This research focuses on a smooth rendering of continuous 2D map based on a smooth 3D vario-scale geographical data structure. A Space Scale Cube (SSC) offers non-redundant geometric data for the different level of details. SSC model represents geographic data as a closed polyhedron, to generate a 2D map; SSC is intersected with the projection plane; resulting in a set of 2D polygons. However, problems emerge when creating maps with a large sized SSC dataset under web environment due to limited bandwidth and decoding speed. Repetitively transmitting data from the server to the client can be time and bandwidth consuming. A preprocess should be applied to a source that allows the follow-up development of an online traffic and time-efficient prototype. After preprocessing, large sized data will be subdivided based on octree algorithm to minimize transmission time from server to the client; moreover, accessible to WebGL. A prototype has been developed which enables smooth and timely vario-scale map rendering against heavy user actions such as fast zooming and panning in a short period. Modified prototype program allows query of only relevant data chunks by current viewport position; it prevents repeated loading of same chunks; what is more, repeated transmission of data from outside to GPU is eliminated. A tree structure is embedded at the client side that facilitates retrieve time. Rendering happens every frame; hence the prototype responses to heavy user actions timely. Also, it can obtain coordinates in RD coordinate system by double clicking. After testing the modified program with a 9km by 9km dataset online, exceptional performance is indicated by a high average frame rate (57 fps) and low main memory occupation (with a network speed at 9MB/s). The prototype performance is significant affected by the client network condition; low network speed can decrease the frame rate. For instance, the web service achieved a frame rate of 47 fps at a network speed at 6MB/s.