Constraining results to preliminary results only

Schrick-Noah_MPI-Tasking.aux

@@ -114,10 +114,10 @@
 \newlabel{fig:overall-efficiency}{{13}{9}{Minimum, Maximum, and Mean Efficiencies of MPI Tasking Across All Problem Sizes}{figure.13}{}}
 \@writefile{lof}{\contentsline {figure}{\numberline {14}{\ignorespaces Mean Speedup and Efficiency for the Exploit Parameter Across the Number of Compute Nodes}}{10}{figure.14}\protected@file@percent }
 \newlabel{fig:param-exploit}{{14}{10}{Mean Speedup and Efficiency for the Exploit Parameter Across the Number of Compute Nodes}{figure.14}{}}
-\@writefile{toc}{\contentsline {section}{\numberline {VIII}Conclusion and Future Work}{10}{section.8}\protected@file@percent }
-\newlabel{sec:FW}{{VIII}{10}{Conclusion and Future Work}{section.8}{}}
 \@writefile{lof}{\contentsline {figure}{\numberline {15}{\ignorespaces Mean Speedup and Efficiency for the Applicability of Exploit Parameter Across the Number of Compute Nodes}}{10}{figure.15}\protected@file@percent }
 \newlabel{fig:param-appl}{{15}{10}{Mean Speedup and Efficiency for the Applicability of Exploit Parameter Across the Number of Compute Nodes}{figure.15}{}}
+\@writefile{toc}{\contentsline {section}{\numberline {VIII}Conclusion and Future Work}{10}{section.8}\protected@file@percent }
+\newlabel{sec:FW}{{VIII}{10}{Conclusion and Future Work}{section.8}{}}
 \citation{Amdahl}
 \citation{Gust}
 \citation{sun}
@@ -147,13 +147,11 @@
 \bibcite{li_concurrency_2019}{23}
 \bibcite{9150145}{24}
 \bibcite{7087377}{25}
-\@writefile{lof}{\contentsline {figure}{\numberline {16}{\ignorespaces Mean Speedup and Efficiency for the Database Load Parameter Across the Number of Compute Nodes}}{11}{figure.16}\protected@file@percent }
-\newlabel{fig:param-load}{{16}{11}{Mean Speedup and Efficiency for the Database Load Parameter Across the Number of Compute Nodes}{figure.16}{}}
-\@writefile{toc}{\contentsline {section}{References}{11}{section*.1}\protected@file@percent }
 \bibcite{li_combining_2019}{26}
 \bibcite{Slurm}{27}
 \bibcite{Amdahl}{28}
 \bibcite{Gust}{29}
 \bibcite{sun}{30}
 \bibstyle{ieeetr}
-\gdef \@abspage@last{12}
+\@writefile{toc}{\contentsline {section}{References}{11}{section*.1}\protected@file@percent }
+\gdef \@abspage@last{11}

Schrick-Noah_MPI-Tasking.log

@@ -1,4 +1,4 @@
-This is pdfTeX, Version 3.141592653-2.6-1.40.25 (TeX Live 2023/Arch Linux) (preloaded format=pdflatex 2023.4.3)  22 APR 2023 17:54
+This is pdfTeX, Version 3.141592653-2.6-1.40.25 (TeX Live 2023/Arch Linux) (preloaded format=pdflatex 2023.9.6)  10 SEP 2023 15:27
 entering extended mode
  restricted \write18 enabled.
  %&-line parsing enabled.
@@ -240,7 +240,7 @@ File: pdftex.def 2022/09/22 v1.2b Graphics/color driver for pdftex
 Package: babel 2023/02/13 3.86 The Babel package
 \babel@savecnt=\count282
 \U@D=\dimen175
-\l@unhyphenated=\language87
+\l@unhyphenated=\language3
 
 (/usr/share/texmf-dist/tex/generic/babel/txtbabel.def)
 \bbl@readstream=\read2
@@ -250,7 +250,7 @@ Package babel Info: You haven't specified a language as a class or package
 
 (/usr/share/texmf-dist/tex/generic/babel/nil.ldf
 Language: nil 2023/02/13 3.86 Nil language
-\l@nil=\language88
+\l@nil=\language4
 ))
 (/usr/share/texmf-dist/tex/latex/base/textcomp.sty
 Package: textcomp 2020/02/02 v2.0n Standard LaTeX package
@@ -611,84 +611,111 @@ Underfull \hbox (badness 7451) in paragraph at lines 299--300
  []
 
 [7 <./images/Applicable_E.png>]
+Underfull \hbox (badness 1675) in paragraph at lines 305--306
+\OT1/ptm/m/n/10 benchmark suite. The preliminary results gathered were
+ []
+
 <./images/para_coords.png, id=250, 1468.48625pt x 451.6875pt>
 File: ./images/para_coords.png Graphic file (type png)
 <use ./images/para_coords.png>
-Package pdftex.def Info: ./images/para_coords.png  used on input line 309.
+Package pdftex.def Info: ./images/para_coords.png  used on input line 311.
 (pdftex.def)             Requested size: 252.0pt x 77.50954pt.
 <./images/nodes-runtime.png, id=251, 431.4519pt x 330.9966pt>
 File: ./images/nodes-runtime.png Graphic file (type png)
 <use ./images/nodes-runtime.png>
-Package pdftex.def Info: ./images/nodes-runtime.png  used on input line 317.
+Package pdftex.def Info: ./images/nodes-runtime.png  used on input line 319.
 (pdftex.def)             Requested size: 252.0pt x 193.33171pt.
 <./images/exploits-runtime.png, id=252, 409.7709pt x 350.5095pt>
 File: ./images/exploits-runtime.png Graphic file (type png)
 <use ./images/exploits-runtime.png>
-Package pdftex.def Info: ./images/exploits-runtime.png  used on input line 318.
+Package pdftex.def Info: ./images/exploits-runtime.png  used on input line 320.
 
 (pdftex.def)             Requested size: 252.0pt x 215.55405pt.
 <./images/applicability-runtime.png, id=253, 431.4519pt x 337.5009pt>
 File: ./images/applicability-runtime.png Graphic file (type png)
 <use ./images/applicability-runtime.png>
 Package pdftex.def Info: ./images/applicability-runtime.png  used on input line
- 325.
+ 327.
 (pdftex.def)             Requested size: 252.0pt x 197.13081pt.
 <./images/dbload-runtime.png, id=254, 431.4519pt x 337.5009pt>
 File: ./images/dbload-runtime.png Graphic file (type png)
 <use ./images/dbload-runtime.png>
-Package pdftex.def Info: ./images/dbload-runtime.png  used on input line 326.
+Package pdftex.def Info: ./images/dbload-runtime.png  used on input line 328.
 (pdftex.def)             Requested size: 252.0pt x 197.13081pt.
 <./images/overall-speedup.png, id=258, 409.7709pt x 346.1733pt>
 File: ./images/overall-speedup.png Graphic file (type png)
 <use ./images/overall-speedup.png>
-Package pdftex.def Info: ./images/overall-speedup.png  used on input line 337.
+Package pdftex.def Info: ./images/overall-speedup.png  used on input line 339.
 (pdftex.def)             Requested size: 252.0pt x 212.88737pt.
- [8 <./images/para_coords.png> <./images/nodes-runtime.png> <./images/exploits-
+[8 <./images/para_coords.png> <./images/nodes-runtime.png> <./images/exploits-r
-runtime.png>]
+untime.png>]
 <./images/overall-efficiency.png, id=274, 409.7709pt x 346.1733pt>
 File: ./images/overall-efficiency.png Graphic file (type png)
 <use ./images/overall-efficiency.png>
 Package pdftex.def Info: ./images/overall-efficiency.png  used on input line 34
-7.
+9.
 (pdftex.def)             Requested size: 252.0pt x 212.88737pt.
 
+
+LaTeX Warning: Reference `fig:param-load' on page 9 undefined on input line 355
+.
+
+
 Underfull \vbox (badness 3815) has occurred while \output is active []
 
  [9 <./images/applicability-runtime.png> <./images/dbload-runtime.png> <./image
 s/overall-speedup.png> <./images/overall-efficiency.png>]
-<./images/exploit-speedup.png, id=292, 414.8298pt x 344.0052pt>
+<./images/exploit-speedup.png, id=287, 419.8887pt x 344.0052pt>
 File: ./images/exploit-speedup.png Graphic file (type png)
 <use ./images/exploit-speedup.png>
-Package pdftex.def Info: ./images/exploit-speedup.png  used on input line 357.
+Package pdftex.def Info: ./images/exploit-speedup.png  used on input line 359.
-(pdftex.def)             Requested size: 252.0pt x 208.97151pt.
+(pdftex.def)             Requested size: 252.0pt x 206.46245pt.
-<./images/exploit-eff.png, id=293, 414.8298pt x 344.0052pt>
+<./images/exploit-eff.png, id=288, 419.8887pt x 344.0052pt>
 File: ./images/exploit-eff.png Graphic file (type png)
 <use ./images/exploit-eff.png>
-Package pdftex.def Info: ./images/exploit-eff.png  used on input line 358.
+Package pdftex.def Info: ./images/exploit-eff.png  used on input line 360.
-(pdftex.def)             Requested size: 252.0pt x 208.97151pt.
+(pdftex.def)             Requested size: 252.0pt x 206.46245pt.
-<./images/appl-speedup.png, id=294, 409.7709pt x 344.0052pt>
+<./images/appl-speedup.png, id=289, 409.7709pt x 344.0052pt>
 File: ./images/appl-speedup.png Graphic file (type png)
 <use ./images/appl-speedup.png>
-Package pdftex.def Info: ./images/appl-speedup.png  used on input line 365.
+Package pdftex.def Info: ./images/appl-speedup.png  used on input line 367.
 (pdftex.def)             Requested size: 252.0pt x 211.55406pt.
-<./images/appl-eff.png, id=295, 409.7709pt x 344.0052pt>
+<./images/appl-eff.png, id=290, 409.7709pt x 344.0052pt>
 File: ./images/appl-eff.png Graphic file (type png)
 <use ./images/appl-eff.png>
-Package pdftex.def Info: ./images/appl-eff.png  used on input line 366.
+Package pdftex.def Info: ./images/appl-eff.png  used on input line 368.
 (pdftex.def)             Requested size: 252.0pt x 211.55406pt.
-<./images/load-speedup.png, id=296, 409.7709pt x 344.0052pt>
+
-File: ./images/load-speedup.png Graphic file (type png)
+Underfull \vbox (badness 10000) has occurred while \output is active []
-<use ./images/load-speedup.png>
+
-Package pdftex.def Info: ./images/load-speedup.png  used on input line 373.
+
-(pdftex.def)             Requested size: 252.0pt x 211.55406pt.
+Underfull \hbox (badness 2941) in paragraph at lines 376--377
-<./images/load-eff.png, id=297, 409.7709pt x 344.0052pt>
+\OT1/ptm/m/n/10 each Task having limited stress during the generation
-File: ./images/load-eff.png Graphic file (type png)
+ []
-<use ./images/load-eff.png>
+
-Package pdftex.def Info: ./images/load-eff.png  used on input line 374.
+
-(pdftex.def)             Requested size: 252.0pt x 211.55406pt.
+Underfull \hbox (badness 2753) in paragraph at lines 376--377
- [10 <./images/exploit-speedup.png> <./images/exploit-eff.png> <./images/appl-s
+\OT1/ptm/m/n/10 process, speedups of over 3.5x can still be obtained.
-peedup.png> <./images/appl-eff.png>]
+ []
-Underfull \hbox (badness 4779) in paragraph at lines 387--388
+
+
+Underfull \hbox (badness 1490) in paragraph at lines 376--377
+\OT1/ptm/m/n/10 Exploit applicability and database load parameters were
+ []
+
+
+Underfull \hbox (badness 1406) in paragraph at lines 376--377
+\OT1/ptm/m/n/10 almost entirely unexplored during the preliminary result
+ []
+
+
+Underfull \hbox (badness 2027) in paragraph at lines 376--377
+\OT1/ptm/m/n/10 applicability and database load are introduced into the
+ []
+
+[10 <./images/exploit-speedup.png> <./images/exploit-eff.png> <./images/appl-sp
+eedup.png> <./images/appl-eff.png>]
+Underfull \hbox (badness 4779) in paragraph at lines 382--383
 []\OT1/ptm/m/n/10 The analysis portion of this work also has room
  []
 
@@ -723,7 +750,7 @@ Underfull \hbox (badness 10000) in paragraph at lines 112--116
 \OT1/ptm/m/n/8 ``The boost graph library, vers. 1.75.0.''
  []
 
-[11 <./images/load-speedup.png> <./images/load-eff.png>]
+
 Underfull \hbox (badness 10000) in paragraph at lines 143--146
 []\OT1/ptm/m/n/8 SchedMD, ``Slurm Workload Manager.''
  []
@@ -740,21 +767,22 @@ Before submitting the final camera ready copy, remember to:
  uses only Type 1 fonts and that every step in the generation
  process uses the appropriate paper size.
 
-[12
+[11] (./Schrick-Noah_MPI-Tasking.aux)
+
+LaTeX Warning: There were undefined references.
 
-] (./Schrick-Noah_MPI-Tasking.aux)
 Package rerunfilecheck Info: File `Schrick-Noah_MPI-Tasking.out' has not change
 d.
 (rerunfilecheck)             Checksum: 267D7A1F68719DE5E57F3E8948B6C837;3398.
  ) 
 Here is how much of TeX's memory you used:
- 12332 strings out of 476025
+ 12315 strings out of 477985
- 197046 string characters out of 5796533
+ 196595 string characters out of 5840059
- 1871388 words of memory out of 5000000
+ 1875388 words of memory out of 5000000
- 32513 multiletter control sequences out of 15000+600000
+ 32246 multiletter control sequences out of 15000+600000
- 550927 words of font info for 105 fonts, out of 8000000 for 9000
+ 552034 words of font info for 106 fonts, out of 8000000 for 9000
- 1141 hyphenation exceptions out of 8191
+ 14 hyphenation exceptions out of 8191
- 75i,14n,76p,2022b,589s stack positions out of 5000i,500n,10000p,200000b,80000s
+ 75i,14n,76p,2022b,459s stack positions out of 10000i,1000n,20000p,200000b,200000s
 </usr/share/texmf-dist/fonts/type1/public/amsfonts/cm/cmmi10.pfb></usr/share/
 texmf-dist/fonts/type1/public/amsfonts/cm/cmr10.pfb></usr/share/texmf-dist/font
 s/type1/public/amsfonts/cm/cmr7.pfb></usr/share/texmf-dist/fonts/type1/public/a
@@ -762,10 +790,10 @@ msfonts/cm/cmsy10.pfb></usr/share/texmf-dist/fonts/type1/public/amsfonts/cm/cms
 y7.pfb></usr/share/texmf-dist/fonts/type1/urw/times/utmb8a.pfb></usr/share/texm
 f-dist/fonts/type1/urw/times/utmbi8a.pfb></usr/share/texmf-dist/fonts/type1/urw
 /times/utmr8a.pfb></usr/share/texmf-dist/fonts/type1/urw/times/utmri8a.pfb>
-Output written on Schrick-Noah_MPI-Tasking.pdf (12 pages, 937784 bytes).
+Output written on Schrick-Noah_MPI-Tasking.pdf (11 pages, 922622 bytes).
 PDF statistics:
- 385 PDF objects out of 1000 (max. 8388607)
+ 373 PDF objects out of 1000 (max. 8388607)
- 308 compressed objects within 4 object streams
+ 301 compressed objects within 4 object streams
- 85 named destinations out of 1000 (max. 500000)
+ 83 named destinations out of 1000 (max. 500000)
- 298 words of extra memory for PDF output out of 10000 (max. 10000000)
+ 288 words of extra memory for PDF output out of 10000 (max. 10000000)
 

Schrick-Noah_MPI-Tasking.tex

@@ -302,6 +302,8 @@ All nodes are connected with a 10Gbps Infiniband interconnect.
 Each parameter discussed in this section was individually changed until all permutations of parameters were explored. In addition to changing the parameters, all tests were conducted on a varying number of nodes. All permutations of parameters were examined on 1 compute node (serially) through 12 compute nodes. A bash script for looping through parameters was created on the distributed computing testing platform, with jobs sent to Slurm Workload Manager \cite{Slurm}. When a job is completed with Slurm, the bash script would use grep on the output file to extract the necessary data, and add it to a CSV file that was used for the data analysis.
 
 \section{Analysis and Results}
+Due to a limited amount of compute time, only preliminary results were gathered, instead of deploying the entire testing benchmark suite. The preliminary results gathered were intended to be the ``slowest" tests that would have the lowest speedup. The ``worst-case", minimum-bound data was collected to determine potential success of this approach. If the slowest tests still yielded promising speedups or efficiencies, then this approach would be viable and appealing for future, in-depth testing that could stress each component of the generation process.
+
 Exploratory data analysis was performed on the resulting data using Python to ascertain data relationships. Due to the multivariate nature of the data, there is difficulty visualizing all four independent variables (parameters) and the outcome (runtime) simultaneously. Figure \ref{fig:para_coords} makes uses of Plotly's parallel coordinates to visualize the parameters and their relationship to the runtime. Using pivot tables, Figures \ref{fig:nodes-exp} and \ref{fig:appl-load} show the runtime as they relate to the average of each individual parameter. These figures display the expected outcome: as the number of nodes increase, the runtime decreases, and as the number of exploits, applicability of exploits, and database load increases, the runtime likewise increases.
 
 \begin{figure}[htp]
@@ -350,7 +352,7 @@ Figure \ref{fig:overall-efficiency} displays the overall minimum, maximum, and m
         \label{fig:overall-efficiency}
 \end{figure}
 
-Speedups and efficiencies were also computed across each parameter. Using pivot tables, mean speedups and mean efficiencies were computed for a parameter across all node configurations. Figures \ref{fig:param-exploit}, \ref{fig:param-appl}, \ref{fig:param-load} display the speedups and efficiencies of the exploit parameter, applicability of exploits parameter, and the database load parameter, respectively. The number of nodes has the largest impact on the exploit parameter, and Figure \ref{fig:param-exploit} illustrates that even when fewer nodes are used, speedup can still be obtained as the exploit list grows in size. Figure \ref{fig:param-appl} demonstrates that though Task 2 has less of an impact on overall runtime and contribution to speedup, speedup is still achievable as more compute nodes are added and as the applicability of exploits increase. Figure \ref{fig:param-load} highlights the increasing runtime as more database operations are performed. By dedicating nodes to solely handle database operations, the tasking pipeline is able to move to new state generation without the need to wait for all preceding database operations to complete.
+Speedups and efficiencies were also computed across each parameter. Using pivot tables, mean speedups and mean efficiencies were computed for a parameter across all node configurations. Figures \ref{fig:param-exploit}, \ref{fig:param-appl}, \ref{fig:param-load} display the speedups and efficiencies of the exploit parameter, applicability of exploits parameter, and the database load parameter, respectively. The number of nodes has the largest impact on the exploit parameter, and Figure \ref{fig:param-exploit} illustrates that even when fewer nodes are used, speedup can still be obtained as the exploit list grows in size. Figure \ref{fig:param-appl} demonstrates that though Task 2 has less of an impact on overall runtime and contribution to speedup, speedup is still achievable as more compute nodes are added and as the applicability of exploits increase. Though database load was not a parameter to easily include in preliminary testing, speedup is expected as this parameter changes. By dedicating nodes to solely handle database operations, the tasking pipeline is able to move to new state generation without the need to wait for all preceding database operations to complete.
 
 \begin{figure}
     \centering
@@ -368,18 +370,11 @@ Speedups and efficiencies were also computed across each parameter. Using pivot
     \label{fig:param-appl}
 \end{figure}
 
-\begin{figure}
-    \centering
-    \includegraphics[width=\linewidth]{"./images/load-speedup.png"}
-    \includegraphics[width=\linewidth]{"./images/load-eff.png"}
-    \caption{Mean Speedup and Efficiency for the Database Load Parameter Across the Number of Compute Nodes}
-    \label{fig:param-load}
-\end{figure}
-
-
 \section{Conclusion and Future Work} \label{sec:FW}
 This work presents a task parallelism approach for large-scale attack and compliance graphs. This approach is a distributed approach rather than a shared-memory approach, allowing for the generation of these large-scale graphs to be deployed on HPC systems. Tasks were identified in the generation process, with parallelization of the tasks clearly identified and incorporated into the tasking algorithm. The results of this approach highlighted its success, showing speedups as generation parameters and the number of nodes increased. Efficiencies were also computed, with various figures illustrating the efficiency across the generation process as a whole, as well as efficiencies across individual parameters.
 
+Though the results presented in this work were preliminary, they still highlight the viability of this approach. Despite each Task having limited stress during the generation process, speedups of over 3.5x can still be obtained. Exploit applicability and database load parameters were almost entirely unexplored during the preliminary result benchmarking, and speedup is still achievable. As exploit applicability and database load are introduced into the generation process, both speedup and efficiency are expected to increase in future testing. With more compute time, the approach of this work can be deployed on a testing platform to examine how speedups and efficiencies change as the complexity of the generation process increases.
+
 Future work can be performed to investigate and improve the method of this work. This work focused on a distributed approach to large-scale graph generation, but leaves room for additional parallelism at the node-level. For example, after work for Tasks 1 or 2 has been distributed to a node, the node can then leverage OpenMP for additional parallelism. Results can be obtained to show how the additional parallelism affects the speedup and efficiency of the approach.
 
 Additional work can be performed to investigate long-term speedup and efficiency of the approach. This work made use of a local HPC cluster with a limited number of compute nodes. The generation algorithm can be deployed to larger clusters to measure speedup and efficiency as more nodes are added to the tasking pipeline. Scalability can likewise be reexamined as the number of nodes increases.

texput.log

@@ -0,0 +1,21 @@
+This is pdfTeX, Version 3.141592653-2.6-1.40.25 (TeX Live 2023/Arch Linux) (preloaded format=pdflatex 2023.9.6)  10 SEP 2023 15:27
+entering extended mode
+ restricted \write18 enabled.
+ %&-line parsing enabled.
+**
+
+! Emergency stop.
+<*> 
+    
+End of file on the terminal!
+
+ 
+Here is how much of TeX's memory you used:
+ 5 strings out of 477985
+ 147 string characters out of 5840059
+ 1849388 words of memory out of 5000000
+ 20305 multiletter control sequences out of 15000+600000
+ 512287 words of font info for 32 fonts, out of 8000000 for 9000
+ 14 hyphenation exceptions out of 8191
+ 0i,0n,0p,43b,6s stack positions out of 10000i,1000n,20000p,200000b,200000s
+!  ==> Fatal error occurred, no output PDF file produced!