Supervisors info:
Μέμα Ρουσσοπούλου, Αναπληρώτρια Καθηγήτρια, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, Ε.Κ.Π.Α.
Summary:
The purpose of this master thesis is the study and the analysis of the novel blockchain technology, as well as a proposal of a new approach on its implementation. In particular, it considers the idea of organizing the blockchain peers into a DHT in order to store the blockchain data: As blockchains get older, their size gets larger. Currently, Bitcoin blockchain is over than 190GB. In addition, as long as some of the ideas about the increase of the transaction rate put into practice, blockchain size will grow even faster. The ever-growing size of the blockchain it is a factor of making the system more centralized, since too few of the nodes will be able to keep a full replica of the data. After all, archival nodes with open ports, have little or no incentives to waste their storage and their bandwidth, in order to keep all the data and serve it to a bootstrapping node. Thus, it is reasonable to mitigate their load (especially that we call “cold data”, some historical blocks) by splitting it into a DHT. The evaluation of this proposal is achieved through simulations of the first and most popular blockchain network, the Bitcoin: We build a simplified DHT that reminds the Chord. We tune our system and we make a system whose characteristics is similar to Bitcoin. We measure several aspects of our system such as capacity savings per node compared to the amount of data that a full node stores today, load balance, query latency, bandwidth usage. We present our results, we make comments on them, and we draw a conclusion.
Keywords:
Blockchain, Bitcoin, Distributed Hash Table, Chord, Scalability, Storage, Simulation, Markov Chain Monte Carlo