CloudAQUA: An Analytic Research on Quantifying Availability (AQUA) for Cloud Resource Provisioning and Allocation

Research Thrusts, Novelty and Expected Contributions

Research Thrusts

• Thrust1 - Application-layer and middleware related issues
  • Service characterization through detailed VM-level instrumentation and process-level measurements on CPU, memory, I/O and network usage
  • VM allocation and placement algorithms
• Thrust2 - Availability prediction models and solution techniques
• Thrust3 - Lower layer issues
  • Collection and analysis of (temporally and spatially correlated) failure data on physical components of the UB-CCR and IU-FG-Cloud
  • Failure modeling
• Thrust4 - Conducting controlled experiments on UB-CCR and IU-FG-Cloud

Novelty and Intellectual Merits

• Rigorous quantification and prediction of availability for a finite duration
• Availability-aware multi-objective optimization
• Experiments on production and testbed systems

Expected Broader Impact and Contributions

• Significant advancement of the state-of-the-art by (a) offering deep insights into factors affecting the availability in cloud computing, (b) developing and disseminating useful data sets, analytic models and computational tools for cost-effective resource provisioning and allocation methods (middleware and algorithms) that will be useful in achieving a high availability levels.
• Development of new tools to collect the following data and provide public access through an integrated secure web interface:
  • Large set of fine-grained failure data on physical components such as the servers and switches/routers and another large set of data
  • Large set of data on workload characteristics of services and performance/usage (e.g., CPU, memory, I/O, Network) at the VM Manager (VMM) level
• Fast and accurate estimation of VM availabilities which can also be useful to solve other such complex and similar models in different research disciplines
• Improvement on Service Level Agreement (SLA) guarantees on service availability
• Stimulate and enable marketing and economic research on constructing service level agreements (SLA) with feasible and acceptable VM price levels and availability guarantees
• Improvement of availability and reliability of cloud services that lead to (a) greater benefits to both cloud service providers and users, (b) eliminate/minimize the millions of dollars of losses in revenue and productivity due to service outages, and (c) avoid energy wastages due to interrupted jobs. More importantly, it will also encourage new users and applications by generating further cost savings and increased productivity to customers and increased revenues to service providers, which in turn positively impact our economy