How about the throughput under Multi-Tenancy mode?

What I want to do

i want to know how about the throughput under Multi-Tenancy mode?
do you have some benchmarks?


Nope, too early to do this. I think the core team have done some bench, but not sure if it was related to MT. But there’s no much difference. MT is just a logical separation. It is like having several predicates spread over the cluster. You could potentially have a MT logic in a normal cluster, but you would have a ton of logic code to deal with.

but will the throughput be increase after increasing alphas (groups) beacuse of more decentralized predicates ?

In theory, yes. If you have a horizontal scaling it will increase the throughput. Also using prefixes in the predicates can also make the querying faster. e.g. Instead of having a “name” predicate. Add a prefix for each context. “”, “”, “”, “” and so on.

A better Graph modeling(planning) always makes things smooth.

“user” “address” is type?
do you have some best practice docs?

No exactly, but could be a way to go. The GraphQL model uses the prefix based on the GraphQL Type. But you can have any modeling you want.

About modeling, I don’t think so. Do we have @damian / @acarey ?

No, but we have a tutorial about Graph data models available on our Dgraph Learn website:

1 Like