Thunder: A private cloud architecture designed for high usability.

摘要

Cloud computing is a technological strategy for saving time, money, and resources within an organization. Underfunded and understaffed organizations benefit the most from a cloud architecture because it can help to alleviate a cost burden allowing funds to be used more effectively. Therefore, we believe that non-profit organizations, such as schools, libraries, non-profit medical facilities, and others have the most to gain from cloud computing. Cloud computing has played a major role in shaping large for-profit businesses like Google, Amazon, and Microsoft. Research has suggested that cultural barriers make it difficult for professionals in non-profits to adopt cloud computing technology.;One key challenge faced by organizations for which a cloud architecture would be beneficial is the deployment and management process. In order for private cloud computing to become a viable solution for struggling organizations, much work needs to be done to simplify and improve the deployment process. We describe a new cloud architecture called THUNDER, which is a recursive backronym meaning "THUNDER Helps Underfunded Nonprofits Distribute Electronic Resources.''.;THUNDER introduces strategies which are meant to help struggling organizations to decrease costs. Virtual machine load balancing attempts to distribute the load across multiple nodes in order to maximize potential performance of virtual machines. Virtual machine consolidation attempts to utilize as few computational resources as possible in order to maximize the potential cost savings of maintaining idle nodes.;We present an evaluation of THUNDER using metrics designed to compare its ease of use with that of other architectures. We also present an empirical evaluation of THUNDER by which users deploy and use the architecture, and then participate in a survey. THUNDER utilizes a new load balancing algorithm, which we have called "RAIN'', meaning "Rating Assisted Instantiation Negotiation''. RAIN attempts to optimize virtual machine instance placement by choosing compute nodes that have the best potential based on real-time metrics aggregation and a rating algorithm. Results show that the THUNDER is preferable for inexperienced users when compared to that of OpenStack and Eucalyptus. Additionally, experimental results show that RAIN is more efficient at placing virtual machines than the more typical approach of round-robin.