I want to use machine learning concepts in cloud computing.
Can any one share learning resource links for implementing Machine Learning algorithms in Java ?
I am not aware how to generate or use existing training and testing data specifically for cloud environment. I will be thankful if anyone can guide me for the same.
As my understanding of your question. You want to implement some ML algorithms for VM consolidation or Scheduling.
1- As CloudSim is just a java based simulator. you can use WEKA java libraries for run some ML algorithms.
For example:
You can implement multiple scheduling algorithms instead of single scheduling algorithm. By using ML classification, you can select efficient scheduling algorithm according to extracted training data from VMs. This data is given as input to the ML algorithm, as result, you will get classification rules help to select efficient scheduling algorithm to execute task. For implementation you can use CloudSim and WEKA for testing data-set
The following GitHub repository contains some Machine Learning algorithms implemented in Java. You may also search GitHub repositories for further implementations.
https://github.com/PacktPublishing/Machine-Learning-in-Java
There are many current examples of deep learning within cloud or SOA. Many of the previously given answers around Github are true. I am very interested in getting more context to your question however. Should you be interested in having a deeper or more clearly contextual discussion, I invite you to https://www.helioscsc.com
This is a great topic for research. I think it is important for management and orchestration especially for learning from monitoring.
I advice to use Python.
Go on. Good luck.
As my understanding of your question. You want to implement some ML algorithms for VM consolidation or Scheduling.
1- As CloudSim is just a java based simulator. you can use WEKA java libraries for run some ML algorithms.
For example:
You can implement multiple scheduling algorithms instead of single scheduling algorithm. By using ML classification, you can select efficient scheduling algorithm according to extracted training data from VMs. This data is given as input to the ML algorithm, as result, you will get classification rules help to select efficient scheduling algorithm to execute task. For implementation you can use CloudSim and WEKA for testing data-set
Well this is how I did it, might be done differently by other.
Also I only used GA and DE - of course there is the debate on whether or not these are actually ML or just metaheuristics.
Assumptions
1. Each PM has 2 PEs.
2. Overprovisioning is not considered.
3. Only applicable in offline workload allocation mode
Steps
1. Each PE represents a gene and collection of PEs form an individual.
2. If a VM (workload) is TO BE allocated to a PE, that PE is set to 1, else set to 0.
3. Randomly allocate VMs to PEs (ensuring you do not go beyond the capacity of the PE). This results in a population of individuals.
4. Calculate the fitness of the population.
In my case, I assumed fitness to be the total number of 0s. The lower the number of 0s, the better the utilization of resources (more resources are free).
5. Reoeatedly perform the various other steps in GA - crossover, mutation...) until a desirable fitness value is obtained.
6. The individual with the best fitness value is your potential VMs-to-PMs allocation map.
7. Use this map to select PMs in the file vmAllocationPolicyAbstract.
just my 2 cents, goodluck
@Olasupo Ajayi thank you so much for detail reply. I really appreciate your efforts for resolving my query. Thanx again.
- Badges
- Science topic