Search code examples
hbasegeomesa

How Geomesa overcome HBase hotspotting problem for spatial indexes


How Geomesa solves the hotspotting problem for Spatial and Spatio-temporal data which is a common problem in space-filling curve?

I read below documents but not able to understand what is Striping? and How it helps.

http://www.eclipse.org/community/eclipse_newsletter/2014/march/article3.php


Solution

  • That article is fairly out of date. Currently, GeoMesa mitigates hot-spotting through prepending a shard prefix to each row, which ensures that some data will be spread out among different table regions. See here and here for details on configuring shards.

    If you know your data distribution up front, GeoMesa supports custom pre-splitting of index tables, which can be used to completely eliminate hot spots.

    In all cases, hot-spotting is further mitigated by automatic table splitting in HBase or Accumulo. Regions will split based on size, so areas of dense features will automatically be split up once they hit the threshold.