In today’s growing cloud world, where users are continuously demanding a large number of services or resources at the same time, cloud providers aim to meet their needs while maintaining service quality, an ideal QoS-based resource provisioning is required. In the consideration of the quality-of-service parameters, it is essential to place a greater emphasis on the scalability attribute, which aids in the design of complex resource provisioning frameworks. This study aims to determine how much work is done in light of scalability as the most important QoS attribute. We first conducted a detailed survey on similar QoS-based resource provisioning proposed frameworks/techniques in this article, which discusses QoS parameters with increasingly growing cloud usage expectations. Second, this paper focuses on scalability as the main QOS characteristic, with types, issues, review questions and research gaps discussed in detail, revealing that less work has been performed thus far. We will try to address scalability and resource provisioning problems with our proposed advance scalable QoS-based resource provisioning framework by integrating new modules resource scheduler, load balancer, resource tracker, and cloud user budget tracker in the resource provisioning process. Cloud providers can easily achieve scalability of resources while performing resource provisioning by integrating the working specialty of these sub modules.

Few scheduling, models focus on service-level agreement (SLA) enforcement, which limits their real-time applicability. Thus, this research work proposes the design of an improved task-side service level (TS2L) agreement model for efficient task-scheduling using elephant herd optimization (EHO) with deadline (BDP) awareness. TS2LBDP incorporates pattern analysis using ensemble hierarchical, k-means, and fuzzy C means (FCM) clustering methods. A combination of these modules assists in improving task diversity and scheduling efficiency. SLA-based distribution enhances scheduling fairness and reduces mean waiting time for different clients. The proposed model was tested on parallel workload archive wherein different cloud workload logs and their respective VM configurations are described. It was observed that the proposed model improves scheduling efficiency by 20%, task diversity by 45%, and deadline hit ratio by 1%. It is scalable and can be used for several types of cloud deployments.