CloudJump II: Optimizing Cloud Databases for Shared Storage

The CloudJump framework aims to shed light on how compute-storage disaggregation can effectively enhance cloud databases' scalability, resilience, and efficiency by only relying on standard cloud storage services. To explore the evolving landscape of cloud-native databases, this study navigates through the design and implementation intricacies of the shared-storage architecture in the CloudJump framework. By focusing on the shared storage model integral to cloud database systems with a leader-follower pattern, we address the challenges of ensuring data consistency across multiple compute nodes-a cornerstone for achieving scalability in cloud environments. At its core, this study proposes Multi-Version Data (MVD), a technique designed to improve node consistency in shared storage environments. MVD enables fine-grained version management and precise reconstruction of data pages by compute nodes, facilitating streamlined recovery and robust data persistence without requiring a custom storage layer. This design simplifies operations, reduces costs, and presents a pragmatic response to the limitations of existing cloud storage approaches. The deployment of MVD in PolarDB demonstrates its practical benefits, including enhanced node decoupling and accelerated failover. The performance and reliability of the enhanced system, CloudJump II, are substantiated through comprehensive experimental evaluation and production usage.