RISCLESS: A Reinforcement Learning Strategy to Exploit Unused Cloud\n Resources

One of the main objectives of Cloud Providers (CP) is to guarantee the\nService-Level Agreement (SLA) of customers while reducing operating costs. To\nachieve this goal, CPs have built large-scale datacenters. This leads, however,\nto underutilized resources and an increase in costs. A way to improve the\nutilization of resources is to reclaim the unused parts and resell them at a\nlower price. Providing SLA guarantees to customers on reclaimed resources is a\nchallenge due to their high volatility. Some state-of-the-art solutions\nconsider keeping a proportion of resources free to absorb sudden variation in\nworkloads. Others consider stable resources on top of the volatile ones to fill\nin for the lost resources. However, these strategies either reduce the amount\nof reclaimable resources or operate on less volatile ones such as Amazon Spot\ninstance. In this paper, we proposed RISCLESS, a Reinforcement Learning\nstrategy to exploit unused Cloud resources. Our approach consists of using a\nsmall proportion of stable on-demand resources alongside the ephemeral ones in\norder to guarantee customers SLA and reduce the overall costs. The approach\ndecides when and how much stable resources to allocate in order to fulfill\ncustomers' demands. RISCLESS improved the CPs' profits by an average of 15.9%\ncompared to state-of-the-art strategies. It also reduced the SLA violation time\nby an average of 36.7% while increasing the amount of used ephemeral resources\nby 19.5% on average\n