The power law for cache misses was first established by C. K. Chow in his 1974 paper,[1] supported by experimental data on hit ratios for stack processing by Richard Mattson in 1971.
[4] The power law can only give an estimate of the miss rate only up to a certain value of cache size.
[3] The validity of the power law of cache misses also depends on the size of the working memory set in a given process and also on the temporal re-reference pattern of cache blocks in a process.
If a process has a small working memory set relative to the cache size, capacity misses are unlikely and the power law does not hold.
Hartstein et al.[4] found that whereas the cache misses for lower levels do not follow a strict power law, as long as the lower level cache is considerably larger than the higher level cache, the miss rate function can be approximated to the power law.