Cayley

What do you like best?

The swappable backend stores are awesome...already have Postgres in your stack? Just use that. Doesn't need long term persistence? It can operate in-memory. LevelDB is there for max performance, and with Bolt support you can have a 100% in-application Go-only codebase.

It was really easy to get setup, and if your familiar with other graph products you probably already know at least the basics of Gremlin.

What do you dislike?

There aren't as many builtin algorithms for graph traversal as competitors, but you can always write those yourself. It's a triplestore, so it also doesn't support node/edge properties which can be fine or annoying depending on your use case.

What problems are you solving with the product? What benefits have you realized?

We used it to house fairly stable connections that we needed to traverse/lookup quickly but would have taken many expensive/disk io heavy joins on our Postgres instance. The latency of those queries was greatly reduced.