Trust-Aware Adaptive Resource Management Framework for SLA-Aware Multi-Tenant Cloud Computing Environments

Abstract

The multi-tenant cloud computing platforms are making significant strides due to their ability to provide scalable and efficient computing power. But, typical resource contention issues due to noisy neighbours, data hungry tenants etc., can result in Service Level Agreement (SLA) violations, performance dropouts, unfair resource use etc. and poor workload isolation to tackle these challenges, this paper proposes a Trust-Aware Adaptive Resource Management Framework (TAARMF) for intelligent resource allocation and SLA-aware workload scheduling in multi-tenant cloud computing environment. The desirable framework combines tenant profiling, assessment of trust, identification of noisy tenants, identification of greedy tenants, flexible resource allocation optimization and optimization under SLA in one single framework. The behaviour and resource usage for the tenants is represented by the 2019 Google Cluster Workload Trace. They solve the problem of noisy tenants and abnormal usage of a resource by greedy ones by using these two methods, isolation forest and XGBoost. Based on the current trust scores and workload priorities, the computational resources are dynamically allocated by the adaptive scheduler without altering the workload isolation policies and with a fairness-aware method. From the experimental results, the proposed framework is seen to be very accurate, with 96% accuracy in the classification of tenants, and SLA and isolation success rates of 97.2% and 95.4% respectively and a high Jain's Fairness Index of 0.96. These findings show that the proposed framework makes much better use of resources, promotes fairness, divides workloads according to fairness, and provides better service reliability than traditional resource management methods. Thus, TAARMF serves as a useful solution for dealing with resource contention and ensuring resource management conforms with SLA requirements in a large-scale multi-tenant cloud computing infrastructure.