EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

1364. A comparison of two approaches to hardware capacity planning for cloud infrastructure under demand uncertainty

Invited abstract in session WD-45: Planning Techniques for Decision Support, stream Decision Support Systems.

Wednesday, 14:30-16:00
Room: 30 (building: 324)

Authors (first author is the speaker)

1. Laura Wolf
Institute for Operations Research (IOR), Karlsruhe Institute of Technology (KIT)
2. Patrick Jahnke
SAP SE
3. Stefan Nickel
Institute for Operations Research (IOR), Karlsruhe Institute of Technology (KIT)

Abstract

Cloud computing and especially Infrastructure-as-a-Service (IaaS) are becoming more and more relevant in industrial applications. In IaaS, cloud providers offer customers resources such as compute, network and storage. From the provider’s point of view, the question arises of how many physical resources are required to cover the uncertain demand, to achieve utilization targets and deal with fluctuating workload demand. This decision must be made from a midterm perspective due to increasing supply chain cycles.
We compare two approaches to deal with this hardware capacity planning problem for cloud infrastructure under demand uncertainty (CPCI_DU) to determine the needed capacity of a datacenter. The decision problem is modeled as a two-stage stochastic optimization problem and solved optimally. The results are evaluated using metrics from stochastic optimization such as the value of stochastic solution and the expected value of perfect information. This method is compared to an optimization-simulation loop. Here, the deterministic optimization problem is solved on the first stage and the impact of uncertainty is evaluated in the subsequent simulation. This optimization-simulation loop is executed until a prespecified stopping criterion is reached. The results show the strengths of the respective approaches to enable robust hardware planning.

Keywords

Status: accepted

Back to the list of papers