Task Scheduler Performance Survey Results

<strong>Task scheduler performance survey</strong> This dataset contains results of task graph scheduler performance survey.<br> The results are stored in the following files, which correspond to simulations performed on <br> the `elementary`, `irw` and `pegasus` task graph datasets published at https://doi.org/10.5281/zenodo.2630384. elementary-result.zip irw-result.zip pegasus-result.zip The files contain compressed pandas dataframes in CSV format, it can be read with the following Python code:<br> ```python<br> import pandas as pd<br> frame = pd.read_csv("elementary-result.zip")<br> ``` Each row in the frame corresponds to a single instance of a task graph that<br> was simulated with a specific configuration (network model, scheduler etc.).<br> The list below summarizes the meaning of the individual columns. <strong>graph_name</strong> - name of the benchmarked task graph <strong>graph_set</strong> - name of the task graph dataset from which the graph originates <strong>graph_id</strong> - unique ID of the graph <strong>cluster_name</strong> - type of cluster used in this instance the format is &lt;number-of-workers&gt;x&lt;number-of-cores&gt;; 32x16 means 32 workers, each with 16 cores <strong>bandwidth</strong> - network bandwidth [MiB] <strong>netmodel</strong> - network model (simple or maxmin) <strong>scheduler_name</strong> - name of the scheduler <strong>imode</strong> - information mode <strong>min_sched_interval</strong> - minimal scheduling delay [s] <strong>sched_time</strong> - duration of each scheduler invocation [s] <strong>time</strong> - simulated makespan of the task graph execution [s] <strong>execution_time </strong>- real duration of all scheduler invocations [s] <strong>total_transfer</strong> - amount of data transferred amongst workers [MiB] The file `charts.zip` contains charts obtained by processing the datasets.<br> On the X axis there is always bandwidth in [MiB/s].<br> There are the following files: [DATASET]-schedulers-time - Absolute makespan produced by schedulers [seconds] [DATASET]-schedulers-score - The same as above but normalized with respect to the best schedule (shortest makespan) for the given configuration. [DATASET]-schedulers-transfer - Sums of transfers between all workers for a given configuration [MiB] [DATASET]-[CLUSTER]-netmodel-time - Comparison of netmodels, absolute times [seconds] [DATASET]-[CLUSTER]-netmodel-score - Comparison of netmodels, normalized to the average of model "simple" [DATASET]-[CLUSTER]-netmodel-transfer - Comparison of netmodels, sum of transfered data between all workers [MiB] [DATASET]-[CLUSTER]-schedtime-time - Comparison of MSD, absolute times [seconds] [DATASET]-[CLUSTER]-schedtime-score - Comparison of MSD, normalized to the average of "MSD=0.0" case [DATASET]-[CLUSTER]-imode-time - Comparison of Imodes, absolute times [seconds] [DATASET]-[CLUSTER]-imode-score - Comparison of Imodes, normalized to the average of "exact" imode <strong>Reproducing the results</strong> <em>1. Download and install Estee (https://github.com/It4innovations/estee)</em> $ git clone https://github.com/It4innovations/estee<br> $ cd estee<br> $ pip install .<br> <br> <em>2. Generate task graphs</em><br> You can either use the provided script `benchmarks/generate.py` to generate graphs<br> from three categories (elementary, irw and pegasus): $ cd benchmarks<br> $ python generate.py elementary.zip elementary<br> $ python generate.py irw.zip irw<br> $ python generate.py pegasus.zip pegasus<br> <br> or use our task graph dataset that is provided at https://doi.org/10.5281/zenodo.2630384. <em>3. Run benchmarks</em><br> To run a benchmark suite, you should prepare a JSON file describing the benchmark.<br> The file that was used to run experiments from the paper is provided in<br> `benchmark.json`. Then you can run the benchmark using this command:<br> <br> $ python pbs.py compute benchmark.json The benchmark script can be interrupted at any time (for example using Ctrl+C).<br> When interrupted, it will store the computed results to the result file and restore<br> the computation when launched again. <em>3. Visualizing results</em><br> <br> $ python view.py --all &lt;result-file&gt;<br> <br> The resulting plots will appear in a folder called `outputs`.