algorithm - good graph/complex networks libraries -

i looking recommendation graph analysis libraries or framework，better in c++ or java. have found graph libraries，

https://stackoverflow.com/questions/3010805/scalable-parallel-large-graph-analysis-library

this webpage gives possible solutions：

• c++ -- viable solutions appear boost graph library , parallel boost graph library. looking @ mtgl, although slanted more toward massively multithreaded hardware architectures cray xmt. lastly, i've added lemon list consideration. • c - igraph , snap (small-world network analysis , partitioning); latter uses openmp parallelism on smp systems. • java - have found no parallel libraries here yet, jgrapht , perhaps jung leading contenders in non-parallel space. • python - igraph , networkx solid options, though neither parallel. there used python bindings bgl, these unsupported; last release in 2005 looks stale now.

but not sure 1 should use based on own need：

1. good data structure , algorithm. can analysis properties of complex networks.
2. scalable： can modify , implement algorithm on it
3. can analysis of graphs/networks million nodes , edges
4. complex networks can produced better, save time true complex networks.

analysis capabilities , scalability primary concern. have recommendations. suggestion helpful.

i have used networkx , works jung. if data in memory these 2 work great problem have both of them lack of persistence(you can save both xml file , upload/download database no in-database analytics limits size of graph size of memory).

lately have been playing around neo4j(and various add ons), gets around this. doesn't scale unlimited if data less 32 billion nodes fast storage engine , can use other cores(or worker machines) analysis graph in parallel. there's alot of performance docs on how fast don't agree type of tests because not test case, suggest download it(free on neo4j.org) , judge yourself.