When Group Buying Meets Cloud Computing

Low cost is the biggest slogan of cloud computing, serving as the major incentive for users to adopt cloud services. There is no doubt that further lowering cost will accelerate the adoption rate of cloud computing. While cloud providers can optimize their resource provisioning and scheduling to offer low-cost services to users, users can also take some actions to reduce cost by themselves, for example, via group buying.Group buying, where multiple buyers jointly make a large-volume order for a better deal, has recently become prevalent in e-commerce. In cloud computing, people have also started to raise the possibility of group buying. Group buying for cloud computing, however, is more complicated than group buying for conventional products because cloud computing demands are multi-attribute and time varying, and involve virtual machine assignment and even communication inside the cloud. Therefore, we see a timely need to understand group buying for cloud computing.Roughly speaking, the benefit of group buying for cloud resources comes directly from the different pricing schemes on the current cloud market, e.g., lower unit price for a bigger volume (or dedicated servers) and lower hourly cost for long-term reserved instances. Hence multiple users can pool their demands together and leverage those pricing schemes to reduce their cost.To implement group buying, we need to tackle several challenges. First, given multiple users, each of which has an arbitrary resource demand, we need an optimal purchasing decision to minimize the total cost for all users. Second, we need to design a fair cost sharing scheme for the users to share the total cost so that every user has an incentive to join group buying. The former issue is a hard optimization problem. While it may look similar to the cloud provider's resource provisioning problem, it is more challenging due to the intricacy of different pricing schemes such as on-demand instances and long-term reservation. The latter issue involves cooperative game theory, which will provide a theoretical foundation to ensure the stability of group buying. At last, based on the research on users’ group buying behaviors, the cloud provider may also manipulate the pricing schemes to retrieve a higher profit.This project aims to address the aforementioned challenges in group buying of cloud resources. The results will certainly provide firm answers on whether group-buying is the future pricing model in cloud computing and how to practice it if the answer is yes.