The present invention discloses a method, an apparatus, and a system for implementing packet loss detection. The method includes: determining, by a detection apparatus, that there is no packet loss when packet transmission is performed at a first priority over a first loopback path that includes a first subpath and a second subpath; sending N1 packets over a second loopback path that includes the first subpath, the second subpath, a third subpath, and a fourth subpath, where the packets are transmitted at the first priority over the first subpath and the second subpath, and are transmitted at a second priority over the third subpath and the fourth subpath; and determining a packet loss rate between a first network element and a second network element according to N1 and a quantity N2 of received packets returned over the second loopback path. Therefore, a packet loss rate between any two network elements can be detected only by implementing a detection function on a detection apparatus, thereby resolving a prior-art problem of complex implementation and relatively high deployment costs because the detection function is required on each detected network element.