Scalable distributed compact trie hashing (CTH*)

Abstract

This last decade, a new class of data structures named Scalable Distributed Data Structures (SDDSs), is appeared completely dedicated to a distributed environment. This type of data structures opened an important axis of research, considering that the data management in a transparent manner is fundamental in a computer network. All the existing methods are mainly based on Linear hashing (LH*) and Range-partitioning (RP*). In this paper, we propose a new method with the constraints of the SDDS. Our approach is an adaptation of the well-known method Trie hashing (TH) for a distributed environment, i.e. a network of interconnected computers. The latter uses a digital tree (trie) as access function. Our major objective is the distribution of file buckets and the tree representing the hashing function. We have considered TH with the tree represented in compact form (CTH) because this option is probably more interesting for the reduction of the message size circulating on the network. Contrary to the majority of the existing methods, the proposed one provides the order of distributed files, then facilitates both the range query operations and the ordered traversal of files. Moreover, the following properties make our method a promising opening towards a new class of SDDS: (a) preservation of the order of records, (b) works without multicast (c) three bytes are sufficient to address a server, (d) the transfer of some bytes is enough for the update of the client trees. The access performances should exceed the ones of traditional files and some competitive scalable and distributed data structures.