This is a very systems implementation oriented question but all we can say from our Econophysical Engineering Physics oriented research using String Field Theory is thaat the Systems Coding should include the cloud specification using location shell search protocols etc. and the Genetic Algorithm should be integrated with this initiating section in the overall algorithm. The ideas behind our Functor Algebra Calculus (Mallick, Hamburger, Mallick (2016)) and Mallick (2007, 2014, 2015) on www.researchgate.net/Soumitra K Mallick,should be used for the implementation.
Soumitra K. Mallick
for Soumitra K. Mallick, Nick Hamburger, Sandipan Mallick
GA: A Package for Genetic Algorithms in R but not in the cloud.
This package contains most of crossover and mutation types in GA for some of the combinatorial problems. You can google the title to have the tutorial pdf file for it
Do you want to implement a genetic algorithm in cloudsim? or do you want to use a genetic algorithm to optimized some parameter set for a cloudsim simulation? In the first case I have no idea, for the second you need to create a smart fitness function for your cloudsim simulation and in some way linked to the GA (As Osman mention there are packages in R, also C++, python etc... I'm not sure if there's one in Java)
You can implement GA at VM level (which is cloudlet to VM mapping) or at Host level (which is VM to Host mapping). In your case, I think you will optimise energy and performance for VM allocation based on GA. For implementation, you can implement your algorithm in power package. To be more specific, in "PowerVmAllocationPolicyMigrationAbstract" class.