Maximizing Profit of Cloud Brokers under Quantized Billing Cycles: a Dynamic Pricing Strategy based on Ski-Rental Problem

TL;DR

This paper introduces dynamic pricing algorithms inspired by the ski-rental problem to maximize cloud broker profits under fixed billing cycle constraints, validated through theoretical analysis and real-world simulations.

Contribution

It presents two novel online algorithms for cloud brokers that optimize profit by regulating demand via dynamic pricing under quantized billing cycles.

Findings

Algorithms achieve favorable competitive ratios.

Simulations demonstrate improved profit over baseline methods.

Effective demand regulation reduces costs significantly.

Abstract

In cloud computing, users scale their resources (computational) based on their need. There is massive literature dealing with such resource scaling algorithms. These works ignore a fundamental constrain imposed by all Cloud Service Providers (CSP), i.e. one has to pay for a fixed minimum duration irrespective of their usage. Such quantization in billing cycles poses problem for users with sporadic workload. In recent literature, Cloud Broker (CB) has been introduced for the benefit of such users. A CB rents resources from CSP and in turn provides service to users to generate profit. Contract between CB and user is that of pay-what-you-use/pay-per-use. However CB faces the challenge of Quantized Billing Cycles as it negotiates with CSP. We design two algorithms, one fully online and the other partially online, which maximizes the profit of the CB. The key idea is to regulate users demand using dynamic pricing. Our algorithm is inspired by the Ski-Rental problem. We derive competitive ratio of these algorithms and also conduct simulations using real world traces to prove the efficiency of our algorithm.