OSPF intro: picture about link down, loopback title, ecmp

ospf-intro/README.md

@@ -348,10 +348,14 @@ Here's mine:
      3. 10.0.1.8                          0.9%   230    0.1   0.1   0.1   0.4   0.0
      4. 10.34.2.5                         0.9%   230    0.1   0.1   0.1   0.4   0.0
 
+![OSPF network, reconvergence](/ospf-intro/ospf-together-hosts-linkdown.png)
+
 When I disabled the interface on R5, BIRD on R5 got notified by netlink that the interface went down. OSPF on R5 had to change its information card immediately and send it out again. But... it was only able to send it out on the 10.0.1.0/24 network. So it did, and R1 and R6 received it. Since R1 had not seen an update on the lower side of the network, it notified routers in there of the change and R2 was able to recalculate the shortest paths to the entire network after changing its view of the complete network topology with the missing link between R5 and the 10.1.2.0/24 network. After doing so, R2 determined that the current open shortest path to 10.34.2.5 had to be via 10.1.2.7 and used the BIRD kernel protocol to retract the route to 10.34.2.0/24 via 10.1.2.56 and inserted a new route into the Linux kernel routing table which points to 10.1.2.7 as next hop for 10.34.2.0/24. And then, mtr noticed there was a change in the path.
 
 Apparently, I lost a ping while the network was busy to get into a stable converged state again. ;-(
 
+## The loopback address
+
 The second thing I want to point out is about the /32 addresses on the loopback interfaces of the routers. I figure you might be wondering what they're useful for. Well, normally, a /32 address on a network interface would not make much sense. But image what happens when we include it in our OSPF process... It suddenly becomes a network subnet whose reachability information is propagated throughout the whole network. Ok, this subnet can only contain a single address, but it's a perfect way to make sure that if any path exists to this single router in the network, OSPF will make you able to use it to connect to the router. So, if I'm the network administrator of the example network we've just built, and `10.50.1.12` is my workstation, I can use `10.9.9.5` to connect to, for example with SSH, to manage this router. Even when I accidentally would disable the link to the `10.1.2.0/24` network, my SSH session would simply stay active, the traffic to and from R5 being rerouted via R1 back to my workstation... :-D Later on, in the BGP tutorial we'll see that there are actually other routing protocols that rely on this mechanism to function correctly.
 
 ## Next...
@@ -361,7 +365,7 @@ There are numerous pages with information about OSPF on the internet. Since I co
 First of all, don't forget to take a look at the BIRD documentation about OSPF. You can find it at User's guide -> Protocols -> OSPF at the [BIRD web page](http://bird.network.cz/). There's a lot more options than "stub". :) While I just proved you don't need to know about them to set up an interesting network with dynamic routing, there must be scenarios in which they can be very useful. For example:
  * If there are untrusted hosts inside your routing vlans, you might want to use password authentication.
  * If you want to decrease the time until the network gets reconfigured when a router crashes without notifying anyone, you might want to play with hello timers, or even bfd.
- * Equal cost multipath routing (ECMP) is a big thing nowadays, which is used a lot to load balance traffic over multiple paths to a destination instead of choosing only one as best path. You can even enable that in the network we just built by just specifying `ecmp yes` in the OSPF configuration and see what effect it has on the output of `ip r` on the linux command line. Just search for information on it on the Internet to learn more.
+ * Equal cost multipath routing (ECMP) is a big thing nowadays, which is used a lot to load balance traffic over multiple paths to a destination instead of choosing only one as best path. You can even enable that in the network we just built by just specifying `ecmp yes` in the OSPF configuration (try it on R2 or R6) and see what effect it has on the output of `ip r` on the linux command line. Just search for information on it on the Internet to learn more.
  * 'Cost' is an aspect that is fundamental to OSPF and the calculation of the shortest paths in the network. Traditionally, cost is related to the bandwith of a link between routers, and causes higher bandwith connections to be prefered above lower bandwith connections. Since we're working with switched Gigabit/s networks by default now, if it's not 10Gb/s, in the datacenter and even in our office, I've just been ignoring that.
 
 Well, after completing this tutorial, I also encourage you to start reading the other "An Introduction to OSPF" like pages on the internet, since they should be a lot easier to understand while having seen it work for real! Have fun.