A novel algorithm which is based on both the flows′ size and transmission rate was proposed for elephant flow detection. In order to meet the performance requirement of high speed networks, the data structure of elephant flow detection was indexed by Hash table, which combines the Hash conflict resolution with the eviction of cached flow entry to identify the elephant flow efficiently. A theoretical analysis was conducted to demonstrate the accuracy, performance and memory overhead of the detection algorithm. Experimental results on real data sets show that the proposed algorithm outperforms least recently used detection algorithm, its derivations and compact space saving detection algorithm in terms of accuracy and performance with similar memory overhead.