基于负载均衡的在线云存储流量调度研究

在线云存储流量的调度问题是当前网络的研究热点。针对在线云存储系统中的文件上传请求调度问题,现有方案很少满足为用户提供不同带宽保证这一需求。针对不同付费级别用户要求不同带宽保证的服务场景,设计了基于请求队列长度的最大权重调度机制以及实际中可用的分布式二次随机选择调度算法,旨在实现系统服务器的流量均衡,同时最大化系统吞吐量。证明了最大权重调度机制能够保证系统稳定性,并且最大化系统吞吐量。实验结果表明,二次随机选择算法能够获得接近于最大权重调度机制的性能,较小的时间延迟和较大的系统吞吐量。其性能优于传统的Round-Robin调度算法。

Flow scheduling for load balance in cloud storage

Load balancing and flow scheduling are important in large-scale on-line cloud storage systems. Currently, there are few solutions consider various bandwidth guarantee for cloud users. We propose a centralized max-weight allocation scheme to schedule the upload traffic in a cloud storage system and balance the workload. We prove that the scheme can stabilize the system and maximize the throughput. We further design a distributed two-random-choice algorithm, which does not require a centralized coordinator in the system nor need to maintain global system information. The distributed algorithm is more cost efficient and robust compared to the centralized one. Simulation results show that the two-random-choice algorithm can achieve low latency and high throughput. Its performance is close to that of the max-weigh allocation scheme, and is much better than the traditional Round-Robin algorithm.