Skip to content
blickly edited this page Dec 12, 2011 · 4 revisions

Fill in what we are trying to measure. (A vs B etc etc?)

Measurements

  • Latency vs. phone signal strength
  • Latency vs. server load (i.e. fake phones)
  • Dropped packets vs. server load
  • Request rate vs. Pacmen drawn

Methodology

How to measure dropped packets?

  • Sent from phone to server, but server never receives
    • Two cause: network OR server
  • Response sent from server to phone, but phone never receives
    • Only one real cause: network drops the packet => indep variable signal strength
    • Maybe we don't need to measure this, since we can measure network in the phone sending case.
    • This is the uncommon failure case.

How to measure the latency between two phones?

Potential solutions:

  • Compare timestamp of sensing GPS on one phone vs. receiving that sensing on another phone (but this would require fairly accurate time synchronization between phones).
  • On a single phone, check roundtrip time with another phone (may require reasoning of concurrent behavior of server to tell when a round-trip has really happened or not).
  • Calculate the latency indirectly by summing up various sub-latencies (such as phone-to-server latency, plus server wait latency, etc.)