About performance

[3ks]

Hey guys:

I learned from the documentation that ent supports both SQL and Germlin.

Because of their structural differences, I want to know how their performance is when performing graph traversal.And the applicable scene of ent？

Thanks.

[alexsn]

Hi @gorda, tl;dr is you should go with a SQL DB.

We initially developed Gremlin support as we assumed it would be a better fit for the ORM however after running with AWS Neptune we decided to switch to MySQL for the following reasons:

• No tenancy support: We needed a separate graph per tanant which in MySQL translates to a logical DB.
• No support for transactions.
• Poor performance compared to RDS: We saw a much better performance in MySQL even though you need to do join for edge traversals.
• People are normally familiar with SQL while gremlin query language is alien to most.

Note that we didn't test other gremlin implementations (i.e neo4j, janus, titan ...) as we wanted something managed from AWS.

[3ks]

Hi @alexsn, thank you for your reply, and the last question, are you planning to update the documentation? I noticed that the current documentation contains some of the unique nouns for the graphics database, and now ent is mainly used for SQL DB. It may be a little hard to understand for users who are misunderstood with unfamiliar graphics databases, such as me.

[alexsn]

The use of graph terminology is intentional as ent models data as a graph. The fact that this graph is stored in SQL is implementation details as we expect users to interact with their data via generated code.

[3ks]

It’s time to learn about the graphics database.Thank you!