In this work we look at Big Data transmission using wireless networks. The first goal is to model the modern communication node equipped with multibeam antenna working on IEEE 802.11ac/ax (5G/6G) standard. We first model the flow of packets through the node using Queuing theory that reflects the more modern communication standard. To make the model practically applicable we also make it computationally simple that it can be executed in real time in smaller processors. This single node model is also extended to a adhoc wireless mesh network scenario and a complete model for the network is derived. We then test the model and use it to optimize a simulated network. The results show that using this accurate model to optimize networks vastly outperforms the traditional models popularly used today. With the queuing model developed, we analyze and develop both Routing and Transport layer protocols for the challenging problem of aerial drone flocks. Drone flocks pose unique challenge in networking where the mobility is really high as the aerial drones can move at speeds of about 70mph. This results in a network topology that is constantly changing. We use prediction, clustering based methods to establish a hierarchical routing scene. The routes developed are designed to use multiple paths to have redundancies built in for robustness. This routing protocol outperforms the popular OLSR protocol for mesh routing by a significant margin. It exchanges much less control packets to maintain the routes and also has better throughput and lesser packet loss because of multiple paths. With this routing protocol established, we also develop a TCP protocol that avoids congestion by choosing different paths for the packets based on congestion in each path. This TCP protocol uses a queuing model similar to the one used in modeling IEEE 802.11ac. To simulate and verify these protocols, a MATLAB based Network Simulator is built which enables the use of complex mobility patterns. This simulator functions similar to the popular Network Simulator 3 (NS3) with the added advantage of being able to use both time and event based simulations. The final contribution of this work is to use the resulting robust communication grid and queuing model to develop a mobile grid computing scene that can function using drones. This mobile grid we establish exploits the mobility and varying computational capability to optimally distribute computations among the various nodes in the network. The simulations demonstrate the performance of this technique in comparison to work stealing algorithm popularly used in grid computing scenarios.