Hello, We are facing some scalability and latency issues with Apache Atlas. Today we have Atlas installed as single node. Some of simple get entity calls (with lots of relationships) are taking lot of time. In our data, we have lot of deep relationships.
Current state: 1. Atlas + JanusGraph (same jvm) Cassandra cluster for persistence ES cluster for indices We are thinking of scaling Atlas Horizontally to reduce latency and improve overall performance. We are thinking of below two options. Option 1 1. Single Atlas instance Multiple instances of JanusGraph servers, living along with Cassandra. Atlas will call JanusGraph though rest API or web sockets. Option 2 Multiple instances of Atlas + JanusGraph (single jvm) Cassandra cluster for persistance What we are trying to achieve horizontal scale with minimal changes to Atlas code and at the same time do not want to have any transaction violations. With graph transactions in mind which option is suggested?. Thanks, Sai.