Otimização da alocação de recursos em datacenters hierarquicamente distribuídos

Abstract

The rapidly increasing volume of services and applications, in addition to the high wireless access demand, are significant challenges for the next generation of mobile networks. The growth in the volume of applications is the reflection of the quantity of “things” that are being connected to the network and generating a huge data traffic. A new paradigm that is gaining recognition in the field of wireless networks, and is also responsible for part of this growth in the volume of services and applications, is the Internet of Things. The high amount of data that is generated by connecting those devices to the network will require significant computational resources to be processed and stored. A prominent approach to handling such large amount of data is the use of Cloud Computing, which uses datacenters for storage and data processing. However, traditional Cloud Computing, which has centralized resources, is not able to handle the high volume of data and the strict latency and Quality of Service requirements. Thus, to address such adversities, a new emerging concept known as Edge Cloud Computing has been proposed as an extension of the traditional Cloud Computing, bringing computational resources to the edge of the network and thereby creating a hierarchy of datacenters. In this way, the stricter requirements from services and applications, such as obtaining near-instant user experience, can be satisfied. In this work, a mathematical formulation for the dimensioning and provisioning of a hierarchy of DCs is proposed. According to the obtained results, the hierarchy of DCs provisioned and dimensioned using the proposed model can be better when compared to the others, being able to allocate 99% of the set of applications that were used in the tests and to decrease the data flow in the backhaul links that is generated by the high number of applications the would circulate through the network. The analysis highlight the necessity of bringing computational resources to the network’s edge in addition to an efficient applications allocation strategy in order to guarantee a better network performance.