Introducing Network Multi-Tenancy for Cloud-Based Enterprise Resource Planning: An IoT Application

The cloud service providers make a considerable investment in setting up the data centers backbone network with the aim to maximize the network resource. However, the actual utilization of the network resources is hard to predict. With the invent of Software Defined Networking (SDN) and OpenFlow protocol, the network control layer has got the capability to communicate with the applications or services which are offered by the service provider. Moreover, a Software Defined Data center suggests resource virtualization at computing, storage, and network layer. The multi-tenancy is a well-accepted architecture in cloud computing where a single instance of a software application serves multiple customers. This work is a first of its kind, which aims at maximizing the network resources with respect to multi-tenancy at the network layer. In this work, with network multitenancy, different customers IoT traffic flows are prioritized, and then network resources are allocated to the traffic flows dynamically based on the priority. We considered a scenario of Enterprise Resource Planning (ERP) solutions deployed in the cloud which offers services in the form of Software as a Service to the customers. The IoT devices deployed at the manufacturing site makes transactions on the cloud ERP. This work focuses on prioritizing the ERP- IoT traffic to meets the demands of a multi-tenant data center network. The ERP-IoT flows are prioritized using a regression based machine learning technique for predicting the response time for execution of a query caused by a traffic flow in the ERP backend server. Later, the ERP-IoT flows are assigned to multiple queues created on each network device in data center. This assignment is performed based on the traffic flow priority and Demand Supply scores, which aims at maximizing network resource utilization. During performance evaluation, we observed that the proposed work with network multi-tenancy shows more than 10% increase in service providers utility with respect to standard data center single queue operations.