Dismiss Notice
What Is This?

It's a constantly growing, global map of 3,342,644 PokéStops and Gyms for Pokémon GO.

The data is from Outgress (an Ingress analytics system).

Declined Cooperate with Lampje

Discussion in 'Feature Requests' started by AlfonsoML, May 23, 2016.

Page 2 of 2
  1. mogmismo

    mogmismo New Member Resistance L16

    Wow, that's pretty awesome. I'm primarily using Solr (which has very good geospatial search) on our SolrCloud cluster, but we also run ElasticSearch for logging and other side projects. Those have to be some beefy ElasticSearch boxes. At last year's Solr/Lucine conference, I saw that Rackspace was bragging on indexing 50K emails / second into Solr in a seminar ( ). Thanks for sharing, I like to know how things are built. I appreciate all your work, and the donation of not insignificant server resources.
     
    mogmismo, Jan 10, 2017
    #21
  2. digitalpoint

    digitalpoint Administrator Staff Member Illuminator Enlightened L15

    The biggest bottleneck on my end is simply the connectivity of the cluster. How many emails can you pass through the network basically. My setup is all interconnected with 54Gbit InfiniBand, so that's how it's able to swallow that much email.

    Rackspace probably is just using normal 1Gb Ethernet or maybe 10Gb.

    On the geospatial stuff, there was a lot of stuff I tried that worked fine but simply didn't scale or couldn't answer some particular types of queries I wanted to do. The geospatial stuff is improved in the new version of ElasticSearch I saw (based on Lucene improvements), but still not better than my purpose built system. One of the biggies I needed to do is when looking at the Intel map in opposition or friendly attack mode. The heat maps are weighted by how often and how recent portals are attacked. It spins through millions of 3D geospatial data (3rd dimension being time) and internally generates a single weighted value for each portal.
     
    digitalpoint, Jan 10, 2017
    #22
Page 2 of 2