Has anybody here ever played with a graph database (NoSQL lols in this thread)

http://www.slideshare.net/emileifrem/neo4j-the-benefits-of-graph-databases-oscon-2009

So I'm probably going to pitch a couple of dudes a ridiculous idea this week and figured why not ask ilcomputers while I'm at it:

- starting with neo4j 'cause it has a wiki already and every other graph DB seems to be strictly commercial/proprietary

- want to begin with 4 billion nodes (one for every IPv4 address)

- relate those nodes to "label" nodes, to form lists (whitelists, blacklists, country codes, ASNs, etc); millions of label nodes possible

- relate those lists to "source" nodes (users, cron jobs, websites) and the sources into source families

- tricky part: relate multiple nodes to "timestamp" nodes, so that a unix time can be associated with an IP' nodes relationship to a label node, e.g. say we have a label node that is just a domain name, and a certain list of addressed resolved from that domain at x time, but did not as of y time (this can be done with a traverser. I am pretty sure.)

- further along down the road, additional relationships for other attributes describing the various nodes; the idea ultimately being I can do almost all of my homework on a network device using a couple of traversals instead of essentially just grepping billions of lines of binary data

So aside from a few presentations and anecdotes from people I definitely don't know, has anybody here worked with Neo4j, or something like it? What eventually kills a graph DB? What do I need to watch out for?

also is it even worth pursuing hadoop & cascading if I only roll with about 50 gigabytes of binfiles a day

hi

hey how you been

o good you know ate some noodle earlier - yrself

busy

cool - i dont really know anything abt this particular technology but i think its sweet u want to have four billion nodes

the od(bms) couple

keeping this b/w us, if you can figure out a way to turn nodes into dollars....oh man

xp

what up tom

So I'm probably going to pitch a couple of nodes a ridiculous idea this week and figured why not ask ilcomputers while I'm at it:

- starting with neo4j 'cause it has a wiki already and every other graph DB seems to be strictly commercial/proprietary

- want to begin with 4 billion dudes (one for every IPv4 address)

- relate those dudes to "label" dudes, to form lists (whitelists, blacklists, country codes, ASNs, etc); millions of label dudes possible

- relate those lists to "source" dudes (users, cron jobs, websites) and the sources into source families

- tricky part: relate multiple dudes to "timestamp" dudes, so that a unix time can be associated with an IP' dudes relationship to a label dude, e.g. say we have a label dude that is just a domain name, and a certain list of addresses resolved from that domain at x time, but did not as of y time (this can be done with a traverser. I am pretty sure.)

- further along down the road, additional relationships for other attributes describing the various dudes; the idea ultimately being I can do almost all of my homework on a network device using a couple of traversals instead of essentially just grepping billions of lines of binary data

totally ignorant in this particular matter, sorry dude.

lol wats a node

I have never played with a graph database but what you're describing is making me want to play with one.

good luck tom

oh and btw I'm all up in yr HOOD next week

marko rodriguez's deck about graph DBs seems like a good overview (to the eye of this particular dangerous amateur) - I'm gonna holler at our grant mofos to see if there's existing research in this area that we can piggyback on

anybody know anything about netcharts? fucking netcharts, man.