Network Bandwidth Optimization Tor Cloud Computing

With the rapid development of information technology, the computing model is changing from self-organization data center to on-demand cloud computing. More and more big enterprise, such as Amazon, Microsoft, Google, etc. begin to provide tenants with the opportunity of migrating services and applications to public cloud platform. Now many cloud providers host thousands of tenants and each tenant has ten to thousand VMs. The traffic between VMs changes dynamically all the time, thus the data center needs to transport tens million flow per second. Due to the overwhelming demand growth of cloud computing network, it’s essential to further study the cloud bandwidth resource optimization problem.The commonly used bandwidth allocation method is static reserved, which means tenants purchase VM of the fixed bandwidth and none can use others’idle bandwidth. Given the burst flow of data center, it will make low network average utilization by using static reserved bandwidth. On the other hand, congestion link may cause low communication rate, SLA violation and additional cost. If cloud providers can avoid link congestion, it will increase the profit of providers and improve the application performance of tenants.In view of the above problems, this paper firstly focuses on how to manage the idle bandwidth in cloud computing. Considering the selfish tenants are not willing to provide resources for free, we proposes an idle bandwidth management mechanism based on reverse auction. It makes tenants voluntary sale their idle bandwidth to cloud providers which can improve network utilization. The mechanism achieves a win-win situation for cloud providers and tenants. It can not only improve the cloud network utilization, but also increase the tenants’benefits. Besides, in order to guarantee the fairness between tenants, we introduces the truthful character of auction mechanism. Finally we proves that this method is truthfulness.At the same time, in order to solve the network congestion problem in cloud computing, this paper proposes a bandwidth allocation method based on congestion avoidance. The novel method consists of two phase. Firstly, tenants reserve required minimum bandwidth to ensure their applications keep running. To encourage tenants only require actual minimum bandwidth, this article requires tenants to pay extra cost of minimum bandwidth. Then we sort all links according to their congestion probability, and assign tenants to the link of smaller congestion probability. In addition, if a link congestion happens, we will migrate some tenants from congested link to appropriate uncongested link.