CS-1 TAS: A DASA Evaluation of the Tele Assistance System

Dimensional Analysis for Software Architecture (DASA) evaluation of the Tele Assistance System (TAS), a service-based self-adaptive exemplar for chronic-care home monitoring with a centralised MAPE-K loop (Weyns and Calinescu, SEAMS 2015).

This repository is a reproducible deliverable, not a library: it produces the metric datasets and figures that ground a DASA case study. It is Phase 1, model-only (analytic + stochastic + dimensional prediction, plus a constructive search), consuming the PyDASA library.

Documentation

The docs/ folder holds the case study in three parts, read in order:

1. docs/case-study.md - what TAS is: the architecture, the analyse-and-act workflow, the ReSeP service catalogue, the quality scenarios, and the S1 / S2 adaptation strategies (the source-faithful reconstruction).
2. docs/procedure.md - how the experiment runs: the DASA method, the hypotheses under test, and the idea behind each of the six notebooks.
3. docs/report.md - what it found: the DASA modelling derivation, the per-method verdicts, the performance-versus-availability trade-off, the constructive search, threats to validity, and research-question coverage.

What it decides

TAS is modelled as a 14-node loss-network queue (13 service stations plus one absorbing FAIL sink); availability is read from the flow leaked to the sink. Two quality-attribute requirements are decided across four adaptation strategies, framed as a performance-versus-availability trade-off:

Requirement	Metric	Threshold	Source
R1 (Availability)	end-to-end failure rate	<= 1.0 %	Weyns 2015
R2 (Performance)	end-to-end response time	<= 26 ms	Camara 2023

(An original cost requirement, R3, is out of scope for this study.)

Adaptation	Failure rate	R1	Response time	R2
baseline (no adaptation)	14.82 %	FAIL	25.74 ms	PASS
S1 (Retry)	0.20 %	PASS	27.72 ms	FAIL
S2 (Select-Reliable)	10.42 %	FAIL	18.53 ms	PASS
aggregate (S1 & S2)	0.07 %	PASS	19.40 ms	PASS
search winner	0.03 %	PASS	19.85 ms	PASS

Only the aggregate passes both among the hand-authored adaptations. A DASA-coefficient-guided search constructs a configuration that beats it and is cross-validated by an independent scipy optimiser.

Methods

Three independent predictive pipelines compute the same verdicts, and a fourth notebook runs the constructive search:

Notebook	Method	What it does
01-analytic.ipynb	analytic	closed-form M/M/c/K Jackson solve (loss network + FAIL sink)
02-stochastic.ipynb	stochastic	SimPy discrete-event simulation of the same network
03-dimensional.ipynb	dimensional	PyDASA Pi-groups + dimensionless viable-region bounds
04-yoly.ipynb	dimensional	design-space sweep across the (mu, c, K) grid
05-search.ipynb	search	coefficient-guided descent + scipy cross-validation
06-comparison.ipynb	comparison	cross-method verdict matrix + numerical agreement

Setup

Requires Python 3.12+.

python -m venv venv
source venv/Scripts/activate         # Git Bash on Windows; use venv/bin/activate on Linux/macOS

# 1. Install PyDASA (not on public PyPI) from source:
pip install "git+https://github.com/DASA-Design/PyDASA.git@v0.7.1"

# 2. Install the rest:
pip install -r requirements.txt

If the PyDASA tag or location differs, adjust the first command; the pin in requirements.txt (pydasa==0.7.1) documents the expected version.

Running

Each method is a module with a small CLI. Adaptation is one of baseline | s1 | s2 | aggregate:

python -m src.methods.analytic     --adaptation aggregate
python -m src.methods.stochastic   --adaptation aggregate
python -m src.methods.dimensional  --adaptation aggregate

Or run the notebooks end-to-end (this regenerates every dataset and figure):

jupyter nbconvert --to notebook --execute --inplace 0*.ipynb

Layout

├── 01-analytic … 06-comparison.ipynb   # the pipeline, in order
├── src/
│   ├── methods/       # per-method orchestrators (run() + CLI)
│   ├── analytic/      # closed-form queueing-network solvers
│   ├── stochastic/    # SimPy DES processes
│   ├── dimensional/   # PyDASA schema, coefficients, routing, search
│   ├── view/          # plotting helpers
│   ├── io/            # config loaders
│   └── utils/         # shared helpers
├── data/
│   ├── config/        # single source of truth for parameters (PyDASA Variable schema)
│   ├── results/       # per-method, per-adaptation metric JSONs + verdicts
│   └── reference/     # ground-truth thresholds
├── docs/              # case study, procedure, report + figures
└── tests/             # pytest, mirrors src/

The method-result figures are included under data/img/ (analytic, stochastic, dimensional, comparison, search), as both PNG and SVG. They are regenerated in place when the notebooks or methods run.

Data convention

data/config/ is the single source of truth. Every config uses the PyDASA Variable-dict schema keyed by LaTeX symbol. Methods read a profile plus an adaptation and write a single JSON per run to data/results/<method>/<adaptation>/, plus a profile-agnostic requirements.json carrying the R1 / R2 verdicts. Files under data/results/ are regenerated by the methods; do not hand-edit them.

Tests

pytest tests/ -q

License

GPL-3.0. See LICENSE.