Deltas and Dynamic Pojos

Question

I have 3 questions :

1.Delta operation in Ignite is basically used to avoid overhead of resending the whole object , but while performing a Delta on a pojo the Binary object is built again. Is nt it a Cliche ? I agree that many other overheads are reduced.

2.If we consider a client server mode with Replication, How are the Deltas transferred over the replicas ?

3.While trying to access a cache using SQL queries , what can be done to implement Dynamic POJO s ?

Answer 1

1. The main advantage here is that you avoid transferring large data objects across the network - typically this is the biggest overhead. Also note that binary object builder is highly optimized, so that when you update one field it will update only this field and cheaply create the copy for everything else in the array . This is not the same as fully deserialize the object and serialize it back.
2. In ATOMIC cache primary node will send the new value to backups, but usually it's not an issue because backups are updated asynchronously by default. 'TRANSACTIONAL' cache always sends entry processor and never sends the value.
3. Dynamic SQL schema is not supported right now, but is on the roadmap.