About Google FI and QoE

Last month, Google introduced Project Fi, which they define as

[…] a program to deliver a fast, easy wireless experience in close partnership with leading carriers, hardware makers, and our users.

Among other features, Google claims it will provide global coverage, seamless handovers between Cellular and WiFi to provide you with the best possible signal strength and it will keep your SIM in the cloud so you can link it to all your devices. An MVNO on steroids.

At the time of writing this post, this project is limited to the US, two carriers and just one device model. This constitutes ‘Lean Startup’ Standard Operating Procedure, so if this experiment succeeds, we will definitively see a wider rollout.

And what will the success factors of this initiative be? Project Fi’s website highlights three aspects: Network, Plan and Experience. We could not agree more. We are all about experience here

But as per the look of its message, Google seems to equate experience to speed, measured indirectly using signal strength as proxy.

Isn’t it too simplistic to limit the handover triggers to signal strength levels? We know that throughput and latency are the key factors to Quality of Experience:

  • Voice calls depend on Quality, Interference and even Power Budget, like the traditional 2G Handover types proved. When carried over IP, they are very sensitive to latency.
  • Browsing does not need very high speed.
  • Video calls and games are more sensitive to latency.
  • FTPs are not so affected by latency.
  • Facebook, twitter are not speed nor latency demanding.
  • Youtube demands consistent latency for a fluid playback.

Optimizing the technology and network in use for all services is not a trivial problem. Google Fi’s handover and reselection algorithms will need calibrating and adjusting to factor a QoE metric of sorts. Furthermore, it will need to consider available capacity as it grows. It is not only about finding the best technology and network for one given customer at one given moment, but also about prioritizing the available resources when many customers using disparate services use the same radio electric space.

For example, the algorithm should be able to decide to keep a voice call on a slower cellular network while sending a concurrent video call to wifi or a browsing session to a faster cellular network.

Will Google work it out? It will be interesting to see the first QoE results.