Supratim...
On 2019-02-26 23:05, Supratim Sanyal wrote:
Thanks Johnny for taking the time to write and explain
(instead of
saying "go read DECnet manuals"). So, basically, CTAKAH cannot decide to
use one path for 1.13 and another one for 1.559 since they are both area 1.
CTAKAH don't even care about where 1.13 is. CTAKAH knows it is not in
area 1, so it tries to find the cheapest path to area 1. It is not
looking at the node number, or anything beyond the area routing at that
point. In short, if you need to go to another area, the only thing
DECnet does is try to find the cheapest path to that area.
And in the case of ties, it takes the node with the highest number, and
1.599 is a higher number than 1.13, even if your end destination is
1.13, it would pick 1.599.
Got it, no easy way out other than move my bunch to a
new area. Please
assign me an area. It will take me a while, though, to move across all
the toys.
Right. Allocated area 31 for you. Go ahead and start moving things over.
Let me know when you want me to update the nodename database.
BTW, interestingly, MIM has also decided that IMPVAX
is the way to reach
area 12 (Tim in Perth, Australia), despite Tim having 3 or more other
direct GRE tunnels to other area routers. Is this because IMPVAX happens
to be in the same area as MIM (unlike his other tunnels)?
No. It just happens to be the cheapest path to area 12. MIM is an area
router, so it has the full picture of all other area routers, and it can
figure out the cheapest way to get to area 12, which is through IMPVAX.
With IMPVAX, it becomes the cost to get to IMPVAX, which carries a cost
of 4 at MIM, and then at IMPVAX, the next hop is IMPRTR on QNA-1. That
hop has a cost of 4 as well. What the cost of the link between IMPRTR
and area 12 is I don't know, but this has a cost of 8+N, whatever that N
is. Any other path to area 12 from MIM will likely go through several
hosts as well, as I do not have any direct connection to area 12. And
other places then have higher costs in the end for the whole path from MIM.
Johnny
DECnet routing is kindof stupid. When a node in one area wants to talk
to a node in another area, the only thing it does is figure out how to
get to that area. And that just means find the lowest cost path to any
node in that area. And here it where it gets silly.
Until a couple of days ago, CTAKAH:: had the same cost for all
circuits. That resulted in CTAKAH choosing IMPVAX for all
communication to area 1. So even though CTAKAH and MIM have a direct
circuit between them, packets from CTAKAH (in Russia) and MIM (in
Sweden) were travelling over New York.
CTAKAH have now dropped the cost for the circuits to MIM to a little
lower, meaning it will prefer this link. But that also means that even
through CTAKAH and IMPVAX have a direct connection, all packets from
CTAKAH to IMPVAX will pass over MIM.
In essence, this is how DECnet works, and you cannot get around this.
When multiple connections to a different area exists, it is going to
be suboptimal. The one thing would could do to improve things a bit
would be to move you to an area of your own. Then both links would
become useful for more than redundancy backup.
?? Johnny
On 2019-02-26 20:53, Supratim Sanyal wrote:
This email is directed to owners of Cisco routers
with GRE tunnels;
to the best of my limited knowledge, Dave McGuire, Ian McLaughlin,
Brian Hechinger, Mark Darvill, Cory Smelosky, Peter L?thberg ... who
else?
I already have a stable GRE tunnel to A12RTR for a while now, and am
looking for end-points on other Cisco routers on HECnet to connect
to. I will be coming in from IP address 45.62.248.66, DECnet node 1.919.
If possible, please let me know what GRE end-points I can connect to.
You can, of course, DM me too.
Thank you.
Supratim
+1-240-483-8572 (cell)
--
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol