Network capacity planning in cloud - Challenges and approaches

We are witnessing the transition of telecom services to cloud, but real-time and high QoS requirements make the transition a slow and difficult process. Through its successful application in IT world, we have seen that cloud offers numerous benefits, including cost reduction, scalability and automation. Furthermore, constantly decreasing hardware costs and auto-scaling capability give an illusion of having infinite resources in cloud. Thus, planning the capacity of a cloud seems unnecessary. In fact, it is one of the most underestimated problems in cloud computing. In order to guarantee service quality to their users, telecom service providers will need to manage network capacity in cloud as well. While in traditional network setting, where resources are static and dedicated, capacity planning involves estimating the optimal amount of hardware equipment to be purchased; in dynamic virtualized environment, it means estimating the amount of needed virtual resources. The main objectives of this thesis work are to: explore traditional network capacity planning and the current cloud capacity planning approaches, compare them in order to define the main requirements for a tool that will manage NW capacity in cloud; and to create and test a prototype based on these requirements. The used research methodology includes literature review, survey and a case study. Based on the research results it was concluded that there are many challenges in network capacity management, such as excessive manual work and the absence of an end-to-end tool. In cloud, solutions need to be auto-scalable, to react in real-time and have integrated performance monitor. For telecom, the biggest challenges in cloud are retaining high QoS and avoiding service interruption. These are the most important factors to be taken into account in capacity planning process. Thus, a network capacity planning tool primarily needs to addresses performance vs. scalability trade-off