November 24, 2002

TiVo Simulator Data

The TiVo Nation simulator finished all the tests in about 10-15 minutes. It generated about 9MB of data. I processed the raw output utility values, dropping the first 24 hours (48 iterations) of the data to allow the cache to somewhat stabilize. I then calculated the average, variance, and standard deviation. At first glance, it seems that the average utility is very close between the two different cache policies tested (TiVo-human and fully-aware approximating) but the variance and standard deviation is much lower with the fully-aware approximating policy. I need to plot some comparative graphs, and also try a third fully-aware approximating compressing cache policy. This policy should show some real gains over the TiVo-human policy.

Posted by josuah at November 24, 2002 10:12 PM UTC+00:00