Skip to content

Use Case I: How our AWS implementation is different

Jump to bottom
dkakkar edited this page Jan 30, 2020 · 1 revision
  • Extremely Fast: 200,000 distances/sec
  • Resource Optimized: Basic EC2 m4.xlarge server
    • Two layered approach:
    • Bottom layer of Geohash based clustering assisting fast data retrieval by the
  • Top layer of R-tree based KNN search
  • Dimensional reduction: 2D coordinates reduced to 1D Geohash providing compressed storage
  • Cost Effective: Infrastructure costs only 75/month
  • Hardware and Software Optimization: Optimization of search algorithm and hard disk storage
  • Customized AMI: Ready to use
  • Compressed Storage: 18 TB reduced to 1.5 TB
  • GPU based: Extremely fast GPU based processing of results
Clone this wiki locally