一种支持负载均衡的存储调度算法

A Scheduling Algorithm for Load Balance Sensitive Storage

针对应用层存储聚合中的调度问题,提出了一种支持负载均衡的存储调度(LBS)算法.LBS是一种基于策略的调度算法,它将应用对存储资源的需求转换为一系列约束,再通过分析约束之间的关系选择合适的存储节点或者已有的调度方案,从而提高了调度方案的复用率,维护了策略复用与节点负载之间的平衡关系,寻找到最佳的负载均衡策略.模拟测试表明,LBS算法在负载均衡方面和策略耦合方面明显优于Least和Random算法,负载均衡指标最高可提升10倍左右.

According to the scheduling problem in storage aggregating of application-level, a scheduling algorithm for load balance sensitive storage （LBS） is proposed based on policy. In the LBS algorithm, requirements from applications for storage are represented as a series of restrictions. The LBS then makes choice between new appropriate storage nodes and existing scheduling scenario by analyzing relations between restrictions so that the reusability of scheduler scenarios can be improved. The LBS maintains the balance between the reusability of policy and the load of storage, and achieves the goal of load balance finally. The simulation and comparisons show that the LBS outperforms the Random and the Least obviously in load balancing, policy decupling and scalability, and that the load balance of the LBS improves 10 times against the two baseline algorithms.