Latency

Hot potato or cold potato? How are the hyperscalers transmitting data (from core to PoP) and how fast are they doing it?