Graph library recommendations for large graphs

[VanL]

I am working on a project that will require building and querying large 
graph objects (initially 8M nodes, 30-40M edges; eventually 40M nodes, 
100M edges). NetworkX seems to be the most popular, but I am concerned 
that a dict representation for nodes would use too much memory -- my 
initial tests suggest that a graph with 3M nodes and 12M edges creates 
substantial memory pressure on my machine.

Can anybody who has worked with large graphs before give a recommendation?
Thanks,

Van

--
http://mail.python.org/mailman/listinfo/python-list