The most significant is that the switch will treat mirrored traffic as a lower priority. While the CPU overhead of copying any individual packet and mirroring it to a port or VLAN is low, those costs add up. So the switch treats each mirrored packet as a lower priority than normal network traffic. During low-to-medium traffic periods, this isn’t a big deal. The switch capably handles both the normal traffic and the mirrored traffic. However, when traffic flow loads up, things can get hairy. The switch will drop mirrored packets first. This means that you’re most likely to lose some network observability during the time when you need it the most.
Another downside to port mirroring is that it requires resources on physical or virtual appliances. These can be costly from a hardware and (in the case of commercial solutions) software licensing point of view. As a result, in most cases, it is only fiscally feasible to deploy forms of port mirroring at selected points in the network.
A cloud-friendly and highly scalable model combines the deployment of lightweight host-based monitoring agents that export packet capture statistics gathered on servers and open source proxy servers.