As the failure of a critical node will directly partition a network

a theorem for critical node identification was proved

which indicates that node degree Ni and elementary loop degree Mi of node i are two decisive factors for the exis-tence of a critical node and shows that Ni-Mi≥ 2 is the necessary and sufficient condition for node i being critical.Based on the theorem

a distributed topology partition detection algorithm（distributed partition detection protocol

DPDP） was presented for large scale networks

which achieves the goal of partition detection efficiently by detecting critical nodes in a local area.Theoretic analysis and experiment results show that DPDP has the advantages of low complexity

high accuracy

low cost as well as good scalability

and is superior to other algorithms.