I am evaluating databases for my new project where existing data is stored in cosmos SQL database.
For our use case graphDB seem to be a good solution. My Options are Gremlin API or TigerGraph.
I heard Gremlin API is built on top of document database so queries will be slower as Graph queries first gets converted to nosql queries, is that true statement? Any pointers here?
The Gremlin query language is part of Apache TinkerPop. TinkerPop does not dictate how a back end store and query engine be built. It ships with a reference in-memory graph written in Java that actually uses a simple HashMap to store data. TinkerPop has been ported to many different back end stores and storage models. It is not very common however for that to be a document store and there is no need to convert Gremlin queries to SQL unless perhaps you port TinkerPop on top of a Relational or other store and implement it that way. Most Graph DBs I have worked with that implement support for Apache TinkerPop use custom built graph engines and have their own query optimizers that are nothing to do with SQL. I should add that I am not familiar with how the CosmosDB Gremlin support is implemented. The main point is that TinkerPop does not dictate the type of store that be used or how the Gremlin support is implemented.