基于Swift的存储负载均衡策略研究与实现

Research and Implementation of Work Load Balancing Strategy Based on Swift Storage System

OpenStack是目前应用广泛的开源云计算平台。OpenStack中的Swift模块通过实现分布式存储来提供对海量数据快速、安全、可靠的存储服务。针对Swift元数据存储是完全均匀随机分布,并且只提供面向数据量的负载平衡造成负载均衡策略问题,提出神经网络和梯度算法结合的寻优方法。神经网络一般来说具有全局寻优的特性,但也存在获取的解易早熟,且可能存在局部寻优性能差的缺陷。本文在传统神经网络的基础上,提出将其与梯度算法结合,通过一系列的迭代过程,保留最优解。实验证明,该方法既解决传统梯度算法中的搜索速度较慢的问题,且保证最后的解收敛于全局最优解。

OpenStack is currently a widely used open source cloud computing platform.The Swift module in OpenStack provides fast,safe,and reliable storage services for massive data by implementing distributed storage.Aiming at the problem of load balancing strategy caused by Swift metadata storage which is completely uniform and random distribution and only provides load balancing for data volume,an optimization method combining neural network and gradient algorithm is proposed.Neural networks generally have the characteristics of global optimization,but they also have the disadvantages that the obtained solutions are easy to mature and may have poor local optimization performance.Based on the traditional neural network,this paper proposes to combine it with the gradient algorithm,and retain the optimal solution through a series of iterative processes.Experiments show that this method not only solves the problem of slow search speed in traditional gradient algorithms,but also ensures that the final solution converges to the global optimal solution.