Large number of acknowledgements from Redis causing high CPU usage in firewall

We have a National Level Fraud and Risk Management application , which handles all Indian Inter-bank payments real-time(ATM, UPI, IMPS, Cards, Aadhaar based etc.). We use Redis extensively here to store transactions and aggregates. For inter-site(Chennai-Hyderabad DC) data replication we use direct posting to the other site using Jedis. The posting goes via network firewall. The payload for message posting works fine, but the acknowledgements from Redis(which are very small in size) are choking the firewall. As per CISCO(firewall provider), firewalls get choked by small packets, as it treats them as hacking/attack.

We work at 3000 Transactions per second(TPS). We have 150 Redis instances where data are sharded and stored. Since for each transaction, we have around 10 different types of data, the TPS for posting to redis is 30000. This works perfectly within a data-center, but across data-center, firewall gets choked due to the small acknowledgements from the remote Redis instances.

I shall be grateful any of you can tell me whether a) The acknowledgement size can be configured in Redis, b) Whether any body has already solved such issue at firewall due to large number of small(<200 bytes) acknowledgements from Redis and what can be the solution.

This doesn’t sound like a Redis problem. It sounds like a firewall problem.

@chinmoyd I am sure you must have resolved the firewall issue so far. (as its 2yrs a post), wanted to check if you have reviewed the Redis Enterprise Active-Active feature, it can help two-way data replication as well as conflict resolution in real-time. let me know if you want to discuss it.

Refer:Active-Active Geo-Distribution (CRDTs-Based) | Redis Enterprise