HCS: A fast and efficient algorithm to smooth hash collision

As a compact data structure with the capacity of supporting fast lookup, Hash table is widely used in many network applications. The accompanying memory accesses introduced by hash collision make hash table works at a slower speed than required. Under the case of serious collision, these applications have little time to proceed the following packet processing, and fail to respond to new arriving packets. In this paper, we introduce an efficient algorithm named HCS to limit the memory accessing cost in an acceptable range. At the core, it dynamically applies for a new hash table, and adopts different hash functions to rehash these serious collide connections. Furthermore, it will reclaim the applied hash table when necessary. Experimental results show that, comparing with existing algorithms, HCS can decrease the lookup times by 18% at least.