Publicação
Byzantine Fault Tolerance in Large Scale Edge Replicated Systems
| Resumo: | Storage systems play an essential role in supporting cloud applications by providing scalable, reliable, and efficient data management solutions. All applications using cloud infrastructures incur the latency of contacting the provider’s data center. This incurred latency may be prohibitive for several application scenarios, notably, real time systems. The advent of edge computing aims to bridge this gap by bringing computations, and hence data manipulated by them, closer to the end users. However, the susceptibility of edge machines to malicious parties poses a significant obstacle. Unlike data centers, edge nodes lack a robust security perimeter, making them vulnerable to be compromised, impacting the integrity and functionality of applications residing at the edge. In this work, we addressed these challenges by specifying and implementing an edge- driven storage system resilient to byzantine faults, considering the security concerns prevalent in the edge environment. Most existing works in this domain either neglect the unique challenges of edge computing, fail to consider byzantine security issues, or lack comprehensive and suitable replication and consistency guarantees. Our solution allows for peers in the edge to communicate efficiently and in a decentralized manner, correctly distributing operations that are applied to the data of applications across the replicas, and is able to support various data structures for many use cases. It employs a secure and partitioned overlay network solution, a replication protocol that enforces byzantine fault-tolerant causal+ consistency and partial replication, and a schema for CRDT creation in this setting. We validated the system through experimental evaluation in an emulated edge sce- nario and comparing it with a state-of-the-art edge-driven storage system. The results demonstrate that our solution can effectively operate in edge environments while main- taining robustness against failures, and with acceptable performance for the security properties it provides. |
|---|---|
| Autores principais: | Almeida, Diogo António Rocha |
| Assunto: | distributed storage systems decentralized systems edge computing byzantine fault tolerance causal+ consistency partial replication |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Storage systems play an essential role in supporting cloud applications by providing scalable, reliable, and efficient data management solutions. All applications using cloud infrastructures incur the latency of contacting the provider’s data center. This incurred latency may be prohibitive for several application scenarios, notably, real time systems. The advent of edge computing aims to bridge this gap by bringing computations, and hence data manipulated by them, closer to the end users. However, the susceptibility of edge machines to malicious parties poses a significant obstacle. Unlike data centers, edge nodes lack a robust security perimeter, making them vulnerable to be compromised, impacting the integrity and functionality of applications residing at the edge. In this work, we addressed these challenges by specifying and implementing an edge- driven storage system resilient to byzantine faults, considering the security concerns prevalent in the edge environment. Most existing works in this domain either neglect the unique challenges of edge computing, fail to consider byzantine security issues, or lack comprehensive and suitable replication and consistency guarantees. Our solution allows for peers in the edge to communicate efficiently and in a decentralized manner, correctly distributing operations that are applied to the data of applications across the replicas, and is able to support various data structures for many use cases. It employs a secure and partitioned overlay network solution, a replication protocol that enforces byzantine fault-tolerant causal+ consistency and partial replication, and a schema for CRDT creation in this setting. We validated the system through experimental evaluation in an emulated edge sce- nario and comparing it with a state-of-the-art edge-driven storage system. The results demonstrate that our solution can effectively operate in edge environments while main- taining robustness against failures, and with acceptable performance for the security properties it provides. |
|---|