Is it possible to use persistent actor with router such as to scale out? If you have elasticity with persistent actor you have a problem with your event sourcing.
Hence I wonder how to scale up or out persistent actors for better throughput?
The question is a bit too open to answer directly, though generally there's two things you can do for starters:
For improving a Persistent Actor's throughput there's techniques you can use inside it already, i.e. using persistAsync which batches up updates into one batch sent to the Journal.
PersistentActors can be easily scaled-out by using them together with ClusterSharding which allows to start actors for given ids (and shard ids) on independent nodes - which sales out the load to those.
Of course for tuning raw throughput of message persisting, you definitely must pick a fast serialization format and configure it as explained in the docs on schema evolution.
Lastly, but very importantly, you should pick a Journal implementation which will scale to your needs. Cassandra is a popular choice, however you'll need to run at least 3 nodes for it to make sense I'd say.