some have had effect on how the routing from
MIMs point of view looks. And I think it's an improvement. A few
destinations farther away that I tested against were noticeable faster now.

	Johnny