We have a network of about 1500 routers and switches, and finding MTU mismatch is quite a chore. Not only you have to collect MTU information from every Physical and Virtual interface, but we need to pair the connections as per connectivity matrix, and also resolve any L3 SVI to attach them to proper L2 interface. My goal is to have a table for Source Dev, Source Int, Source MTU, Destination  Dev, Destination Int, Destination MTU, Protocol for all STP, CDP, OSPF, EIGRP etc.   Do you use scripts? Is it possible with Prime? Thanks ... View more

I'm CCIE R&S 23372 who got together with a double R&S/SP 19975, to try to address a problem I continuously had when building and managing dozens of F500 scale networks. As you well know, the way those management systems run for 50 devices is not the way they run for 50000. I was yet to witness a system that could actually understand what I am doing, and free me from this drudgery of entering changes into this or that system, writing this or that document, or writing this or that script to do a straight-forward analysis like "Show me all the hosts for this particular vlan, and show their edge switch and their active gateway" or "Which routes have changed and where?". I was dreaming of a moment when spanning tree topology mapping would become instantaneous, when I could query the whole network for current error rates on all interfaces, and get results instantaneously, not in two weeks time. When I could see the IP network for what it truly is. Routing topologies. Bridging topologies. Protocol topologies. When SDN is just tunelling. But that seemed so far away when even adding a single device requires a human to actually manually clicking that device into CMDB. So we built a thing, NIMPEE, which you can carry around on your laptop, and access through your phone. I tried to describe it here: https://www.youtube.com/watch?v=kYnNPlVuj3I I was really curious of what you guys think. We ran it in 5000-device inter-as mpls vpn firewalled virtual networks that we've built ourselves on UNL (we had to rewrite part of the wrapper for our needs), running in a rack in a garage on about 100 physical cores of x5690 and 600 GRAM of commodity x86 (X3550 M3 is under $200, and so are X5675), with openstack and our own automation and scaling system, which enables to spawn customers or carries with a single command. Basically we had to write our own ISP scripts, adjusted to specific virtual infrastructure and double-nested virtualization, combined with NETNS namespaces to emulate true hosts. We then piloted in production networks of enterprises and an ISP. Looks good so far, but needs more work, and we're hiring Node.js/Arango, math with graph theory, UI/UX, basically all developers that would help us build something that creates documents for you, submits them to the CAB board automatically, and let's network engeneers be network engeneers. You know, the true R&S way, laser focused on the Internet Protocol - the true foundation of all digital communication. The thin waist. Just like Deering imagined it. Looking forward to the day I can get back to managing networks, and not being a slave to them! ... View more