Best type of chart to visualize geographic distribution of servers

May 28, 2023 8:15 AM

I am doing some research on a provider's network geographical distribution. If my data is for 70 countries and 2,000 servers, how could I best visually display this? The relationship I'm trying to show is the potential even or uneven geographic distribution of servers while also taking in to consideration the fact that the USA would have more servers than a country than the size of Ireland or Hungary. This seems like the least complicated way of controlling for size rather than getting into population, etc.
posted by ascetic to Computers & Internet (5 answers total)

I haven't done this myself, but you probably want to set up some sort of Cartogram distribution, which is the general term for distorting a geographic map to match some other variable like population density. I don't know how you're generally planning on visualizing things, but something like this page explaining how to do Cartogram distributions in the GIS system could be helpful.
posted by JZig at 9:47 AM on May 28, 2023 [1 favorite]

You might want to do servers per capita?

Also see tile grid maps.
posted by oceano at 10:20 AM on May 28, 2023 [1 favorite]

3d map. You might be able to do this in Excel, by enabling Power Maps then 3D Maps.

If you don't have that, then you might use Tableau for free. Two options:

1. If you have the city or zip or precise location of each server, plot each server as 1 dot, and reduce each dot's transparency. This results in a map that shows darker areas where you have a concentration of servers, while still showing you their geographic extent.

2. If you only know where the country is (imprecise geographic location), I'd plot the sum of servers per country as a dot, one dot per country. Then, scale the darkness of the dot to to the sum of servers. This will have a darker dot where there are more servers.
posted by nicodine at 5:18 PM on May 28, 2023

You need either country size or country population data to "normalize" the number of servers in each country. It's not inherently easier to normalize by size vs population, unless you already have size data. Population data is a better choice as "people per server" is a more direct measure of potential server traffic than "square miles per server".

The generic term for plotting a map with each region color coded is choropleth. More info here: https://datavizcatalogue.com/methods/choropleth.html.
posted by grog at 8:14 AM on May 29, 2023

Does it make sense to normalize over population or over something like data throughput (or whatever technical metric reflects the rationale for multiple servers)?
posted by kokaku at 2:23 PM on May 29, 2023