Automated generation of policies to support elastic scaling in cloud environments
Richard Gil Martinez, Antónia Lopes, Luı́s Rodrigues · 2017
Cloud computing has enabled myriad of applications to benefit from dynamic allocation of resources. Yet, as the offer of resources and pricing schema varies, the selection of the best configuration to adapt to environment conditions and business goals becomes a harder problem to solve manually. Humans may fail to anticipate conditions that arise at run-time or to elaborate complex adaptation plans. Hence, expert's know-how should be complemented with the use of intelligent tools. Automated planning presents itself as a promising solution to complex decision-making, when finding strategic plans over a multidimensional space is required. We address the challenges of relying on adaptation planning for the generation of high-level policies, namely: representing the adaptation problem using a planning language, taking advantage of off-the-shelf planners to generate and select adequate plans, and incorporating techniques to scan initial system states and merge plans under common conditions.