How to increase the performance of the kafka input in indexer

Hello everyone, I have a problem I've been working on for days and I can't find a compatible solution. I need to consume about 150k messages per second via a kafka input / topic / groupid. I have 8 logstash on separate machines and performers of both cpu and memory (I already have about fifteen active pipelines but they are not a problem at the moment). Basically I can consume about 8k per second of messages, using the consumer_thread value of 5 for each node (with topic partition 48). I always arrive at a maximum of about 42k messages per second and the rest goes into lag. I've tried working with the worker and batch.size parameter but they haven't changed the context. I ask if you have tested solutions that can do for me, without increasing other logstash indexers. Thank you all