Inter-regional Performance Analysis of Data Center Networks in Public Cloud Systems

摘要

Recently, due to the development of the service using the network, it has become possible to provide services not only to the country where the service was developed but also to other countries. Building a system in a service area can result in poor quality of service due to a large amount of capital and personnel spent on costs, operations, and maintenance. Therefore, many service providers construct and operate a system using a public cloud system. Public cloud systems allow users to borrow hardware and software resources from cloud providers to build the systems they need. The public cloud system has a data center in each region, and it is possible for a user to create a virtual machine in the area and configure the desired type of system. There are various types of performance metrics that evaluate the system such as hardware performance, disk performance and network performance. However, if the service is provided at various locations, the user of the service should show the least level of satisfactory for system performance, and it is necessary to evaluate the performance of the network, which is a measure for evaluating the fast response speed and the amount of transmission to the user. In this paper, we compare and analyze network performance among data centers in public cloud systems. RTT and SCP throughput were measured to compare and analyze the network performance between data centers in a public cloud system. In order to measure network performance between data centers in a public cloud system, a virtual machine was created for each region. RTT test and throughput test were performed as virtual machines in the data centers of other regions in the created virtual machines.