After a few months of silence, I’m back guys!
I came across something super interesting last week, which I have been hearing about for a while, but never had the chance to investigate.
Everyone: assume you are sitting in the NOC, on a night-shift, and a failure occurs. I.e. packet loss from endpoint A to B.
What tools do you use to troubleshoot the issue?
Ping? Traceroute? Some SNMP via your OSS/BSS platforms?
Probably you have a good and comprehensive diagram of the E2E service (although I seriously doubt), so you have to log in onto all the intermediate systems to check for packet loss, port speed negotiation, queue drops, etc.
Really? 21st century? This is what we were all using 20 years ago, so I think the times are really ready for a change.
I came across this VERY interesting startup: Forward Networks.
They are a bunch of Stanford CS PhDs coming from (guess what??) Nick McKeown group who put together a platform that is – essentially – a search engine for networks.
Let me explain.
The breakthrough of search engines in the early 2000s, was that, instead of us having to browse through a complicated and often incomplete directory tree, there was a “tool” who built a “model” of the resources he was able to crawl and then return results to our queries to give us a list of results that we are likely to be searching for.
Can you imagine that applied onto networking?
You “connect” all your devices, your entire network, to this magic tin that does analyses at constant intervals (in the order of minutes) and creates a “model” of your network, including ACLs, ports, LLDP adjacencies, ECMP paths, RIBs/FIBs, etc.
Using such a tool you could reconstruct very easily the path of a packet in your network, including which ACLs are applied to the flow, which ports the flows go through, which ECMP paths, etc.
A demo works more than a 1000 words though…