Scholarlink

<p>Most existing Byzantine fault-tolerant algorithms are slow and not designed for large participant sets trying to reach consensus. Consequently, distributed databases that use consensus mechanisms to process transactions face significant limitations in scalability and throughput. These limitations can be substantially improved using sharding, a technique that partitions a state into multiple shards, each handled in parallel by a subset of the network. Sharding has already been implemented in several data replication systems. While it has demonstrated notable potential for enhancing performance and scalability, current sharding techniques still face critical scalability and security issues. In this article, we propose a novel, fault-tolerant, self-configurable, scalable, secure, decentralized, high-performance distributed NoSQL database architecture. Our approach employs an innovative sharding technique to enable Byzantine fault-tolerant consensus mechanisms in very large-scale networks. We introduce a new sharding method for data replication that leverages a classic consensus mechanism, such as PBFT, to process transactions. Our approach modifies node allocation among shards through the public key generation process, effectively reducing the frequency of cross-shard transactions, which are generally more complex and costly than intra-shard transactions. Additionally, our method eliminates the need for a shared ledger between shards, which typically imposes further scalability and security challenges on the network. We also explain how to automatically form new committees based on the availability of candidate processor nodes. This technique optimizes network capacity by employing inactive surplus processors from one committee’s queue in forming new committees, thereby increasing system throughput and efficiency. Our approach maximizes processor node utilization as well as computational and storage capacity across the network, enhancing both processing and storage sharding to their fullest potential. With our approach, a network using a classic consensus mechanism can scale significantly in the number of nodes while remaining permissionless.We call this novel architecture the Parallel Committees Database, or simply PCDB.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Parallel Committees: a scalable, secure, and fault-tolerant distributed NoSQL database architecture

  • 期刊论文

  • 发表日期：

    2025年2月25日

查看全文
  • Siamak Solat

摘要

Most existing Byzantine fault-tolerant algorithms are slow and not designed for large participant sets trying to reach consensus. Consequently, distributed databases that use consensus mechanisms to process transactions face significant limitations in scalability and throughput. These limitations can be substantially improved using sharding, a technique that partitions a state into multiple shards, each handled in parallel by a subset of the network. Sharding has already been implemented in several data replication systems. While it has demonstrated notable potential for enhancing performance and scalability, current sharding techniques still face critical scalability and security issues. In this article, we propose a novel, fault-tolerant, self-configurable, scalable, secure, decentralized, high-performance distributed NoSQL database architecture. Our approach employs an innovative sharding technique to enable Byzantine fault-tolerant consensus mechanisms in very large-scale networks. We introduce a new sharding method for data replication that leverages a classic consensus mechanism, such as PBFT, to process transactions. Our approach modifies node allocation among shards through the public key generation process, effectively reducing the frequency of cross-shard transactions, which are generally more complex and costly than intra-shard transactions. Additionally, our method eliminates the need for a shared ledger between shards, which typically imposes further scalability and security challenges on the network. We also explain how to automatically form new committees based on the availability of candidate processor nodes. This technique optimizes network capacity by employing inactive surplus processors from one committee’s queue in forming new committees, thereby increasing system throughput and efficiency. Our approach maximizes processor node utilization as well as computational and storage capacity across the network, enhancing both processing and storage sharding to their fullest potential. With our approach, a network using a classic consensus mechanism can scale significantly in the number of nodes while remaining permissionless.We call this novel architecture the Parallel Committees Database, or simply PCDB.