7. Conclusions
In this work we address the problem of finding a suitable technique to automatize the deployment of complex systems
assembled from a large number of interconnected components. We propose an algorithm able to compute in polynomial
time the actions needed to deploy such a system and we prove soundness and completeness of this novel approach. We then
present METIS, a proof of concept implementation of the proposed approach and we validate its efficiency and effectiveness
in two ways. On the one hand, by testing its performances on extremely large synthetic problem instances and, on the other
hand, by discussing its integration within a framework for automatic deployment of cloud applications called Blender [42].
The developed technique shares some commonalities with both the top–down holistic and the bottom–up DevOps ap
proaches since, given a (partial) global description of the final configuration to obtain, the deployment plan is automatically
inferred from a declarative description of individual components. Results are encouraging: METIS is able to produce plans
in less than a minute, for scenarios involving hundreds of components.
7. ConclusionsIn this work we address the problem of finding a suitable technique to automatize the deployment of complex systemsassembled from a large number of interconnected components. We propose an algorithm able to compute in polynomialtime the actions needed to deploy such a system and we prove soundness and completeness of this novel approach. We thenpresent METIS, a proof of concept implementation of the proposed approach and we validate its efficiency and effectivenessin two ways. On the one hand, by testing its performances on extremely large synthetic problem instances and, on the otherhand, by discussing its integration within a framework for automatic deployment of cloud applications called Blender [42].The developed technique shares some commonalities with both the top–down holistic and the bottom–up DevOps approaches since, given a (partial) global description of the final configuration to obtain, the deployment plan is automaticallyinferred from a declarative description of individual components. Results are encouraging: METIS is able to produce plansin less than a minute, for scenarios involving hundreds of components.
การแปล กรุณารอสักครู่..
