examples of sqlg in the wild

[rhubarb]

Are there some examples of sqlg in the wild?
Does it generally run behind gremlin-server or in some other configuration?
I'm thinking of using it for a project with the possiblity of swapping out for neptune or neo4j in the future, so relying on tinkerpop to abstract my dependencies... does that seem reasonable?

[pietermartin]

Hi,

I recommend running Sqlg embedded in your server. On postgresql Sqlg supports a distributed mode where you can run it embedded in multiple jvms. It uses postgres's notify mechanism to keep its meta data synchronised.

Most of Sqlgs custom features are not available via gremlin-server. The schema creation api is not available via gremlin-server, so no indexes / partitions / custom identifiers... The various bulk modes are also not possible via gremlin-server.

gremlin-server is only required if you are planning on using a non jvm language.

The production systems I run Sqlg on, are very large and its working well. Sqlg manages databases with 100s of schemas and thousands of tables each with 100s of columns. Last I checked the total amount of tables was about 50 000 and total columns over 1 000 000. The db sizes are generally around 5TB or so.

However if you plan to swap out later for another provider then using any of the custom features are risky. They all have their own custom features also, so not sure if its possible to avoid.

[rhubarb]

Thanks for the quick answer.
The main reasons I'm interested in gremlin-server are

1. because I want to put a REST API in front of my graph database for querying... Is there another way of doing that?
2. I figured that would help with the swap-it-out-for-neo4j-later options... but perhaps if I stick to gremlin, using the server doesn't make a difference there?

On the original question of 'out in the wild' can you point to any uses of this in other products or applications on GitHub or elsewhere?

thanks again

[pietermartin]

"REST API in front of my graph database for querying".
In this case yes gremlin-server will be best.

I do not design applications like that, nor is it the intended use of Sqlg. I view Sqlg and gremlin equivalent to JDBC. No ones wraps JDBC inside a rest server. Ideally REST calls are at the application level and at a much higher level of abstraction.

That said Sqlg implements Tinkerpop exactly so there is no problem using gremlin-server.
"but perhaps if I stick to gremlin, using the server doesn't make a difference there?" yes,

Regarding out in the wild there was a US medical company using it in production. I have not heard from them for some time so don't know what is happening there.
Apart from stars on github I have no further insight to Sqlg out there in the wild.