The video service has already become one of the Internet killer applications. To deliver the ever growing video traffic with the Quality of Experience (QoE) guarantee, deploying the infrastructure of Content Delivery Network (CDN) closer and closer to users has been a noticeable trend. And this trend becomes more promising when recently CDNs providers start utilizing the storage and network resources of the edge network devices (e.g., the Wi-Fi Access Points) for content delivery [13], i.e., crowdsourced CDN. In this article, we envision the crowdsourced CDN and propose a set of practical strategies to guide the implementation of such new paradigm. Firstly, we perform a large-scale measurement to explore the geo-content popularity and the geo-heterogeneity of the resource distribution. Benefiting from the measurement observations, we design the key solutions for the system operation, including regional content popularity prediction based on Markov Decision Process (MDP), region partition for the edge device organization and the collaborative content replication. Moreover, we consider the economic issues and propose a sequential auction mechanism to cope with the resource allocation when multiple Content Providers (CPs) compete for the edge resources for their content caching. We also demonstrate the performance gain of our design with data-driven simulation.