Prevent Throughput Rule oscillation and cache pollution.
Created by: dsparacio
- To prevent throughput oscillation. (Defined as, when an end user's bandwidth sits between two advertised bandwidth of two renditions and the client keeps switching between each quality at a fast pace.
- One solution is to lock out the upper quality for X amount of time. If we switch from quality x to x+1 and back to x within 10 seconds, lockoutCount for index x+1 is incremented. If lockoutCount === 3 you lock out x+1 and do not allow a switch up to that quality for a few minutes.
- Do not let a cached fragment's load time pollute the average throughput values.
- Solution is to look at load speed and if > then X do you use in averaging. Question is what should the magic number be?