Find the latency for locations around the world from each datacenter in GCP and AWS. Process those latencies, and create an interactive map that lets you explore the expected latency to a datacenter from anywhere in the world.
You can see this at https://bit.io/latency
There's a blog post on this as well - https://innerjoin.bit.io/exploring-cloud-datacenter-latency-e6245278e71b
The web site, which is static, and you can literally open the app.html file in your browswer to explore it. The geojson files checked into this repo should be good enough.
To create/update the coverage geojson, you need to run the more interesting bits of this: the tracer & the locator.
The tracer uses mtr and the mtrpacket libraries to send, in parallel, traceroutes to
IPs spread throughout the world. The tracer records those results per-IP in the probes table.
Then, when the locator is run, the locator goes through the probes (joined with the ip_location
table to prevent re-locating IPs we already have located) and tries two methods to geolocate
the IP (the open maxmind database and the commerical ipinfo data). Once the IP is located we updated
the public database's latency_from_datacenter table with the latitude, longitude, and latency.
Then one last script creates the geojson files by computing the alphashape around the points, one alphashape for every millisecond of latency between 10 and 150 milliseconds for all regions and cloud providers.
You need a GCP account & and AWS account, two bit.io databases, and an ipinfo account.
Set the environment variables:
TRACER_PRIV_DB_CONNSTR- the bit.io (or Postgres) connect string for the private databaseTRACER_PUBLIC_DB_CONNSTR- the bit.io (or Postgres) connect string for the public databaseTRACER_IPINFO_KEY- the ipinfo keyTRACER_IPINFO_BATCH_URL- the ipinfo batch URL
Then edit tracer/aws_json/* and put your AWS customer ID in the right places. Do the same in tracer/build-aws.sh
Make sure you are logged into gcloud and aws - both those CLI tools are required.
Dump the files in schema/ to the public & private bit.io databases.
Running tracer/build-aws.sh will deploy the tracer image to everywhere in AWS and run 10 tasks for each region, twice a day.
Running tracer/build.sh will do mostly the same for GCP.
Every so often (or as a scheduled job) run ./update_geojson.sh