Once a relatively restricted subdomain of data analytics, GIS (Geographic Information Science) use cases are suddenly front-and-center due to the proliferation of geospatial data from sensor, smart phone, social media and transportation data. These new forms of data are not only often of much higher volume and velocity that traditional GIS datasets but often require a mixture of GIS and traditional analytics.
MapD bridges this gap by providing an ultra-fast analytics platform with advanced geospatial analytics and visualization features. Unlike most GIS platforms it allows exploration of both spatial and non-spatial data at speeds that leave traditional systems in the dust.
The first area where traditional GIS platforms fail is with visualizing mass amounts of geospatial data. Whether due to legacy CPU rendering or network “gaps” between server and client, traditional tools fall over when faced with even moderate (read tens of thousands of points) sized datasets.
Compare that to MapD, which, when needed can not only run the necessary query over the data but leverage the native graphics pipeline of the GPUs to render the result without copying the data to the CPU. This allows the system to scale to rendering tens of millions of points, lines or polygons with latencies measured in milliseconds and only requires sending the rendered result in compress PNG form to the client rather than potentially gigabytes of raw data.
The second area where GIS platforms often fail is when required to perform numerically complex geospatial calculations at scale (potentially over billions of records). For example, merely converting modest amounts data from one geographic projection to another (an arithmetically intensive operation requiring multiple trigonometric calculations) can stop legacy systems in their tracks due to a lack of CPU horsepower. MapD, on the other hand, has tens of thousands of cores at its disposal across multiple GPUs and can perform such calculations on-the-fly with little performance overhead. This approach also applies to complex calculations such as ad-hoc filters based on user-drawn polygons. By leveraging the immense computational throughput of the GPUs, users of MapD can perform such operations fully interactively and in real-time.
To see some of MapD’s geospatial prowess in action, please see our live Tweetmap demo or our political donations demo - showing how MapD can render and allow interactive filtering on billions of rows of data without lag.