Proteo Dataset (2023) for article "Proteo: A Framework for the Generation and Evaluation of Malleable MPI Applications"

<p>This dataset is the one used by the publication "Proteo: A Framework for the Generation and Evaluation of Malleable MPI Applications". The data are the data processed after the experiments to obtain the graphs and perform the analysis of the two applications shown.</p><p>On the one hand, there are data from the conjugate gradient (CG) execution and on the other hand, data from the Proteo framework, configured to emulate the CG.</p><p>The set is divided into 4 sections, each representing the results obtained for each result section of the article:</p><ul><li><strong>5.3</strong>: Similarity results between CG and Proteo configured to emulate CG. These results were obtained from the previous paper "Configurable synthetic application for studying malleability in HPC" at the PDP2023 conference. It consists of 4 datasheets:<ul><li><i>Nasp CG-VS-Proteo.xlsx</i>: Datasheet showing the total execution, iteration and computation times of Nasp (System_1) with the CG and Proteo. Used to create Figures 7 and 8.</li><li><i>JupiterThor CG-VS-Proteo.xlsx: </i>Datasheet showing the total execution, iteration and computation times of Jupiter (System_2) and Thor (System_3) with the CG and Proteus. Used to create Figures 7 and 8.</li><li><i>Comparison Comms CG Only.xlsx</i>: Datasheet showing MPI_Allgatherv communication times for different sizes on Nasp, Jupiter and Thor systems (System_1, System_2 and System_3). Used to create Table 1.</li><li><i>Comms CV_VS_Proteo.xlsx</i>: Datasheet showing MPI_Allgatherv communication times between the CG and Proteo for different sizes in Nasp, Jupiter and Thor systems (System_1, System_2 and System_3). Used to create Table 2.</li></ul></li><li><strong>5.4.1</strong>: Evaluation of the reconfiguration techniques in isolation of Proteo for synchronous methods. Used to create Figure 9. It is divided into three files:<ul><li><i>dataG.pkl</i>: Contains data on the complete runtimes in Proteo. The description can be found in the file <i>Proteo_dataG_description.txt</i></li><li><i>dataM.pkl</i>: Contains data on reconfiguration times in Proteo. The description can be found in the file <i>Proteo_dataM_description.txt</i></li><li><i>dataL.pkl</i>: Contains data on iteration times in Proteo. The description can be found in the file <i>Proteo_dataL_description.txt</i></li></ul></li><li><strong>5.4.2</strong>: Evaluation of the reconfiguration in a malleable application, CG against Proteo. Used to create Figure 10 and 11. It is divided into two directories, differentiating between CG and Proteus. In total there are the following 5 files:<ul><li><i>dataG.pkl</i>: Contains data on the complete runtimes in Proteo. The description can be found in the file <i>Proteo_dataG_description.txt</i></li><li><i>dataM.pkl</i>: Contains data on reconfiguration times in Proteo. The description can be found in the file <i>Proteo_dataM_description.txt</i></li><li><i>dataL.pkl</i>: Contains data on iteration times in Proteo. The description can be found in the file <i>Proteo_dataL_description.txt</i></li><li><i>dataCG_G.pkl</i>: Contains data on the complete execution times in the CG. The description can be found in the CG<i>_dataG_description.txt</i></li><li><i>dataCG_M.pkl</i>: Contains data on reconfiguration times in the CG. The description can be found in the CG<i>_dataM_description.txt</i></li></ul></li><li><strong>5.4.3</strong>: Evaluation of a malleable emulated application, CG against Proteo. Used to create Figure 12, 13 and 14. It is divided into two directories, differentiating between CG and Proteus. In total there are the following 5 files:<ul><li><i>dataG.pkl</i>: Contains data on the complete runtimes in Proteo. The description can be found in the file <i>Proteo_dataG_description.txt</i></li><li><i>dataM.pkl</i>: Contains data on reconfiguration times in Proteo. The description can be found in the file <i>Proteo_dataM_description.txt</i></li><li><i>dataL.pkl</i>: Contains data on iteration times in Proteo. The description can be found in the file <i>Proteo_dataL_description.txt</i></li><li><i>dataCG_G.pkl</i>: Contains data on the complete execution times in the CG. The description can be found in the CG<i>_dataG_description.txt</i></li><li><i>dataCG_M.pkl</i>: Contains data on reconfiguration times in the CG. The description can be found in the CG<i>_dataM_description.txt</i></li></ul></li></ul>