The cloud computing is evolving as a high-performance computing platform due to broad-scale, flexible computational architecture and heterogeneous collection of autonomous systems. Cloud technology uses concept of virtualization for managing resources, which develops resource scheduling as a key issue. The scheduling of cloud tasks is an NP-complete problem and therefore irreconcilable with particular solution. Also, with the huge collection of a database system, the management of resources and tasks becomes complex with specific to the completion time requirements and cost constraints. To resolve this problem, a number of meta-heuristic algorithms have been developed. Due to redundant wastage of resources and time, the under and over-provisioning is one kind of issues leads to either degradation in performance or wastage of cloud resources. To overcome these kinds of problems, we introduce a task scheduling approach by incorporating reinforcement learning along with the nature-inspired meta-heuristic optimization to maximizing cloud throughput, minimizing completion time & production cost in IaaS cloud. By reinforcement learning, the agent will choose appropriate action among a set of available actions and the scheduler succeeds towards task allocation resulted to decrease makespan and increasing system utilization rate.