The memory footprint for each client session is really huge right now. The main benefit is an increase of performance due to not having to allocate any tracking packet for QoS > 0 by allocating a 65536 array to track all possible concurrent inflight packets. A midway between having to malloc/free each new received packet that must be tracked till the transaction is completed and the current pre-allocate all solution would be great.
The memory footprint for each client session is really huge right now. The main benefit is an increase of performance due to not having to allocate any tracking packet for QoS > 0 by allocating a 65536 array to track all possible concurrent inflight packets. A midway between having to
malloc/freeeach new received packet that must be tracked till the transaction is completed and the current pre-allocate all solution would be great.