Introduction
This project approaches the problem of optimizing infrastructure technologies as a continuous multi objective combinatorial optimization problem in a dynamic environment
Meta-Heuristics, e.g. Evolutionary Algorithms (EA), have been found operating well in such optimization problem fields.
Technologies throughout the stack are targeted. A focus is put on open technologies first.
Overall optimizing, rearranging, calibrating of technology settings throughout the life-cycle of an instance of a technology are what is seeked to address with godon.
Objective
It is
- augmenting human operation engineers at bringing about performance improvements
- it simplifies the process through standardization and industrialization
- reduces prior knowledge needed about configuration changes and implications
- less toil in terms of engineering hours spent
- tackles the wide spread neglection of broader performance tuning
- pragmatic operations engineering complementing instrument
- focussed on open technologies first
- approximating an optimal state in a dynamic environment changing over time
- leveraging metaheuristics algorithms of all kinds to explore combinatorial configuration spaces
- betting on parallelization and acceleration of metaheuristics
It is not
- fully off-hands automation as human setup, supervision and planning is required
- a machine learning or data analysis oriented technology
- kept to a minimum
- ideally only used if needed in the implementation details of a metaheuristic
- guaranteeing a global optimum in the search space of configurations
- rather approximating a better than untouched state
- a metaheuristics framework
Sponsors
Greatest esteem to:
- OSU Open Source Lab (https://osuosl.org) for bestowing generously with resources on their openstack infrastructure
References
[1] Inspiring POC work for the project;Autonomous_Configuration_of_Network_Parameters_in_Operating_Systems_using_Evolutionary_Algorithms
[2] Underpinning the potential to accelerate metaheurstics by parallelization;A unified view of parallel multi-objective evolutionary algorithms