Spacetime数据库在其最新技术文章中详细阐述了实现数据库水平扩展的设计方案1。该数据库采用单线程执行模型，经测试在单线程性能上优于并行执行方案1。在热键场景下，单线程解决方案的吞吐量比分布式系统快300倍，达到300,000 TPS对比1,000 TPS1。
Spacetime通过actor模型启发的架构支持跨数据库分片，使事务在单个分片内保持快速、本地和单线程的特性1。SpacetimeDB Cloud支持分布式状态机复制，基准测试显示复制数据库吞吐量约为300k TPS1。文章对CockroachDB等通用水平扩展关系型数据库进行了分析，指出它们在高竞争交易下存在巨大协调开销1。
根据产品规划，Spacetime计划在2026年10月31日发布分层存储和异步数据库间通信功能（Spacetime Continuum），以进一步提升系统的扩展性1。
Clockwork Labs has detailed its approach to horizontal scaling for Spacetime, a database system designed to handle performance challenges that plague traditional distributed systems.1 The company's strategy centers on a single-threaded execution model that, according to benchmarks, outperforms parallel execution in certain scenarios.1
In hot-key workloads—situations where many transactions compete for access to the same data—Spacetime's single-threaded approach achieves approximately 300,000 transactions per second, compared to around 1,000 tps for distributed systems, representing a 300-fold performance advantage.1 This design reflects a deliberate trade-off; while systems like CockroachDB enable general-purpose horizontal scaling across multiple nodes, they incur substantial coordination overhead under high-contention transaction loads.1 Spacetime addresses scalability across three dimensions—computation, storage, and networking—through an actor-inspired architecture that supports cross-database sharding while maintaining fast, local, and single-threaded transactions within individual shards.1 SpacetimeDB Cloud employs distributed state machine replication, with benchmark results showing replicated databases achieving approximately 300,000 tps.1
Looking ahead, Spacetime plans to release tiered storage and asynchronous inter-database communication capabilities on October 31, 2026, features intended to further enhance the system's scalability profile.1
评论
还没有评论，欢迎留下第一条。