The can widely vary on your nature of data, peak spikes and other requirements.
You can start with allocating more resources to clickhouse and otel-collectors. Rest of the components will not be taking up too much resources.
And update the allotted resource as you see.
For very high production workload, it is recommend have message queuing tool like Kafka/Redpanda or any other compatiable tools.
This way, you can handle random spikes better and have a reliable set up with no data loss even when SigNoz clickhouse/collectors are down.