It is all good to have access to large volume of data generated with high velocity which normally spans variety of domains and usually comes with levels of veracity. However, it is all useful only when we turn it into a value. In particular, in the domain of “Big Spatial Data” we deal with massive amounts of real-time spatial and spatio-temporal data obtained from billions of sensors and location-aware devices. The use of Big Spatial Data spans a variety of applications including social networks, earth sciences, transportation, communication networks, online maps, smart cities and urban planning, remote sensing, and crisis and evacuation management, to name but a few. Turning Big Spatial Data into value is challenging and requires introduction of fundamentally new spatio-temporal algorithms, methods, and systems that can process, mine, and analyze massive amounts of fast and heterogeneous spatio-temporal data in a timely manner.

Keynote: Big Spatial Data @ Facebook

Big Geospatial Data at scale has all the challenges of data at scale along with some quirks very specific to spatio-temporal data. However, these very quirks (like the bounds of latitude/longitude, Euclidean vs. great circle distances, the "true" shape of the earth and the extremely skewed distribution of geospatial features) can be leveraged into interesting and productive trade-offs to offset and address these challenges. With more and more mobile devices thrown into the mix (both as producers and consumers of spatio-temporal data), realtime and accurate lookup of points and polygons based on GPS locations and queries about k-nearest and Top-K based on geospatial contexts are a very common and relevant problem. At the same time, providing scalable offline aggregation and query capabilities of spatio-temporal data for analytics use cases becomes vital to making sense of it.
The Facebook Location Infrastructure team handles spatio-temporal data at Facebook scale (using a mix of in-house and open source technologies and pragmatic trade-offs/decisions). This presentation will cover various design decisions and architectural choices taken to ramp up Trillions of operations per day on a heterogeneous mix of spatio-temporal data (for both online and analytics oriented use cases).

• ● Xiaoming Gao, Research Scientist at Facebook Inc.
  

  Xiaoming graduated from Indiana University with a PhD in Jan 2015 and currently works at Facebook as a Research Scientist in the Maps team. His thesis work and previous research were related to topics in Distributed Systems, Cloud Computing, Big Data processing, and Social Media Data analysis. His current project is about applying deep learning techniques for object recognition in large scale satellite imagery processing. Google scholar profile

  
  
• ● Saurav Mohapatra, Software Engineer at Facebook Inc.

  Saurav works for Facebook on the Location Infrastructure team building Geospatial Indexing and Querying services that handle Spatio-temporal Data at Facebook Scale. Previously he worked for Salesforce.com as an Architect for their real-time infrastructure and prior to that was the co-founder/Director of Technology for the web conferencing startup Dimdim (acquired by Salesforce.com in 2011).

  
  
• ● Lihan Bin, Software Engineer at Facebook Inc.

  Lihan works for Facebook on the Location Infrastructure team. Prior to Facebook, he worked on GPGPU standard OpenCL at Qualcomm and served as Qualcomm representative in Khronos Group and HSA foundation. Prior to that he worked for AMD on GPGPU performance and tooling.