I want graph database for my project. I wish to perform least path operation around the graph. I looked google and located neo4j, InfoGrid and sones. I don't know which works the very best.

Neo4j is insisting on transaction for read queries also, and so i am worried it'll cause delays. I would like high end graph database engine. It will likely be awesome whether it supports least path also.

To be sure with @Reinderien, you need to bench these items. However I would include OrientDB within the mix. Estimating its author in this comment:

I am trying to OrientDB. It is a NoSQL hybrid document-graph dbms with special operators for graph procedures. The primary difference is that you could query it using SQL language with a few extension like:

select from People where friends TRAVERSE(1,7) (name = 'Ayende' and surname = 'Rahien')

What this means is get all of the people who have relationship of buddies with you as much as the seventh degree of separation.

Based on its website, the performances looks pretty decent. Well worth the take a look at least.

I believe you are able to give consideration at Flockdb too.

Fortunately there's a typical interface for many of them: Blueprints. On the top from it you will find lots of technologies like the Gremlin language and Relaxation accessibility graph. Visit Tinkerpop to understand more.

About benchmarks there's a piece-in-progress project known as graphdb-bench (always by Tinkerpop), but AFAIK is within development.

Why don't you try out the 3 and use a benchmark for the particular data? Otherwise it's tough to express which would carry out the quickest.

From version 1.1 read data without needing a transaction in Neo4j. Just in case you need to read uncommitted data though, you are able to only do this from the same transaction (and clearly require a transaction to begin with). A lot of transactions can be found in the Neo4j wiki.