Named Data Networking (NDN) is a new content-centric architecture.Its original Leave Copy Everywhere (LCE) strategy has many shortcomings

for instance

massive cache redundancy and poor utilization of neighbors' cache

which waste cache space and lower cache efficiency.We proposed a routing scheme named K-Medoids Hash Routing (KMHR) combining on-path non-cooperation and off-path cooperation schemes.K-medoids algorithm is used to select some content routers as medoids in hierarchical clusters.And for different popularity of contents

we choose Hash routing or the shortest path routing separately.KMHR locates and insures the uniqueness of contents with high popularity in the cluster

which significantly reduces cache redundancy and improves cache efficiency.Based on the real world network topology

simulation results show that KMHR has the shortest request time

optimal routing gain and fewer cached contents.