Oslo meeting (23, 24 Sept. 2014)


Venue: SINTEF ICT, Oslo

Tuesday Sept, 23: experimentation

9h00 – 9h45: introduction / what’s new (all)
9h45 – 12h00: status of the case study (TCD, all). Status of current developments, demo, plan for future developments, hands-on.
13h30 – 14h15: Constrained evolution and its cost related to architecture diversity (SINTEF)
14h15 – 15h00: Algorithm diversification (TCD)
15h00 – 15h30: Status about sosiefication (INRIA)
15h30 – 17h30: plan for experiments with the smart city application (all)
17h30: dinner

Wednesday Sept, 24: Ecology

9h30 – 10h30: clonal plants and food webs plasticity (UR1)
10h30 – 11h15: bipartite graph simulation (SINTEF)
11h15 – 12h00: evolutionnary diversification (INRIA)
13h30 – 14h30: extinction sequences as a robustness indicator for Cloud topologies (SINTEF)
14h30 – 17h30: plans for bipartite graphs and collaborations with ecology

Constrained evolution and its cost related to architecture diversity (SINTEF)

In this talk, we will continue the presentation of using constraint solving on self-adaptation. We will talk about how to model an architecture adaptation problem as a Satisfaction Modulo Theory (SMT) problem, with both hard and soft constraints, and also show how to use constraint solving to automatically calculate adaptation results. We will go over our work in this summer to use this idea to “diversify” the system’s architecture. Furthermore, we will also report an initial experiment we are doing in order to see what benefit we can get from the “diversified” architecture. The idea behind this experiment is that a more diverse system, with spare functionalities and configuration elements, should be “cheaper” to be adapted, in average, when the context is changed. Here “cheap” means less changes on the current system architecture. We also use constraint solving to calculate the minimal cost of adaptation on a particular system configuration: we introduce a set of soft constraints which require every configuration to be unchanged after adaptation, and the cost of adaptation is the total weights of the violated constraints in the best result found by the constraint solver.

Algorithm diversification (TCD)

We are trying to investigate mechanisms and effects of diversifying the
underlying algorithms, in a running piece of software. Most algorithms
have some assumptions about the nature of their input space, and in
dynamic real-world systems, these assumptions may sometimes be invalid.
Diversifying/changing the algorithm that a working software uses, has interesting
effects on various non-functional attributes of the software. The idea
is to develop a systematic means of diversification such that the effect
on some selected non-functional attribute is predictably positive.

In this particular presentation, we show that there is a measurable
effect of genetic/algorithmic diversification, and we would like to
discuss means of measuring the diversity so achieved

Status about sosiefication (INRIA)

The latest works on sosies have focused on the analysis of the sosies and the correctness envelops that define the scope for synthesis. We have developed a set of metrics and associated static and dynamic analysis procedures to have a more precise idea about the extent of the correctness envelop. We have also analyzed a number of sosies manually to classify the different types of transformations that lead to diversity. Finally, we’ll present initial results on multi-point sosies, which increase the diversity w.r.t the original program.