Every 20 minutes our server (located in Germany) sends 7 different RPC requests to each known public RPC node, and records the time each takes to respond.
This scatter chart shows the time for responses in milliseconds over the last 12 hours.
When the response time is set to -300ms (negative!) it actually means the request was unsuccessful and returned an error, whereas if the response time is set to 5000ms it means the request took at least 5000ms.
Hovering over the node labels highlights those in the chart, and hovering over data-points shows which node the response was for.
You can also (de)select individual nodes by clicking on them in the key, which makes it easier to compare just a few nodes.
If any gaps appear in the data, it simply means our monitoring process wasn't working for a while, it doesn't mean no RPCs were working. In this unlikely case the errors line would be full of data-points!