Research On The Methods Of Bandwidth Guarantee Of Multi-tenant In Cloud Data Center Networks

Cloud computing is a new model for service and computing, the core concept of cloud computing is no longer relying on the local computers to do the computing, but depending on the third party who offer the computing and storage resources. The structure of the cloud data center is a distributed network, which is connected by computing nodes(such as servers)、storage nodes and network devices. The component of each node device is made of a series of resources.In cloud datacenter, the approach to scheduling the resources is the hottest point,especially the bandwidth guarantee of multi-tenant in datacenter network, causing a lot of discussions. While cloud computing providers offer guaranteed allocations for resources such as CPU and memory, they do not offer any guarantees for network resources. The lack of network guarantees prevents tenants from predicting lower bounds on the performance of their applications. While all the current methods of bandwidth guarantee have some shortages: some has to modify the hardware and software of switches, some needs to modify the kernel of the servers’ system, some has the requirements of topology, some needs the tenants’ request of sending and receiving bandwidth equal, and some can’t make full use of the bottle links.However, the method of bandwidth guarantee of multi-tenant in cloud data center networks in this thesis can only deploy in servers, use the unmodified commodity switches. The server nodes use the full-virtualization hypervisor with SDN technology, and don’t need to modify the kernel of the server’s system, which makes it easy to deploy and implement. Meanwhile, it can support the current network topologies, and satisfy the unequal requests of sending and receiving bandwidth of multi-tenant, as well as realize the dynamic adjustment of bandwidth, and make full use of the residual bandwidth of the bottle link.This thesis firstly makes an introduction of the technology using in deployment, including multi-tenant virtual data center, OpenFlow protocol, Open vSwitch and server virtualization technology. Then the thesis does the research and analysis on current bandwidth guarantee in cloud data center, and based on it to propose a modified method of bandwidth guarantee with servers only, as well as propose the global design structure of the server node. According to the function, the method is divided into three modules:(1) the tenants’ request model: clear the tenants’ requests and realize the initial setting;(2) the minimum bandwidth guarantee module: finish the computing and transforming from tenants’ request model to VM-to-VM minimum bandwidth guarantee;(3) the bandwidth allocation and adjustment module: finish the allocation of each VM-to-VM flow’s sending rate, and the adjustment of feedback. The thesis makes a detailed algorithm design of the three modules’ implementation principles, and proposes the specific methods of their real deployments. Then use the OPNET simulation tool to design nodes, uses two different network topologies and three different tenants’ requests to verify the principles and analyze the simulation results. At last, this thesis sets up a small cloud data center, verifies and tests in real conditions, and finishes the preliminary implementation of the tenants’ requests, minimum bandwidth guarantee and bandwidth allocation and adjustment.