Abstract
The rapid development of cloud computing requires improvement of pricing and allocation mechanisms of cloud resources. Dynamic pricing and allocation mechanisms are considered convenient, due to characteristics of cloud resources and the fact that demand for cloud resources is not uniform. The aims of such a mechanism are to optimize the utilization of cloud resources, to maximize cloud providers' revenues, and to minimize prices for cloud customers. Auction-based pricing and allocation mechanisms are often used since resources are allocated to the customers that value them the most, and prices are determined depending on the supply and demand conditions. Selection of an appropriate bidding strategy is a very important issue and requires the comprehensive approach. This chapter analyses the benefits of auction-based pricing and allocation mechanism in the cloud environment. In addition, the effects of different bidding strategies application are addressed.
TopIntroduction
The rapid development of technologies triggered the emergence of cloud computing. Cloud computing enables network access to a scalable and elastic pool of shareable physical or virtual resources. Resources in the cloud environment are generally provisioned as cloud instances, i.e. Virtual Machines (VMs) with the required CPU, memory and I/O resources. The main advantages of cloud computing are scalability, low costs, dependability, service-driven, self-healing and multi-faceted. However, there are also numerous challenges including a vulnerability, bandwidth costs, transparency access, trust, reliability, security, availability, resource management, task scheduling, and performance (Gullhav & Nygreen, 2016; Sebastio, Gnecco & Bemporad, 2017). The dominant factor in supporting customers to use virtual resources and move towards cloud computing is pricing. Cloud customers pay for required cloud resources depending on instance type and pricing mechanism applied. In addition, different levels of guarantees in terms of availability and termination are possible. Pricing mechanisms in the cloud environment can be generally classified into static and dynamic. The static pricing mechanisms are the dominant scheme in the cloud environment and involve the charging the fixed price per hour and per instance. The dynamic pricing mechanisms include charging the usage of cloud resources as long as the cloud customers' bids are equal or exceed the price per instance and per time interval, determined by the cloud provider (Alzhouri & Agarwal, 2015). For example, Amazon Elastic Compute Cloud (EC2) uses three different pricing options for cloud resources: Reserved instances, On-demand instances, and Spot instances. Reserved instances are provided with long-term availability guarantees, On-demand instances are provided for specific time periods and are guaranteed to be available once assigned, and Spot instances are provided through an auction-like mechanism, but their availability depends on the time-varying Spot price. Reservation and On-demand pricing mechanisms are considered as static since prices for cloud resources are invariant for both under-utilization and over-utilization of the resources. Idle cloud resources can be offered as Spot instances. This purchasing option offers no guarantees on minimum sustained availability. Submitting a bid for the access to Spot instances, cloud customers specify the maximum price they are willing to pay. The actual price they pay is not the value of the bid, but the value of the Spot price. Cloud provider set Spot price depending on the availability of the idle resources and current demand. When the Spot price exceeds the value of the bid, the Spot instance is terminated. Cloud provider can also revoke Spot Instances when the supply of available Spot Instances decreases. Recently, a new variety of Spot instances, Spot Block instances, are introduced. These instances run continuously for a finite duration, up to 6 hours. When cloud customer submits a bid, two parameters need to be specified: the value of the bid (the maximum price willing to pay per hour) and the Block duration parameter (the number of hours that Spot Block instance will run). The Spot Block instance is terminated only at the end of the Spot Block duration.
As a dynamic form of pricing, auction mechanisms can be an effective and promising solution for profit maximization of cloud providers. These mechanisms provide price variation depending on the changes in supply and demand by creating competition among cloud customers. Therefore, auctions can balance supply and demand for cloud resources. They support cloud customers for truthful bidding and allocate the resources to the customers that value them the most. Various auction-based allocation and pricing mechanisms can be used for cloud resources: Uniform price auctions, Second-price auctions, Combinatorial auctions, Double auctions etc. In this chapter, the analysis of different auction-based mechanisms for pricing and allocation of cloud resources is provided.
Key Terms in this Chapter
Cloud Resources: Virtual machines (VM) with specific computational and storage capacities provided to the cloud customers in the form of cloud instances. Cloud customers can rent those instances depending on their availability and applied pricing mechanism.
Auction Mechanism: Often used pricing mechanism for cloud provider's idle resources. If properly set, an auction mechanism should support customers to bid truthfully for cloud resources and allocate resources to customers that value them the most.
Cloud Customer: Content/service provider or end user that rents cloud resources for execution specific tasks depending on their requirements.
Spot Block Instances: Subset of cloud provider's Spot instances. Cloud customers submit bids, determining the maximum price they are willing to pay for specific cloud instance per hour and the Spot Block duration up to six hours. The Spot Block runs uninterrupted up to Spot Block durations expires, regardless on the Spot Block prices.
Spot Instances: Cloud provider's idle resources provided through an auction-like mechanism, where cloud customers submit bids, i.e. maximum price they are willing to pay for specific cloud instance of one-hour duration. The spot instance can be terminated whenever the submitted bid exceeds the Spot price determined by the cloud provider.
Pricing and Allocation Mechanism: Mechanism that determines which cloud resources will be allocated to which cloud customers at a specific price and specific duration.
Cloud Provider: Undertaking that provides storage and computational resources to cloud customers depending on appropriate pricing and allocation mechanism and resources' availability.