“…The works in [ 46 , 49 , 55 , 59 , 64 , 68 , 69 ] report processing latencies of 1.23 μs, 200 μs, 25 μs, 0.458 μs, 6 μs, 5 μs, and 0.739 μs, respectively. The study in [ 56 ] reports a latency reduction of 77% compared to common software implementations, while [ 61 ] shows a drop of 3.75x compared to a host based UPF deployment using DPDK. Furthermore, [ 44 ] specifies an end-to-end latency of 50 μs, which is useful for checking compliance with the KPIs from [ 10 ].…”