Performance Analysis of Non-cryptographic Hash Functions for Real-Time Storage and Lookup of URLs.

In this work, the performance of various non-cryptographic hash functions has been investigated to identify their random nature when employed in the implementation of hash tables for real-time storage and lookup of uniform resource locators. The performance analysis is performed mainly using statistical studies on the sequences generated using five widely used non-cryptographic hash functions: 1) CRC, 2) Adler, 3) FNV, 4) DJBX33A, and 5) Murmur. The comparative analysis of tested non-cryptographic hash functions shows that the Adler hash function is not suitable for hash table implementation, whereas, the rest of non-cryptographic hash functions exhibit similar and better randomizing features which make them an attractive choice for hash table implementation. The results of these statistical studies have been verified by the implementation of hash table using these non-cryptographic hash functions. The implementation results show that the average number of probes for Adler based hash table varies between 1.25 and 2.75 for different load factors and hash table sizes, whereas, for the rest of non-cryptographic hash functions the average number of probes in a hash table is sim1, which is highly desirable for real-time network applications. Thus proving that 1) CRC, 2) FNV, 3) DJBX33A, and 4) Murmur non-cryptographic hash functions are good choices for hash table based implementation for real-time storage and lookup of uniform resource locators.