I suggest (1) ease of implementation (2) effectiveness in maintaining output close to optimum (3) ability to handle successfully multiple power maxima as you can get in partial shade conditions. Speed is not usually an issue - solar radiance doesn't change that fast, may be with a time constant of seconds as a cloud covers the sun. Most computer driven algorithms are surely going to be hundreds of times faster than this.
As the colleague Tony hinted it is required that the MPPT controller has high tracking speed with less tracking errors and lowest possible complexity. The ability to track the global peak in case of the presence of multiple peaks. A review of the MPPT techniques is given in paper at the link:Article Enhancing the tracking techniques for the global maximum pow...