Towards an efficient indexing and searching model for service discovery in a decentralised environment.

Given the growth and outreach of new information, communication, computing and electronic technologies in various dimensions, the amount of data has explosively increased in the recent years. Centralised systems suffer some limitations to dealing with this issue due to all data is stored in central data centres. Thus, decentralised systems are getting more attention and increasing in popularity. Moreover, efficient service discovery mechanisms have naturally become an essential component in both large-scale and small-scale decentralised systems and. This research study is aimed at modelling a novel efficient indexing and searching model for service discovery in decentralised environments comprising numerous repositories with massive stored services. The main contributions of this research study can be summarised in three components: a novel distributed multilevel indexing model, an optimised searching algorithm and a new simulation environment. Indexing model has been widely used for efficient service discovery. For instance; the inverted index is one of the popular indexing models used for service retrieval in consistent repositories. However, redundancies are inevitable in the inverted index which is significantly time-consuming in the service discovery and retrieval process. This theeis proposes a novel distributed multilevel indexing model (DM-index), which offers an efficient solution for service discovery and retrieval in distributed service repositories comprising massive stored services. The architecture of the proposed indexing model encompasses four hierarchical levels to eliminate redundancy information in service repositories, to narrow the searching space and to reduce the number of traversed services whilst discovering services. Distributed Hash Tables have been widely used to provide data lookup services with logarithmic message costs which only require maintenance of limited amounts of routing states. This thesis develops an optimised searching algorithm, named Double-layer No-redundancy Enhanced Bi-direction Chord (DNEB-Chord), to handle retrieval requests in distributed destination repositories efficiently. This DNEB-Chord algorithm achieves faster routing performances with the double-layer routing mechanism and optimal routing index. The efficiency of the developed indexing and searching model is evaluated through theoretical analysis and experimental evaluation in a newly developed simulation environment, named Distributed Multilevel Bi-direction Simulator (DMBSim), which can be used as cost efficient tool for exploring various service configurations, user retrieval requirements and other parameter settings. Both the theoretical validation and experimental evaluations demonstrate that the service discovery efficiency of the DM-index outperforms the sequential index and inverted index configurations. Furthermore, the experimental evaluation results demostrate that the DNEB-Chord algorithm performs better than the Chord in terms of reducing the incurred hop counts. Finally, simulation results demonstrate that the proposed indexing and searching model can achieve better service discovery performances in large-scale decentralised environments comprising numerous repositories with massive stored services.

Given the growth and outreach of new information, communication, computing and electronic technologies in various dimensions, the amount of data has explosively increased in the recent years. Centralised systems suffer some limitations to dealing with this issue due to all data is stored in central data centres. Thus, decentralised systems are getting more attention and increasing in popularity. Moreover, efficient service discovery mechanisms have naturally become an essential component in both large-scale and small-scale decentralised systems and. This research study is aimed at modelling a novel efficient indexing and searching model for service discovery in decentralised environments comprising numerous repositories with massive stored services. The main contributions of this research study can be summarised in three components: a novel distributed multilevel indexing model, an optimised searching algorithm and a new simulation environment. Indexing model has been widely used for efficient service discovery. For instance; the inverted index is one of the popular indexing models used for service retrieval in consistent repositories. However, redundancies are inevitable in the inverted index which is significantly time-consuming in the service discovery and retrieval process. This theeis proposes a novel distributed multilevel indexing model (DM-index), which offers an efficient

solution for service discovery and retrieval in distributed service repositories comprising massive stored services. The architecture of the proposed indexing model encompasses four hierarchical levels to eliminate redundancy information in service repositories, to narrow the searching space and to reduce the number of traversed services whilst discovering services. Distributed Hash Tables have been widely used to provide data lookup services with logarithmic message costs which only require maintenance of limited amounts of routing states. This thesis develops an optimised searching algorithm, named Double-layer No-redundancy Enhanced Bi-direction Chord (DNEB-Chord), to handle retrieval requests in distributed destination repositories efficiently. This DNEB-Chord algorithm achieves faster routing performances with the double-layer routing mechanism and optimal routing index. The efficiency of the developed indexing and searching model is evaluated through theoretical analysis and experimental evaluation in a newly developed simulation environment, named Distributed Multilevel Bi-direction Simulator (DMBSim), which can be used as cost efficient tool for exploring various service configurations, user retrieval requirements and other parameter settings. Both the theoretical validation and experimental evaluations demonstrate that the service discovery efficiency of the DM-index outperforms the sequential index and inverted index configurations. Furthermore, the experimental evaluation results demostrate that the DNEB-Chord algorithm performs better than the Chord in terms of reducing the incurred hop counts. Finally, simulation results demonstrate that the proposed indexing and searching model can achieve better service discovery performances in large-scale decentralised environments comprising numerous repositories with massive stored services.