Reseach On The Tenants’ Network Performance Guarantee Method In Cloud Data Center With Using Servers And Access Switches

With the rise and rapid development of cloud computing technology, data center become more and more important infrastructure. But because of a series of factors: size, performance, and sercurity and so on, in the traditional data center, make it no longer to meet the various requirements of the increasing users’ businesses, so the cloud data center comes.Cloud data center can consist of hundreds of thousands of server resources, as well as the interconnection network resources between servers, provide on-demand access to computing resources, allow a large number of users to migrate their applications or businesses to the cloud. These applications have more or less involved in the communication between tenants’ virtual machines in the internal network of cloud data center. In order to meet the various applications or businesses requirements of tenants, the cloud data center provide various types of virtual machine for users to choose and use. All the virtual machines have a clear definition of physical resources such as CPU, storage and memory, but the network resources which is vital for some tenants’ businesses is not defined or guaranteed explicitly. However, to some businesses of tenants, the communication network performance between tenants’ VM is a key factor to run their businesses efficiently. Therefore, the research content of this article is to study how to make tenants’ network performance guaranteed efficiently in the cloud data center environment.First, current tenants’ network performance guarantee methods are introduced synoptically in this paper, and in accordance with their design schemes, this paper summarizes three design objectives: limit the tenant’s network bandwidth quota; guarantee the tenant’s minimum network bandwidth; tenant’s network performance must be proportional to its network expense. Then, this paper combines these three objectives and designs a novel tenants’ network performance guarantee scheme in the cloud data center. This scheme is achieved by combining servers and access switches which are directly connecting with servers. On the server side, this paper adds the tenant’s VM corresponding buffer queue, and runs an improved DRR scheduling algorithm for each queue to make the VM’s egress bandwidth guaranteed. On the switch side, this paper improves the CIOQ switch structure, adds the tenant corresponding descriptor buffer queue in each egress port, and runs an improved DRR scheduling algorithm. Based on the scheduling results, then the packets which are cached in the switch ingress port will be sent out, so the VM’s ingress bandwidth can be guaranteed. At the same time, this paper uses back pressure flow control mechanism to produce feedback package to send to the data generate source side in the server, to slow down the data transmission from the source side, which aim to avoid the happening of congestion to the greatest extent. This paper also uses the mapping mechanism of tenant’s network payment to divide the tenant’s network payment into two parts: guaranteed bandwidth cost and additional bandwidth cost, the guaranteed bandwidth is determined by the minimum guaranteed bandwidth and tenant’s network weight. These two parameters will be used in the scheduler, eventually improve the performance of tenant’s network reasonably and efficiently.This paper uses OPNET simulation software to simulate the scheme model proposed in this paper, and builds a FatTree network topology structure which consists of 80 switches of eight ports and 128 servers. On this basis, this paper establishes a series of simulation scenarios, compares and analyzes the simulation results, finally validates the proposed scheme in this paper can efficiently guarantee the tenants’ network performance.