[Paper Review] On the Benefits of QoS-Differentiated Posted Pricing in Cloud Computing: An Analytical Model.

Xiaohu Wu, Francesco De Pellegrini|arXiv (Cornell University)|Sep 26, 2017

Cloud Computing and Resource Management3 citations

TL;DR

This paper proposes an analytical model for QoS-differentiated posted pricing in cloud computing, where fixed SLAs and pre-set prices allow users to choose service levels based on cost and performance. The model optimizes pricing and capacity allocation to maximize provider revenue and resource utilization, achieving up to an eleven-fold increase in utilization and a five-fold revenue boost compared to pay-as-you-go models.

ABSTRACT

When designing service and pricing models, an important goal is to both enable user-friendly and cost-effective access to services and achieve a high system's utilization. Such design could attract more users and improve the revenue. QoS-differentiated pricing represents an important direction to address the above design goal as illustrated by numerous studies in other fields such as Internet. In this paper, we propose the first analytical model of QoS-differentiated posted pricing in the context of cloud computing. In this model, a cloud provider offers a stable set of SLAs and their prices are also posted to users in advance, where higher QoSs are charged higher prices. As a result, users can directly choose their preferred SLA with a certain price, in contrast to many existing dynamic pricing mechanisms with the uncertain prices and availability of cloud resource. In order to maximizing the revenue of a cloud provider, the questions that arise here include: (1) setting the prices properly to direct users to the right SLAs best fitting their need, and, (2) given a fixed cloud capacity, determining how many servers should be assigned to each SLA and which users and how many of their jobs are admitted to be served. To this respect, we propose optimal schemes to jointly determine SLA-based prices and perform capacity planning. In spite of the usability of such a pricing model, simulations also show that it could improve the utilization of computing resource by up to eleven-fold increase, compared with the standard pay-as-you-go pricing model; furthermore, the revenue of a cloud provider could be improved by up to a five-fold increase.

Motivation & Objective

• To address the challenge of achieving high cloud resource utilization while ensuring cost-effective and user-friendly access for diverse users.
• To overcome the limitations of dynamic pricing models with uncertain prices and availability by introducing a stable, posted pricing mechanism.
• To jointly optimize SLA pricing and capacity planning to maximize cloud provider revenue under fixed resource capacity.
• To provide a framework for cloud providers to set prices and allocate servers to SLAs based on user demand and performance requirements.

Proposed method

• Develops an analytical model that formulates the joint optimization of SLA-based pricing and server allocation across multiple QoS tiers.
• Uses a posted pricing mechanism where each SLA has a fixed price corresponding to its QoS level, enabling transparent and predictable user choices.
• Applies optimization techniques to determine the optimal number of servers assigned to each SLA based on demand and cost constraints.
• Integrates user admission control to select which jobs are admitted based on SLA preferences and system capacity.
• Models user behavior by assuming users select SLAs based on price and QoS trade-offs, aiming to minimize cost while meeting performance needs.
• Employs mathematical analysis to derive optimal pricing and capacity allocation strategies that maximize provider revenue under capacity constraints.

Experimental results

Research questions

• RQ1 How can a cloud provider set optimal prices for QoS-differentiated SLAs to maximize revenue while matching user demand?
• RQ2 What is the optimal allocation of server capacity across different SLAs to balance utilization and revenue?
• RQ3 How does posted pricing with fixed SLAs compare to dynamic or pay-as-you-go pricing in terms of resource utilization?
• RQ4 What impact does QoS-differentiated pricing have on the overall revenue of a cloud provider?
• RQ5 How can user admission be managed to ensure efficient utilization and revenue generation under capacity limits?

Key findings

• The proposed QoS-differentiated posted pricing model increases cloud resource utilization by up to eleven-fold compared to the standard pay-as-you-go model.
• Provider revenue can be improved by up to five-fold under the proposed pricing and capacity allocation scheme.
• The model enables stable and predictable pricing, reducing uncertainty for users compared to dynamic pricing mechanisms.
• Joint optimization of pricing and capacity planning leads to better alignment between user SLA choices and system capacity.
• The analytical model successfully identifies optimal price points and server allocations that maximize revenue while maintaining high utilization.
• Simulations confirm that the model outperforms traditional pricing models in both efficiency and profitability metrics.