Optimizing decomposition of software architecture for local recovery

Abstract

The increasing size and complexity of software systems has led to an amplified number of potential failures and as such makes it harder to ensure software reliability. Since it is usually hard to prevent all the failures, fault tolerance techniques have become more important. An essential element of fault tolerance is the recovery from failures. Local recovery is an effective approach whereby only the erroneous parts of the system are recovered while the other parts remain available. For achieving local recovery, the architecture needs to be decomposed into separate units that can be recovered in isolation. Usually, there are many different alternative ways to decompose the system into recoverable units. It appears that each of these decomposition alternatives performs differently with respect to availability and performance metrics. We propose a systematic approach dedicated to optimizing the decomposition of software architecture for local recovery. The approach provides systematic guidelines to depict the design space of the possible decomposition alternatives, to reduce the design space with respect to domain and stakeholder constraints and to balance the feasible alternatives with respect to availability and performance. The approach is supported by an integrated set of tools and illustrated for the open-source MPlayer software.

Safety Perspective for Supporting Architectural Design of Safety-Critical Systems

“…Addressing quality concerns at the architecture view level has been actually based on either defining a new viewpoint [2] or using architecture perspectives [7], each with their own merits. In our earlier work we have considered the explicit modeling of viewpoints for quality concerns [9][10] [11]. Unfortunately, so far no architectural perspective has been defined for the safety concern.…”

Availability analysis of software architecture decomposition alternatives for local recovery

“…Previously, we introduced Recovery Designer (Sozer et al 2013) for automatically finding the optimal decomposition of software architecture for local recovery. This tool addresses a set of system constraints and balances the achieved availability and performance overhead for decomposition alternatives.…”

“…The availability of the system, that is, the percentage of time the system is up and running (Avizienis et al 2004), heavily depends on the chosen software decomposition alternative, i.e., the way in which the software modules are grouped into RUs (Sozer et al 2013). We have previously introduced a framework, called FLORA ) that supports the implementation of local recovery for a particular software decomposition.…”

“…Despite the support of this framework, the implementation of local recovery for a decomposition alternative is still a time-consuming and a non-trivial task ), Hence, it is important to follow a systematic method to compare various alternatives and only implement the best one. We previously developed a tool named Recovery Designer (Sozer et al 2013;Sozer 2009), which optimizes software architecture decomposition for local recovery based on quality estimations for decomposition alternatives. However, optimization needs input from a systematic and sound method that provides availability estimations regarding decomposition alternatives at design time.…”

Architecture Framework for Software Safety

“…In our earlier work, we have considered the explicit modeling of viewpoints for quality concerns [13] [15]. Hereby, each quality concern, such as adaptability and recoverability, require a different decomposition of the architecture.…”