The Zephyr OS is for small and embedded devices which have limited memory and processing power but have IP network connectivity.
Typically there is no dynamic memory in these devices and all the memory must be allocated statically beforehand. This makes it difficult to utilize memory best possible way when network packets are sent or received. Zephyr OS IP stack handles this problem by allocating memory pool of small network buffers optimized for used network bearer type. The network buffers are then linked together to store bigger network packets.
In order to get best network data handling performance, it is possible in certain cases to use zero copy when transferring or receiving data from the network. This is done by pre-allocating space for the link layer headers and partitioning the data so that there is no need to copy network data between different parts of the Zephyr OS.
