Reliable IoT storage for sensor monitoring applications: Trading off early redundancy injection costs and repair costs

Providing reliable storage for sensor monitoring in IoT systems is key to manage nodes unavailability and lack of connectivity between the sensor devices and outside networks. Although reliable distributed storage systems typically assume that new devices are available to cope with disk, node or rack losses, this is not generally the case in monitoring systems. This paper proposes a data protection scheme based on network coding for monitoring networks that considers the cost of repair and the cost of setting up the initial protection for the data. We provide a theoretical characterization of the data protection problem with different operating regimes. We also analyze the mechanisms to trade-off network costs of the initial protection phase and the repair phases. Our numerical results show that a judicious choice between the preparation and repair phase costs can result in a reduction of protection cost as high as four times with respect to a naive selection, e.g., only focused on the repair costs.