Abstrakt: |
Data center storage systems have different workloads and characteristics. In a centralized managed storage paradigm (software defined data center), providing a fault-tolerant system requires the best possible placement, and the least possible numbers of software defined storage (SDS) controllers. The separation of storage intelligence from storage resources raises the SDS controller's placement problem, which in turn, raises the reliability's issue (the main controller's failure leads to unavailability). The unavailability—down time—cost too much to an organization. To protect against the single point of failure, and at the same time, facilitate a fault-tolerant storage network, center of gravity technique is presented for optimal location selection. The method analyzes different work-loads, distance, and link bandwidth among controllers and OpenFlow enabled switches, and proposes multiple optimal locations for the controllers' deployment. We show the best optimal controllers' placement for various scenarios, e.g., single optimal placement for a big enterprise, and multiple proposed optimal placements for multiple SDS controllers containing various openFlow switches. These controllers can be the part of a large distributed Internet of Things (IoT) architecture. Experiments show that the work-load, distance, and the bandwidth are three important factors that can affect the whole network's reliability (the controllers' placement). [ABSTRACT FROM AUTHOR] |