MDDPro: Model-Driven Dependability Provisioning in Enterprise Distributed Real-Time and Embedded Systems

Abstract

Service oriented architecture (SOA) design principles are increasingly being adopted to develop distributed real-time and embedded (DRE) systems, such as avionics mission computing, due to the availability of real-time component middleware platforms. Traditional approaches to fault tolerance that rely on replication and recovery of a single server or a single host do not work in this paradigm since the fault management schemes must now account for the timely and simultaneous failover of groups of entities while improving system availability by minimizing the risk of simultaneous failures of replicated entities. This paper describes MDDPro, a model-driven dependability provisioning tool for DRE systems. MDDPro provides intuitive modeling abstractions to specify failover requirements of DRE systems at different granularities. MDDPro enables plugging in different replica placement algorithms to improve system availability. Finally, its generative capabilities automate the deployment and configuration of the DRE system on the underlying platforms.

Supporting component-based failover units in middleware for distributed real-time and embedded systems

“…This section describes how the structure and functionality of CORFU provides DRE systems with passive replication of a group of components treated as a logical failover unit [13], which contains a set of components that have dependencies with respect to failure propagation. Failover units allow for failure reaction times suitable for DRE systems by terminating a group of components with a single logical execution step, rather than reacting on slow failure propagations.…”

“…Since CORFU is geared to integrate with and leverage the analysis of model-based frameworks (such as MDDPro [13]) we compare it with related modeling efforts.…”

MoPED: A Model-Based Provisioning Engine for Dependability in Component-Based Distributed Real-Time Embedded Systems

“…Unlike the traditional client/server model of designing distributed systems, component-based systems often have more than one component arranged in a workflow-like pattern (assembly) to realize critical application functionality. To simultaneously meet the predictability and availability requirements of the end-to-end application workflows, group-failover [23] protocol has been developed. In the event of a failure, group-failover protocol allows the clients to failover to a replica assembly to maintain the state consistency and timeliness of application data.…”

“…While doing so, it also automatically generates complex connection topology interconnecting the generated components, which is dictated by the replication degree of the primary component and replication degree of components that it interacts with. A detailed discussion of the model transformation algorithms that generate the replica component topologies and complex interconnections between them are described in [24], [26].…”

Component Replication Based on Failover Units

“…This section describes the structure and functionality of CORFU, focusing on how it provides DRE systems with passive replication of a group of components treated as a logical failover unit [13]. A failover unit contains a set of components have dependencies with respect to failure propagation.…”