<h1 id="fase-2023-artifact">FASE 2023 Artifact</h1>
<p>NOTE: this README is available also in HTML or PDF format.</p>
<h2 id="summary">Summary</h2>
<p>This artifact contains a prototype of a runtime monitoring middleware
called VAMOS as discussed in the FASE 2023 paper <em>VAMOS: Middleware
for Best-Effort Third-Party Monitoring</em> by Marek Chalupa, Stefanie
Muroya Lei, Fabian Muehlboeck, and Thomas A. Henzinger. The artifact
consist of the software, benchmarks discussed in the paper, and a set of
scripts to run experiments and present the results.</p>
<h2 id="hardware-requirements">Hardware Requirements</h2>
<p>For all benchmarks to make sense, the artifact requires a reasonably
modern (~past 5? years) x86-64 processor with multiple cores (ideally at
least 6 cores).</p>
<h2 id="using-a-bundled-docker-image">Using a bundled docker image</h2>
<p>You can load the bundled docker image with <code>docker load</code>
command and run it:</p>
<pre><code>docker load &lt; vamos-fase2023.tar.gz
docker run -ti -v &quot;$(pwd)/results&quot;:/opt/results vamos:fase</code></pre>
<p>Note that the image was build for x86 architectures. If your
architecture is different, you either need to specify
<code>--platform</code> when running/creating the container to emulate
your architecture or build the image yourself. Note that emulating the
architecture brings non-negligible overhead and so building the image
may be a better option.</p>
<h2 id="building-the-artifact">Building the artifact</h2>
<p>To build and run the artifact run this command from the top-level
directory:</p>
<pre><code>docker build . -t vamos:fase</code></pre>
<p>If you are on a new enough Linux system, you may use this command to
get faster builds:</p>
<pre><code>DOCKER_BUILDKIT=1 docker build . -t vamos:fase</code></pre>
<p>The building process can take more than 10 minutes, based on the used
machine and the speed of the internet connection.</p>
<p>To run the built image, use:</p>
<pre><code>docker run -ti -v &quot;$(pwd)/results&quot;:/opt/results vamos:fase</code></pre>
<p>This command starts a docker container with experiments ready to run
and gives you a shell in this container. It also creates a folder
<code>results/</code> in the current directory to the container where
will the results of the experiments appear. Feel free to change
<code>$(pwd)/results</code> to a directory of your choice.</p>
<p>Once in the docker container, continue with the test instructions
below.</p>
<h2 id="test-instructions">Test Instructions</h2>
<p>We have prepared script to run three versions of experiments: short,
mid, and full experiments. Short experiments run just a few minutes, but
the results may diverge from the results in the paper. Mid experiments
run longer but still in the order of tens of minutes. Full experiments
can run for several hours and should reproduce numbers from the paper.
Note that running experiments will overwrite results of previous runs of
experiments.</p>
<p>You can run short/mid/full experiments with these scripts:</p>
<pre><code>./short-experiments.sh
./mid-experiments.sh
./full-experiments.sh</code></pre>
<p>A guide on how to run just some experiments is at the end of this
document.</p>
<p>Once experiments are finished, plots in the PDF format should be
automatically generated and put in the results folder from where the
docker was started (or in a folder of your choice if you changed the
command). The plots are named after the figures in the paper.</p>
<p>You can also generate the plots manually by running make in the
<code>plots</code> directory. Moreover, you can generate the plots from
the original data that we measured. The data are in
<code>plots/original-data</code> and you generate the plots by</p>
<pre><code>cd plots
make original</code></pre>
<p>For more details about plots, check the README.md inside
<code>plots</code> directory.</p>
<h2 id="running-just-some-experiments">Running just some
experiments</h2>
<p>To run just some experiments, you can comment out lines with
experiments in <code>{short,full}-experiments.sh</code> scripts and
rerun these scripts or you can follow what these script do:</p>
<p>First, pick if you want short of full experiments and according to
that, copy <code>scripts/setup-short.sh</code> or
<code>scripts/setup-full.sh</code> into <code>setup.sh</code> (we are in
the top-level directory now):</p>
<pre><code>cp scripts/setup-short.sh setup.sh   # short experiments
cp scripts/setup-full.sh setup.sh   # full experiments</code></pre>
<p>Then, go into directories with experiments and run
<code>make experiments</code> in the directory, for example:</p>
<pre><code>cd scalability
make experiments</code></pre>
<p>Available experiment directories are <code>scalability</code>,
<code>primes</code>, <code>bank</code>, <code>bank-tessla</code>, and
<code>dataraces</code>. The only exception from this pattern are
primes-tessla experiments that are run from the <code>primes</code>
directory using <code>make experiments-tessla</code>:</p>
<pre><code>cd primes
make experiments-tessla</code></pre>