Analyzing CSP Trustworthiness and Predicting Cloud Service Performance

Abstract

Analytics firm Cyence estimated Amazon's four-hour cloud computing outage in 2017 “cost S&P 500 companies at least $ \$ $150 million” and traffic monitoring firm Apica claimed “54 of the top 100 online retailers saw site performance slump by at least 20 percent”. According to Ponemon, 2015 data center outages cost Fortune 1000 companies between $ \$ $1.25 and $ \$ $2.5 billion. Despite potential risks, the cloud computing industry continues to grow. For example, Internet of Things, which is projected to grow 266% between 2013 and 2020, will drive increased demand on cloud computing as data across multiple industries is collected and sent back to cloud data centers for processing. RightScale estimates enterprises will continue to increase cloud demand with 85% having multi-cloud strategies. This growth and dependency will influence risk exposure and potential for impact (e.g. availability, performance, security, financial). The research in this paper and proposed solution calculates cloud service provider (CSP) trustworthiness levels and predicts cloud service and cloud service level agreement (SLA) availability performance. Evolving industry standards (e.g. NIST, ISO/IEC) for cloud SLAs and existing work regarding CSP trustworthiness will be leveraged as regression-based predictive models are constructed to analyze CSP cloud computing services, SLA performance and CSP trustworthiness.

Keywords

• CSP
• cloud service provider
• cloud service
• cloud security
• machine learning
• predictive model