A Joint Chance Constraints Stochastic Model for Reliable Network-Aware IaaS Provisioning in Datacenters | ||
| Journal of Engineering Science and Military Technologies | ||
| Articles in Press, Accepted Manuscript, Available Online from 05 November 2025 | ||
| Document Type: Original Article | ||
| DOI: 10.21608/ejmtc.2025.430960.1343 | ||
| Author | ||
| Khalid Metwally* | ||
| Military Technical College | ||
| Abstract | ||
| Cloud service providers (CSPs) increasingly confront the challenge of provisioning Infrastructure-as-a-Service (IaaS) amid volatile and bursty request arrivals. Static, mean-based allocation policies are ill-suited to this environment: provisioning at expected demand can maximize nominal throughput yet incurs a heightened risk of capacity violations, while worst-case robustness preserves SLAs but strands substantial idle spaces and inflates cost. To reconcile reliability and efficiency, we cast joint server-network resource allocation as a chance-constrained optimization with distinct reliability targets for node and link resources. Within this framework, we employ Chance-Constrained Programming (CCP) to model demand variability and uncertainty levels, and derive deterministic equivalents that convert stochastic constraints into per-resource quantile buffers, thereby enabling tractable, SLA-aligned provisioning and routing decisions. Using a VL2 datacenter topology, we empirically compare four policies: deterministic, robust, node-only chance constraints, and joint chance constraints. The joint formulation consistently reduces server- and link-level violations relative to deterministic provisioning and achieves higher acceptance than robust schemes at comparable risk, especially when bandwidth and memory are the binding resources. We also examine modeling assumptions, the sensitivity of outcomes to confidence parameters, and practical deployment considerations. Overall, probabilistic buffering emerges as a capacity-efficient alternative to uniform robustness, delivering reliable performance while improving resource utilization in cloud datacenters. | ||
| Keywords | ||
| Chance Constraints; Resource Allocation; IaaS; Datacenters; Stochastic Model | ||
|
Statistics Article View: 20 |
||
