Creating a peer population determines the criteria that the testing agents use to locate the appropriate public peers to use for a particular test run. Our agents receive the test information and then search for available peers that fall into the categories specified within a particular peer population.
When machines initially sign up to become public peers, they specify certain characteristics that they possess. An example would be that someone who signs up is located in California and has a high broadband connection type.
When an agent locates a peer that matches the criteria in the population, it runs addition tests to determine if that peer is what it says it is. Using the example, this would mean that the agent determines by the peer's IP whether it is located in California, and runs a speed test to confirm that it has a high broadband connection type.
If the tests show that this peer is using a low broadband connection type, then that test run is stored for charting, but it does not fit into a peer population that requested only high broadband machines. If this situation occurs three times in a row for a particular machine, then the peer is automatically reclassified as a different machine profile (i.e. it will now be a low broadband machine instead of high broadband).
This peer testing can cause discrepancies during Last Mile charting: an initial chart may contain more data points than a chart broken down by the intended peer populations because the initial chart may represent peers that were excluded after the initial tests.
Breaking down a chart by peer populations ensures that you are viewing data from tests that matched your peer population criteria.