Researchers at The George Washington University are developing a system and method for applying statistical and machine learning models to calculate and rank cloud service provider (CSP) trustworthiness levels and predict cloud computing service performance and availability compliance against industry standards based cloud service level agreements (SLAs).
Considering the rate of adoption and dependency on cloud services along with potential risk and impact, important topics need to be considered on behalf of cloud service customers (CSCs) and cloud service providers (CSPs). Topics include CSP trustworthiness, cloud services performance, cloud SLAs and compliance. No consistent cloud SLA structure or technique exists across CSPs for measuring, calculating, predicting, reporting and comparing SLA compliance of cloud services. The challenges intensify as CSCs establish multi-CSP and hybrid cloud strategies. When evaluating CSPs and SLA compliance for cloud services, a common system is required which considers multiple dimensions such as historical performance (cloud service and cloud SLA compliance), cloud security (required controls and regulatory compliance), cloud service characteristics (location and redundancy), CSP reputation, integrity and capabilities (can and do CSPs meet commitments), and CSC expectations (are they realistic and obtainable).
The uniqueness of this technology is the effective management of cloud computing service levels (e.g. availability, performance, security) and financial risk, both of which are dependent on the capability, performance and trustworthiness of CSPs. Cloud SLAs from standards bodies, cloud security requirements and compliance from Cloud Security Alliance, along with CSP performance (SLAs and cloud services) are analyzed via Graph Theory and Multi-Criteria Decision Analysis/Analytic Hierarchy Process to calculate CSP trustworthiness levels. CSP trustworthiness levels are input along with multiple factors (e.g. CSP SLA content and performance, cloud service characteristics and performance) into machine learning models to predict cloud service performance and cloud SLA compliance.
Applications:
1. Compare CSPs, related cloud services, and predict their performance.
2. Evaluate private vs public vs hybrid cloud services.
3. Manage relationships and tradeoffs between cloud SLA service levels and investment levels.
4. Identify cloud service improvement programs in support of cloud SLA service levels.
Advantages:
1. Identify opportunities to right size cloud service capabilities based on financial drivers, competitive market conditions, customer requirements (service level expectations).
2. Calculates and rank CSP trustworthiness levels based on multiple dimensions related to availability, performance, security and CSP capabilities.
3. Identify and manage risk, and set expectations by predicting performance of cloud services and cloud SLA compliance based on multiple factors (e.g. CSP trustworthiness, performance`).
4. Establish a defensible CSP and cloud computing strategy (e.g. public, private, hybrid) based on quantitative and qualitative statistical and machine learning models and analysis.
