Thousands of graphs for multi-tenant scenarios

[luxianlong]

Hi, all

Scenario: I've set up a JanusGraph to support multi-tenant scenario. The tenant number is a big number which will be more than 10000. I create one graph in the cluster for each tenant.

Issue: Then the question is that there will be thousands message pullers of "systemlog" (each graph needs one message puller to fetch system logs per 5 seconds). So many message pullers will impact the performance of each JanusGraph instance.

It seems to be a scalability issue to support many graphs in one cluster. Are there any ideas/proposals to support such case?

Thanks in advance.