Smart cities have been developed by integrating IoT networks that generate huge data from heterogeneous sources, including environmental sensors, surveillance systems, and utility meters. Processing this vast volume of data faces challenges regarding nonstandard device protocols, limited network capacity, and the need for subsecond latency in critical tasks like autonomous vehicle coordination. Therefore, appropriate strategies are required for optimizing real-time sensor data processing through the integration of edge and cloud computing with advanced analytics. In this study, we aim to review optimization technologies, including adaptive sampling. It is critical to adjust collection rates while considering anomalies and construct tiered data storage architectures that separate hot, warm, and cold data for efficiency. Furthermore, reinforcement learning and container orchestration platforms such as Kubernetes must be introduced for dynamic resource distribution and load balancing across the computing continuum. The practical application of these architectures in Shenzhen, China, is highlighted. Shenzhen’s City Brain (named after the Hangzhou City Brain), utilizing distributed edge-cloud processing, reduced average traffic intersection waiting times by 12.6% and achieved a 92% detection rate for environmental anomalies using fog-computing gateways. Transforms from batch-oriented processes to real-time methods led to significant changes in urban governance. However, sustaining these systems requires addressing security vulnerabilities, ensuring data privacy through differential privacy, and standardizing interoperability policies to maximize the return on investment for smart cities.