A Formal Model for the Availability Analysis of Cloud Deployed Multi-tiered Applications

Abstract

High availability is a critical requirement for cloud deployed services. Cloud providers offer different availability zones with geo-redundancy to protect their infrastructure and consequently their tenants against failures and natural disasters. Nevertheless, different zones may have different reliability levels depending on the hardware equipment, the geo-location, the energy source powering the facility, etc. Hence, the ability to assess the expected availability of a given deployment is extremely important for both the cloud tenants and providers that are bound by a service level agreement. Due to the stochastic nature of failures, a formal stochastic model is needed to quantify the expected availability offered by an application deployment. This paper presents a Stochastic Petri Net model to evaluate the availability of cloud services and their deployment in geographically distributed data centers. The proposed Stochastic Petri Net model captures the characteristics of the cloud provider and user. It translates them into elements of an availability model that can be solved to calculate the expected availability and subsequently be used to guide the cloud scheduling solution.