QPides! is an application whose main target is to satisfy the need of a more fluid and detailed information between restaurants and their clients. This information includes the restaurant's location, table allocation, availability, menu content and the customers' comments. Additionally, provides a reliable and secure channel to submit customized orders and payments. This enables to reduce the workload on restaurant staff and invest such resources in maximizing the quality of the service provided. One of the most important distinguishing features of QPides! application is the table allocation process, since an effective table allocation can be crucial to a restaurant's profitability. Inefficient use of tables means that the restaurant is losing potential customers, but overbooking means that customers are delayed or feel cramped, and so are unlikely to return. In addition, customer behavior is dynamic, and so table allocation should be flexible or quickly reconfigurable, to avoid delays. Restaurant table allocation could be improved if the software can be used by staff with less expertise and knowledge, and that can help the automation and optimization of the allocation process. Specially, for these reasons the idea of QPides! was proposed. This work aims to develop the backend of QPides! focusing on table allocation problem. Our proposed solution involves designing different table-allocation algorithms. These algorithms solve a constraint satisfaction problem, looking the best combination of tables at any given time that maximize the occupancy rate. The web environment used, in which the algorithms are implemented, comprises Meteor, an open-source JavaScript web application framework optimized for real-time apps, and MongoDB database. The main objective is to improve restaurants performance in terms of occupancy rate and response time. Evaluation of QPides! performance is done by applying load testing for each table allocation algorithm. All testing carried out in this work are based on dummy data that try to simulate real-world scenarios. The results obtained in the experimental measurements prove that searching with backtracking (BT) is much more efficient. Performing a properly pruning of the search tree does not prevent find the best solution and reduces the computational cost.