A Link Layer Protocol for Quantum Networks

<h1>
QLinkLayer simulation data
</h1>
<p>
The .zip-files in this DataVerse contains data used in the paper:
A Link Layer Protocol for Quantum Networks (https://arxiv.org/abs/1903.09778)
</p>

<h2>
General info
</h2>
<p>
All the raw-data can be found as sqlite3 (.db) files.
Physical parameters used in the simulations can be found in each .zip file in the folder config.
Input parameters for the simulation runs are in the files paramcombinations.json.
</p>

<h2>
Reading the data
</h2>
<p>
If you just want to have a look at the performance metrics such as fidelity, latency, throughput, number of generated pairs etc, there are genereted excel files for all the simulation runs. In each .zip-folder there is a file metrics/metrics.xlsx with multiple sheets containing this data.
The script that generates the excel file from the raw data can be found at <a href="https://github.com/SoftwareQuTech/QLinkLayer/blob/Develop/simulations/generate_metrics_file.py">generate_metrics_file.py</a>.
Scripts that generated plots and tables used in the paper can be found in the folder <a href="https://github.com/SoftwareQuTech/QLinkLayer/tree/Develop/simulations/data_analysis_scripts">data_analysis_scripts</a>
</p>

<h2>
Other notes
</h2>
<p>
<ul>
<li> The fidelity estimation unit used in these simulations was very simplistic and therefore not correct. It did not do any estimation based on measurement outcomes but only using a simple model for what the generated states are based on bright state populations, detection efficiencies and dark counts, not taking gate and measurement noise into account.</li>

<li> These simulations used a rather pessimistic coherence time for the electron (T1=2.68 ms, T2=1 ms) and nuclear spin (T1=inf, T2=3.5 ms), whereas coherence times of seconds have been experimentally verified using dynamical decoupling.</li>

<li> Only one of the carbons (nr 1) was used, so whenever an type K OK was returned to the higher layer it was consumed and measured such that the same carbon could be used again. Thus multiple entangled pairs were never stored at the same time. </li>

<li> During the earlier of these simulations we noticed a bug with the distributed queue in some rare cases, which amounted to some warnings in the logs. This is however now fixed in the latest version of the repo at <a href="https://github.com/SoftwareQuTech/QLinkLayer">QLinkLayer</a>.</li>

<li> To more efficiently estimate QBER, the measurement-bases used for type M requests were iterated over (X, Y, Z) between the nodes such that the two nodes always measured in the same basis. The nodes choose the measurement-basis based on the current MHP-cycle modulo 3. However, this is not what a real implementation of a link layer protocol would do, see specification of the fields RBC, ROTX1, ROTY and ROTX2 in the CREATE header specified in
<a href="https://datatracker.ietf.org/doc/draft-dahlberg-ll-quantum/">draft-dahlberg-ll-quantum</a>.</li>
</ul>
</p>

<h2>
Data
</h2>
<p>
The DataVerse contains the following zip files:
<ul>
<li> Major simulation (run1) (used in the paper)
<li> 2019-01-16T11:10:28CET_major_simulation.zip </li>
</ul>
</li>

<li> Major simulation (run2)
<ul>
<li> 2019-01-15T23:56:55CET_major_simulation.zip </li>
</ul>
</li>

<li> High loss run (used in the paper)
<ul>
<li> 2019-01-28T16:49:28CET_major_simulation.zip </li>
</ul>
</li>

<li> Request frequency sweep (used in the paper)
<ul>
<li> 2019-01-24T19:59:34CET_req_freq_sweep.zip </li>
<li> 2019-01-26T15:59:44CET_req_freq_sweep.zip </li>
<li> 2019-01-26T16:23:25CET_req_freq_sweep.zip </li>
</ul>
</li>

<li> Fidelity sweep (used in the paper)
<ul>
<li> 2019-01-28T11:32:51CET_fidelity_sweep.zip </li>
<li> 2019-02-25T19:45:16UTC_fidelity_sweep.zip </li>
<li> 2019-02-27T17:14:02UTC_fidelity_sweep.zip </li>
</ul>
</li>
</ul>
</p>