面向多租户数据存储的分段多维性能模型

Piecewise Multiple Performance Model for Multi-Tenant Data Storage

多租户数据存储对提高用户体验质量,优化系统运行效率,降低服务提供商成本至关重要。面对大规模租户,共享数据存储节点如何满足不同租户的不同性能需求是需要解决的关键问题。为此,针对多租户数据存储不同性能需求的特点,提出了一种面向多租户数据存储的考虑租户之间相互干扰的分段多维性能模型,包括分段多维性能边界函数和多租户数据存储性能预测分析方法。通过基准测试数据采集、模型计算和性能分析,来判断是否每个数据存储节点能够综合地满足不同租户的不同性能需求。实验系统的分析结果表明,该方法能够判断在特定租户负载组合下每个数据节点是否过载,为分布式多租户数据布局提供基础。

The multi-tenant data storage is crucial for improving user experience, enhancing system executing efficiency and reducing operator cost. In consideration of massive tenants, how to satisfy the performance requirements of different tenants for shared data node is key problem to be solved. Aiming at the characteristics of different performaace requirement of multi-tenant data storage, this paper proposes a piecewise multiple performance model, containing piecewise multiple performance boundary function and performance prediction method for multi-tenant data storage. This method contains benchmark data collecting, model calculation, load forecasting and performance analysis. The goal is to determine whether each data node is able to meet the performance requirements of different tenants. The experiment system results show that the method can predict the workload in future and judge whether each data node is overloaded or not, which provides the basis for the distributed multi-tenant data layout.