Ray Of Hope

dynamic scheduler that can self-learn – 2

On thinking further (based on the earlier article: http://www.justkernel.com/Blogs/?p=322), I am not sure whether what I described is just an optimization or learning. I have a data and the prediction confirms to the set of the data that I have. That is just an optimization and not learning. But what I need is some model that can also handle future scenarios for eg change in workload affecting all the related features i.e a model that can self learn to handle future scenarios. (user can use any type of workload).

Further, also, in this scenario I don’t have much data to train the model. I can , at the max, run the scheduler on 2 or 3 systems under few workloads and collect the relevant data.So, I can;t say that I have large amount of samples to go for supervised learning. I have to go with the unsupervised learning.

Now the question is which model to use. Right now evaluating K Algo.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.