The earlier weblog put up on this sequence mentioned how TCP/IP and CLNP attain adjoining nodes and construct ARP/ND/ES caches. Now let’s transfer one step additional: how do nodes operating IPv4/IPv6 or CLNP uncover the first-hop router that would ahead their site visitors to off-subnet nodes they need to talk with?
We’ve already lined work out if a node is non-adjacent. In TCP/IP, a node is non-adjacent if its subnet doesn’t match ours. In OSI, a node is non-adjacent if we don’t have it in the long run system good day cache.
No commonplace mechanism was constructed into IPv4 to find out who the adjoining router was. Every host within the IPv4 world has a routing desk that it makes use of to ahead packets, and one of many entries within the routing desk is often the default route – the route you employ whenever you don’t know the place to ship the packets.
Nevertheless, each route wants a subsequent hop. How do you determine the following hop for a default route? Within the early IPv4 days, the reply was easy: handbook configuration. Years later, after they lastly applied a centralized IPv4 configuration with DHCP, they added an choice to DHCP to specify the default route. Nonetheless, the default route all the time had some statically configured subsequent hop, which needed to match the IPv4 tackle of the adjoining router.
What occurs when you’ve got two routers? We needed to implement one other clutch as a result of this was not constructed into IPv4. Cisco Methods was the primary vendor to understand there was an issue if we had a number of routers that needed to share the identical IPv4 next-hop tackle, and developed a proprietary protocol (HSRP) to permit routers to determine which one in every of them would personal the shared IP tackle. As a aspect impact, all hosts would use only one router as a result of a number of routers couldn’t have the identical IP tackle on the identical time – it was inconceivable to implement load balancing the place all routers hooked up to a subnet would ahead a few of the off-subnet site visitors.
To handle that, Cisco applied GLBP, which used a number of MAC addresses for a similar IP tackle. Alternatively, you may use the soiled trick of getting two IP addresses with HSRP and routers backing one another up, however that was a nightmare since you needed to inform finish hosts to make use of two totally different subsequent hops in DHCP replies.
Anyway, a couple of years later, we acquired VRRP, a regular first-hop redundancy protocol (FHRP). VRRP is now on its third model and helps IPv4 and IPv6.
CLNP (OSI networking stack) by no means had that downside. Routers ship periodic Intermediate System Hellos (ISH), and the top hosts are all the time conscious of all of the routers linked to their subnet. If a number, based mostly on the top system good day cache, found {that a} node was non-adjacent, it might simply ship the packet to the intermediate system marketed with an intermediate system good day. Within the OSI world, you may have many routers hooked up to a LAN phase, and the host would determine which routers to make use of.
IPv6 copied the CLNP performance. It has Router Commercial (RA) messages, that are particular ICMPv6 messages through which the routers say, “Hey, I’m a router (and that is the IPv6 prefix we use on this hyperlink).” After receiving RA messages from a number of routers, the hosts can determine to make use of one or all routers; the routers might additionally promote their relative precedence within the RA messages.
Nevertheless, OSI and IPv6 share the identical downside: when a router goes down, its RA commercial takes a short while to outing. Throughout this era, the host might ship site visitors to the disappeared router, leading to a brief blackout. CLNP by no means acquired an answer for that downside (everybody stopped engaged on the OSI stack nicely earlier than path excessive availability turned a major difficulty); within the IPv6 world, we nonetheless use VRRPv3 once we need to have quick failover in case of router failures.
Lastly, it might be unfair to not point out the IPv4 ICMP Router Discovery Messages laid out in RFC 1256, an equal to IPv6 RA messages, however created years earlier than IPv6. It was accessible on some routers (together with Cisco IOS), however I’ve by no means seen it utilized in a manufacturing setting. Did you? Please go away a remark!
What I’ve seen, although, was a nasty hack. You’ll run RIP on the sting segments, and the hosts would hearken to RIP updates. They wouldn’t have an interest within the RIP replace’s contents; they might simply hearken to who was sending RIP messages and determine to make use of the supply IP tackle of the RIP messages because the first-hop router.
