We are at a tipping point where more data is being produced by the Internet of Things (IoT) than can be meaningfully processed, as this requires tackling the sheer Volume, Velocity and Variety of the data simultaneously. Semantic web reasoning has proven to be ideal for targeting Variety, enabling the integration of data from various sources. However, reasoning is too computationally expensive, compared to the velocity of the IoT. Therefore, a paradigm shift is happening from cloud computing towards computing the data as close as possible to its source, i.e. at the edge. The objective of this proposal is to enable efficient and scalable real-time processing of IoT data with Cascading Reasoning (CR), i.e. a layered reasoning approach consisting of various levels of complexity. CR naturally matches the edge processing paradigm, as the low-complexity processing techniques can be pushed down to the edge, while complex layers can be processed on intermediate & cloud nodes with more resources. Due to the size and complexity of the edge environment, it is extremely hard for users to issue efficient queries across such a CR layered framework. I will investigate how queries can be optimized and automatically translated to the various levels of the cascade to enable efficient and real-time edge analytics. As IoT data requires to solve an integration problem, efficient reasoning techniques will be investigated that can deal with the velocity, volume, and variety of IoT data.