Pattern based configuration generation for highly available COTS components based systems

Abstract

ContextService availability is an important quality factor that distinguishes between providers. High availability is achieved when the service is available at least 99.999% of the time. The Service Availability Forum has defined several middleware services, among them the Availability Management Framework (AMF) aims at ensuring the high availability of services provided by applications under its control. AMF requires a configuration. The design of an AMF configuration is complex and error prone. For improving the quality of such configurations, an automated generation technique has been proposed in the literature. However, this technique may generate several configurations among which some may not meet the required level of availability. Therefore, the generated configurations are evaluated using availability analysis tools to select an appropriate one for deployment. ObjectiveInstead of generating configurations, analyze them and select one that meets the required level of availability if any, we target directly configurations that are estimated to satisfy the required level of availability. MethodWe investigated the different aspects/attributes of an AMF configuration that may affect service recovery when a failure happens to come up with configuration design patterns and integrate them into an enhanced configuration generation technique. ResultsA set of configuration design patterns and two lightweight methods for availability estimation have been defined. These configuration design patterns and analysis methods are embedded into an enhanced configuration generation technique to target only the configurations that are estimated to satisfy the required level of availability or the configuration that is the closest to the required level of availability when it is not possible to meet the requirements. ConclusionWe conclude it is possible to target directly the best configurations in terms of estimated availability using configuration design patterns and estimation methods. The enhanced configuration generation technique is less resource and time consuming than related approaches.