Scholarlink

AI-as-a-Service (AIaaS) redefines cloud computing by enabling scalable and autonomous AI-based solutions. Optimizing AI workloads on heterogeneous infrastructure for the multi-cloud context is challenging because of performance heterogeneity, management of resources, and cost. Experimental performance evaluation for the performance of AIaaS in the context of the multi-cloud environment is presented in this paper with the MLPerf Benchmarks Dataset. Google Cloud's TPU v4 offers the highest throughput at 60 TFLOPS but at the highest cost per execution at $0.40, while AWS balances cost and performance at 45 TFLOPS at $0.35 per execution. Azure offers the lowest cost at $0.30, but the lowest throughput at 38 TFOPs. A workload balancing strategy for the multi-cloud context improves performance by 15% and cost savings by 12%, demonstrating the possibility for AI-optimized AIaaS deployment. This work contributes to intelligent AI workload orchestration, efficiency, and cost savings for autonomous AI-based cloud computing.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Agent AI-as-a-Service (AIaaS) in Multi-Cloud Environments: Challenges, Opportunities, and the Future of Autonomous AI-Driven Cloud Computing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Rahul Vadisetty

摘要

AI-as-a-Service (AIaaS) redefines cloud computing by enabling scalable and autonomous AI-based solutions. Optimizing AI workloads on heterogeneous infrastructure for the multi-cloud context is challenging because of performance heterogeneity, management of resources, and cost. Experimental performance evaluation for the performance of AIaaS in the context of the multi-cloud environment is presented in this paper with the MLPerf Benchmarks Dataset. Google Cloud's TPU v4 offers the highest throughput at 60 TFLOPS but at the highest cost per execution at $0.40, while AWS balances cost and performance at 45 TFLOPS at $0.35 per execution. Azure offers the lowest cost at $0.30, but the lowest throughput at 38 TFOPs. A workload balancing strategy for the multi-cloud context improves performance by 15% and cost savings by 12%, demonstrating the possibility for AI-optimized AIaaS deployment. This work contributes to intelligent AI workload orchestration, efficiency, and cost savings for autonomous AI-based cloud computing.