HIGH-PERFORMANCE COMPUTING ARCHITECTURES TO STRENGTHEN CLOUD INFRASTRUCTURE SECURITY

Abstract

This study examined how high-performance computing architectures enhanced cloud infrastructure security by analyzing the computational, networking, storage, and isolation factors that shaped cryptographic efficiency and security-analytics responsiveness. Experimental evaluation was conducted using CPU-only, GPU-accelerated, and FPGA-enabled configurations, supported by multiple interconnects and storage hierarchies. Quantitative findings showed that GPU-based nodes achieved the highest encryption throughput, averaging 1,860 MB/s, compared to 1,240 MB/s on FPGA-enabled systems and 420 MB/s on CPU-only nodes. Encryption latency demonstrated similar patterns, with GPU nodes averaging 7.9 MS, FPGA nodes 6.2 MS, and CPU-based systems 18.4 Ms. Interconnect performance emerged as a critical determinant of secure data movement; InfiniBand links recorded packet-loss rates as low as 0.43%, compared to 2.81% on Ethernet under matched workloads. Storage subsystems also demonstrated substantial performance variation, with NV Me-based configurations supporting log-ingestion rates of 184,000 events per second, nearly doubling the rate observed on SSD-based systems. Virtualization and containerization strategies influenced isolation stability, with cross-tenant interference measuring 18.6% under KVM and 24.9% in Docker-based environments. Regression results confirmed accelerator type (β = +0.61), interconnect type (β = –0.73), and storage configuration (β = +0.81) as significant predictors of key security-performance metrics, indicating that architectural decisions directly shaped encryption stability, packet-flow reliability, and detection timeliness. By integrating experimental results with insights from 312 reviewed studies, the research demonstrated that coordinated improvements across compute, network, storage, and virtualization layers substantially strengthened cloud-security performance in high-demand environments.