Abstract: A small percentage of high rate large-sized flows consume most of the network bandwidth. It is of great significance to efficiently identify these flows for traffic engineering, so as to alleviate network congestion and improve network transport performance. With the development of network technology, the capacity of transmission links and the transfer rate of data flows become higher and higher. So the network equipment with high-speed packets forwarding capability put forward high performance requirements for flows identifying algorithms. The flows whose size and transmission rate both exceed certain thresholds are usually defined as elephant flows. In this paper, a novel algorithm is proposed for elephant flow detection, in which the data structure of flow entries is indexed by d-Left Hash table to meet the performance requirements of high speed packet processing. The proposed detection algorithm combines the d-Left Hash data structure with the eviction of low-rate flows’ entries in order to identify elephant flows efficiently. The accuracy of the proposed detection algorithm is improved by the eviction of low rate flows’ entries. Theoretical analysis is conducted to demonstrate the accuracy, performance and memory overhead of the proposed detection algorithm. Experimental results on real data sets show that the proposed algorithm outperforms CSS, WCSS, S-LRU and L-LRU algorithms in terms of accuracy and performance at comparable memory overhead.