Yes. I've enabled S3 checkpointing, I'm dealing with massive data ingestion where windowing isn't feasible. To prevent state explosion due to the real-time changes in our substantial stateful data, I'm considering storing the state in an external source rather than in memory. This approach could potentially facilitate easier scaling of our solution. I also have identified lookup joins but since data is expected in less than 30 seconds. I am not able to use this as well. Could you please help me with any other suggestions to deal with this situation apart from scaling up the memory