OPAL regarding perf/avail, limitations, recommended data

[hongbo-miao]

I am wondering if adding OPAL to this OPA External Data table, how would it be regarding perf/avail, limitations, recommended data? Thanks

[asafc]

Hi @hongbo-miao, this table is from the OPA project site, and lists all the built-in ways to sync data into OPA. I’m not sure if the OPA maintainers would like a reference to an external project to be part of their table, but if you simply want to know how OPAL compares to these approaches, here how i would think about it:

• Perf/Avail: OPAL is highly performant/available - because you can scale both OPAL server and client and they are both stateless + OPAL updates data with a back channel so it does not affect OPA’s authorization query evaluation performance:
• Limitations: only limitations are OPA limitations with size. If i am not mistaken OPA can hold up to 5GB of data in one agent before starting to degrade. However, OPAL can mitigate this issue by enabling you to shard large datasets into multiple OPA agents.
• Recommended Data: both static and dynamic are well supported. OPAL can push data into OPA from multiple sources (aggregation of state) and do it in realtime (realtime updates of app critical data). The reason we built OPAL is to be able to use OPA for user facing applications, which often aggregates state from multiple sources (e.g: billing data from stripe - who is a paying user) and needs to sync in realtime (i.e: just invited a user to a document in realtime - polling is not fast enough for these use cases).