1

I have encountered the following sequence of words: 'routing table formed on the basis of network history information'. I know that initially, routers discover neighbor nodes and therby the possible ways to reach every other destination in the network, based on distance.

1. So, I conclude using network history means the tables are dynamic?

2. But then for e.g., if path quality (SNR) is a metric then how often the routers would update their table?

3. And when they do, do they just update the table for links connected to neighbors?

Background: Actually, I saw papers where tabular Q-learning is applied to find a route from a given source to destination node assuming a centralized SDN. The Q-learning process took into account the network metric. And I later found the above-said bolded terms that confused me as to what advantage would table Q-learning routing would offer if we already have some routing based on network history. That led to the above 3 questions.

2
  • Routers do not discover anything by default, or out of the box. Dynamic routing protocol processes can share information if they are configured. Directly connected networks, meaning locally configured layer 3 interfaces will show up by default in the routing table of a router. Jun 13 at 0:58
  • @RonnieRoyston By directly connected networks, do you mean wired networks? Or the only definition is what you wrote in italics? And by 'will show' up do you mean even the network metric of the corresponding links? Like congestion value, etc.. Excuse me if I sound naive as I am trying to make some analogies with wired networks, and is really imp for me. Jun 13 at 2:04

1 Answer 1

2

So, I conclude using network history means the tables are dynamic?

Yes, if you use a routing protocol to exchange routing information, but routes can be statically configured, and directly connected networks are not normally dynamic if the link does not go down.

But then for e.g., if path quality (SNR) is a metric then how often the routers would update their table?

The path quality is not a normal metric, although it can be one of the metrics for EIGRP.

And when they do, do they just update the table for links connected to neighbors?

That depends on the routing protocol used. If static or directly connected networks are not redistributed or participating in the routing protocol, then nothing of those is communicated. Neither are any routes communicated without a common routing protocol between neighbors.

Routers route packets individually by the destination address in the packet header, using the routing table. Routers learn routes in three ways:

  1. Directly connected networks
  2. Statically configured routes
  3. Dynamic routing protocols

The routes are, by default, believed in that order. Also, remember that routing protocols do not route, they only exchange routing information between routers. It is up to each router configuration as to which routes are advertised to neighbors in a routing protocol. Also, there is a hierarchy for which routing protocols are believed over others. Different routing protocols have different update times; some are almost immediate, while others may take minutes to converge.

3
  • Can you please elaborate on what you mean by, "That depends on..." for the second question? Path quality was just an example but consider any metric that is different at time t and t+1 due to dynamic network operation. Then say a router A having B,C,D as neighbors thinks like, "based on the metric value, transmitting to B doesn't seem good now & so let me update my table so that I avoid forwarding to this node in my next trasmission" Does this happen I mean? So I asked if it really that dynamic such that the table updates happen at every t+1,t+2.. or there is a gap? Jun 12 at 20:00
  • Also, I am assuming there is a centralized SDN controller than can tell the nodes such metric values of their corresponding links. I come from wired communications background so I may be wrong to say something like a centralized controller in wireless communications where it is 'hop & then decide the next hop node type transmission'. ... Jun 12 at 20:01
  • Routers each make their own routing decisions independently. What gets advertised or not to neighbors depends on how the router is configured and which routing protocol is used. SDN is not often used in business networks. Routing protocols have different metrics and algorithms for propagating routes. For example. RIP uses the Bellman-Ford algorithm, while OSPF uses Dijkstra's algorithm. OSPF routers have a full understanding of all routes and routers in an area, while RIP routers only know their neighbors.
    – Ron Maupin
    Jun 12 at 20:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.