X-Git-Url: https://bilbo.iut-bm.univ-fcomte.fr/and/gitweb/ThesisAhmed.git/blobdiff_plain/c82561816531e7092e9f8a3751df2408d3f91f68..3dfa5bd4d94afee02f58fc41b9e95b4f2fc78133:/thesis-presentation/AhmedSlides.tex diff --git a/thesis-presentation/AhmedSlides.tex b/thesis-presentation/AhmedSlides.tex index 0a61337..2a57d28 100644 --- a/thesis-presentation/AhmedSlides.tex +++ b/thesis-presentation/AhmedSlides.tex @@ -109,13 +109,60 @@ \tableofcontents \end{frame} - %%%%%%%%%%%%%%%%%%%% %% SLIDE 03 %% %%%%%%%%%%%%%%%%%%%% \begin{frame}{Introduction and problem definition} - \section{\small {Introduction and Problem definition}} - \bf \textcolor{blue}{To get more computing power:} +\section{\small {Introduction and Problem definition}} + \centering + \includegraphics[width=0.99\textwidth]{para.pdf} +\end{frame} + + + + + + + + +\begin{frame}{Execution of synchronous parallel tasks} +\vspace{-0.5 cm} +\begin{figure} + \centering + \subfloat[Synchronous imbalanced communications]{% + \includegraphics[scale=0.49]{c1/commtasks}\label{fig:h1}} + \subfloat[Synchronous imbalanced computations]{% + \includegraphics[scale=0.49]{c1/compt}\label{fig:h2}} + % \caption{Parallel tasks on homogeneous platform} + \label{fig:homo} +\end{figure} + + \end{frame} + + +%%%%%%%%%%%%%%%%%%%% +%% SLIDE 07 %% +%%%%%%%%%%%%%%%%%%%% + + +\begin{frame}{\large Synchronous and asynchronous iterative methods } +\vspace{-0.5 cm} +\begin{figure} + +\includegraphics[scale=0.42]{syn_tasks.pdf} +\vspace{0.6 cm} +\includegraphics[scale=0.42]{Asyn_tasks.pdf} +\end{figure} + + + \end{frame} + + %%%%%%%%%%%%%%%%%%%% +%% SLIDE 03 %% +%%%%%%%%%%%%%%%%%%%% +\begin{frame}{Approaches to get more computing power} + + %\bf \textcolor{blue}{} \begin{minipage}{0.5\textwidth} \textcolor{blue}{1)} \small \bf \textcolor{black}{Increase the frequency of a processor.\\ (limited due to overheating)} \end{minipage}% @@ -128,7 +175,7 @@ \end{minipage}% \vspace{0.2cm} \begin{minipage}{0.5\textwidth} - \textcolor{blue}{2)} \small \bf \textcolor{black}{Use more nodes.} + \textcolor{blue}{2)} \small \bf \textcolor{black}{Increase the number of nodes.} \textcolor{black}{The supercomputer Tianhe-2 has more than 3 million cores and consumes around 17.8 megawatts.} @@ -142,7 +189,6 @@ - %%%%%%%%%%%%%%%%%%% %% SLIDE 04 %% %%%%%%%%%%%%%%%%%%%% @@ -151,7 +197,7 @@ \textcolor{blue}{1)} \bf \textcolor{black}{Switch-off idle nodes method} \vspace{-0.9cm} \begin{figure} - \animategraphics[autopause,loop,controls,scale=0.25,buttonsize=0.2cm]{200}{on-off/a-}{0}{69} + \animategraphics[autopause,controls,scale=0.26,buttonsize=0.2cm]{200}{on-off/a-}{0}{111} %\includegraphics[width=0.6\textwidth]{on-off/a-69} \end{figure} \end{frame} @@ -164,15 +210,16 @@ \textcolor{blue}{2)} \bf \textcolor{black}{Dynamic Voltage and Frequency Scaling (DVFS)} \vspace{-0.9cm} \begin{figure} - \animategraphics[autopause,controls,scale=0.25,buttonsize=0.2cm]{10}{DVFS-meq/a-}{0}{109} + \animategraphics[autopause,controls,scale=0.26,buttonsize=0.2cm]{10}{DVFS-meq/a-}{0}{175} %\includegraphics[width=0.6\textwidth]{DVFS-meq/a-109} \end{figure} \end{frame} - - %%%%%%%%%%%%%%%%%%%% -%% SLIDE 06 %% +%% SLIDE 06 %% +%%%%%%%%%%%%%%%%%%%% +%%%%%%%%%%%%%%%%%%%% +%% SLIDE 07 %% %%%%%%%%%%%%%%%%%%%% \begin{frame}{Motivations} \vspace{0.05cm} @@ -242,7 +289,7 @@ for a warehouse-sized computer. \begin{itemize} \small \justifying - \item Study the effect of the scaling factor on the \textbf{energy consumption and performance } of parallel applications with iterations. \medskip + \item Studying the effect of the scaling factor on the \textbf{energy consumption and performance } of parallel applications with iterations. \medskip \item Discovering the \textbf{energy-performance trade-off relation} when changing the frequency of the processor.\medskip \item Proposing an algorithm for selecting the scaling factor that produces \textbf {the optimal trade-off} between the energy consumption and the performance. \medskip @@ -259,24 +306,7 @@ for a warehouse-sized computer. -%%%%%%%%%%%%%%%%%%%% -%% SLIDE 10 %% -%%%%%%%%%%%%%%%%%%%% - - -\begin{frame}{Execution of synchronous parallel tasks} -\vspace{-0.5 cm} -\begin{figure} - \centering - \subfloat[Synchronous imbalanced communications]{% - \includegraphics[scale=0.49]{c1/commtasks}\label{fig:h1}} - \subfloat[Synchronous imbalanced computations]{% - \includegraphics[scale=0.49]{c1/compt}\label{fig:h2}} - % \caption{Parallel tasks on homogeneous platform} - \label{fig:homo} -\end{figure} - \end{frame} @@ -285,7 +315,7 @@ for a warehouse-sized computer. %% SLIDE 11 %% %%%%%%%%%%%%%%%%%%%% \begin{frame}{Energy model for a homogeneous platform} - The power consumed by a processor divided into two power metrics: the dynamic (\textcolor{red}{$P_d$}) and static + The power consumed by a processor is divided into two power metrics: the dynamic (\textcolor{red}{$P_d$}) and the static (\textcolor{red}{$P_s$}) power. \begin{equation} \label{eq:pd} @@ -302,6 +332,8 @@ for a warehouse-sized computer. \scriptsize{ \textcolor{blue}{$V$}: the supply voltage. \hspace{28 mm} \textcolor{blue}{$N_{trans}$}: number of transistors. \\ \textcolor{blue}{$K_{design}$}: design dependent parameter. \hspace{8 mm} \textcolor{blue}{$I_{leak}$}: technology dependent parameter.} + + The frequency scaling factor is the ratio between the maximum and the new frequency, \textcolor{blue}{$S = \frac{F_{max}}{F_{new}}$}. \end{frame} %%%%%%%%%%%%%%%%%%%% @@ -309,22 +341,25 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% \begin{frame}{Energy model for a homogeneous platform} - - The frequency scaling factor is the ratio between the maximum and the new frequency, \textcolor{blue}{$S = \frac{F_{max}}{F_{new}}$}. \medskip - - + \vspace{-0.77cm} + \begin{figure} + \animategraphics[autopause,controls,scale=0.3,buttonsize=0.2cm]{10}{homo-model/a-}{0}{441} + %\includegraphics[width=0.6\textwidth]{homo-model/a-356} + \end{figure} - \begin{block}{\small Rauber and Rünger's energy model} - $ E = P_{d} \cdot S_1^{-2} \cdot - \left( T_1 + \sum_{i=2}^{N} \frac{T_i^3}{T_1^2} \right) + - P_{s} \cdot S_1 \cdot T_1 \cdot N$ - \end{block} - \textcolor{blue}{$S_1$}: the maximum scaling factor.\\ - \textcolor{blue}{$P_{d}$}: the dynamic power.\\ - \textcolor{blue}{$P_{s}$}: the static power.\\ - \textcolor{blue}{$T_I$}: the execution time of the slower task.\\ - \textcolor{blue}{$T_i$}: the execution time of task i.\\ - \textcolor{blue}{$N$}: the number of nodes. + % \begin{block}{\small Rauber and Rünger's energy model} + %$ E = P_{d} \cdot S_1^{-2} \cdot + %\left( T_1 + \sum_{i=2}^{N} \frac{T_i^3}{T_1^2} \right) + + % P_{s} \cdot S_1 \cdot T_1 \cdot N$ + %\end{block} + % \textcolor{blue}{$S_1$}: the maximum scaling factor.\\ + % \textcolor{blue}{$P_{d}$}: the dynamic power.\\ + % \textcolor{blue}{$P_{s}$}: the static power.\\ + % \textcolor{blue}{$T_I$}: the execution time of the slower task.\\ + % \textcolor{blue}{$T_i$}: the execution time of task i.\\ + % \textcolor{blue}{$N$}: the number of nodes. + + \end{frame} @@ -384,23 +419,23 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 15 %% %%%%%%%%%%%%%%%%%%%% - \begin{frame}{Scaling factor selection algorithm} -\vspace{-0.75cm} - \begin{center} - \includegraphics[width=.56 \textwidth]{c1/algo-homo} - \end{center} + %\begin{frame}{Scaling factor selection algorithm} +%\vspace{-0.75cm} + % \begin{center} + %\includegraphics[width=.56 \textwidth]{c1/algo-homo} + %\end{center} -\end{frame} +%\end{frame} %%%%%%%%%%%%%%%%%%%% %% SLIDE 16 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{Scaling algorithm example} +\begin{frame}{Scaling factor selection algorithm} \vspace{-0.75cm} \begin{figure} - \animategraphics[autopause,controls,scale=0.28,buttonsize=0.2cm]{10}{dvfs-homo/a-}{0}{159} + \animategraphics[autopause,controls,scale=0.29,buttonsize=0.2cm]{10}{dvfs-homo/a-}{0}{335} %\includegraphics[width=0.6\textwidth]{dvfs-homo/a-159} \end{figure} \end{frame} @@ -432,6 +467,8 @@ for a warehouse-sized computer. \includegraphics[width=.35\textwidth]{c1/cg} \includegraphics[width=.35\textwidth]{c1/bt}} +\hspace{0.5cm} + \centering {\includegraphics[width=.55\textwidth]{c1/results.pdf}} \end{femtoBlock} \end{frame} @@ -461,25 +498,25 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 20 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The proposed new energy model} - \vspace{-0.75cm} - \begin{figure} - \animategraphics[autopause,controls,scale=0.28,buttonsize=0.2cm]{10}{homo-model/a-}{0}{356} +%\begin{frame}{The proposed new energy model} + % \vspace{-0.75cm} + %\begin{figure} + % \animategraphics[autopause,controls,scale=0.28,buttonsize=0.2cm]{10}{homo-model/a-}{0}{356} %\includegraphics[width=0.6\textwidth]{homo-model/a-356} - \end{figure} -\end{frame} + % \end{figure} +%\end{frame} %%%%%%%%%%%%%%%%%%%% %% SLIDE 21 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{\large Comparing the new model with Rauber's model } - \vspace{0.1cm} - \centering - \includegraphics[width=.45\textwidth]{c1/energy_con} +%\begin{frame}{\large Comparing the new model with Rauber's model } +% \vspace{0.1cm} +% \centering + %\includegraphics[width=.45\textwidth]{c1/energy_con} - \includegraphics[width=.5\textwidth]{c1/compare-scales} -\end{frame} + %\includegraphics[width=.5\textwidth]{c1/compare-scales} +%\end{frame} @@ -582,10 +619,10 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 26 %% %%%%%%%%%%%%%%%%%%%% - \begin{frame}{The energy model for heterogeneous cluster} - \vspace{-0.5cm} + \begin{frame}{The energy model for heterogeneous cluster} + \vspace{-0.77cm} \begin{figure} - \animategraphics[autopause,controls,scale=0.28,buttonsize=0.2cm]{10}{heter-model/a-}{0}{272} + \animategraphics[autopause,controls,scale=0.3,buttonsize=0.2cm]{10}{heter-model/a-}{0}{350} %\includegraphics[width=0.6\textwidth]{heter-model/a-272} \end{figure} \end{frame} @@ -621,22 +658,22 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 28 %% %%%%%%%%%%%%%%%%%%%% - \begin{frame}{The scaling algorithm for heter. cluster} + %\begin{frame}{The scaling algorithm for heter. cluster} - \centering - \includegraphics[width=.52\textwidth]{algo-heter} - \end{frame} + %\centering + %\includegraphics[width=.52\textwidth]{algo-heter} + %\end{frame} %%%%%%%%%%%%%%%%%%%% %% SLIDE 29 %% %%%%%%%%%%%%%%%%%%%% - \begin{frame}{The scaling algorithm example} - \vspace{-0.5cm} + \begin{frame}{The scaling algorithm for heter. cluster} + \vspace{-0.77cm} \centering \begin{figure} - \animategraphics[autopause,controls,scale=0.28,buttonsize=0.2cm]{10}{dvfs-heter/a-}{0}{650} + \animategraphics[autopause,controls,scale=0.3,buttonsize=0.2cm]{10}{dvfs-heter/a-}{0}{836} % \includegraphics[width=0.6\textwidth]{dvfs-heter/a-650} \end{figure} \end{frame} @@ -665,67 +702,39 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 31 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The simulation results} - \vspace{-5 mm} - \begin{figure}[!t] - \centering - \includegraphics[width=0.8\textwidth]{c2/energy_saving.pdf} +%\begin{frame}{The simulation results} + % \vspace{-5 mm} + % \begin{figure}[!t] + %\centering + %\includegraphics[width=0.8\textwidth]{c2/energy_saving.pdf} - \textcolor{blue}{On average, it reduces the energy consumption by \textcolor{red}{29\%} - for the class C of the NAS Benchmarks executed over 8 nodes} + % \textcolor{blue}{On average, it reduces the energy consumption by \textcolor{red}{29\%} + %for the class C of the NAS Benchmarks executed over 8 nodes} - \end{figure} -\end{frame} + % \end{figure} +%\end{frame} %%%%%%%%%%%%%%%%%%%% %% SLIDE 32 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The simulation results} - \vspace{-5 mm} - \begin{figure}[!t] - \centering +%\begin{frame}{The simulation results} + % \vspace{-5 mm} + % \begin{figure}[!t] + % \centering - \includegraphics[width=.8\textwidth]{c2/perf_degra.pdf} + % \includegraphics[width=.8\textwidth]{c2/perf_degra.pdf} - \textcolor{blue}{On average, it degrades by \textcolor{red}{3.8\%} the performance - of NAS Benchmarks class C executed over 8 nodes} - \end{figure} -\end{frame} + % \textcolor{blue}{On average, it degrades by \textcolor{red}{3.8\%} the performance + % of NAS Benchmarks class C executed over 8 nodes} + % \end{figure} +%\end{frame} -%%%%%%%%%%%%%%%%%%%% -%% SLIDE 33 %% -%%%%%%%%%%%%%%%%%%%% -\begin{frame}{The results of the three power scenarios} - \vspace{-5 mm} - \begin{figure}[!t] - \centering - \includegraphics[width=.55\textwidth]{c2/three_power.pdf} - \vspace{10 mm} - \includegraphics[width=.55\textwidth]{c2/three_scenarios.pdf} - \end{figure} -\end{frame} - -%%%%%%%%%%%%%%%%%%%% -%% SLIDE 34 %% -%%%%%%%%%%%%%%%%%%%% -\begin{frame}{Comparing the objective function to EDP} - - EDP is the products between the energy consumption and the delay. - \vspace{-5 mm} - \begin{figure}[!t] - \centering - \includegraphics[width=.55\textwidth]{c2/avg_compare.pdf} - - \includegraphics[width=.55\textwidth]{c2/compare_with_EDP.pdf} - \end{figure} -\end{frame} - @@ -745,10 +754,10 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 36 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The grid architecture} -\begin{center} -\includegraphics[width=.8\textwidth]{c2/init_freq.pdf} -\end{center} +%\begin{frame}{The grid architecture} +%\begin{center} +%\includegraphics[width=.8\textwidth]{c2/init_freq.pdf} +%\end{center} %\begin{frame}{Performance, Energy and trade-off models} \small %\begin{block}{\small The performance model of grid} @@ -780,7 +789,7 @@ for a warehouse-sized computer. % \end{block} - \end{frame} + %\end{frame} @@ -838,11 +847,32 @@ for a warehouse-sized computer. +%%%%%%%%%%%%%%%%%%%% +%% SLIDE 33 %% +%%%%%%%%%%%%%%%%%%%% +\begin{frame}{The results of the three power scenarios} + \vspace{-5 mm} + \begin{figure}[!t] + \centering + \includegraphics[width=.45\textwidth]{c2/eng_pow.eps} + \hspace{0.3cm} + \includegraphics[width=.45\textwidth]{c2/per_pow.eps} + \vspace{4 mm} + \includegraphics[width=.7\textwidth]{c2/three_scenarios.pdf} + \end{figure} +\end{frame} + + + + + + + %%%%%%%%%%%%%%%%%%%% %% SLIDE 39 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{Experiments over Grid'5000} - \textcolor{blue}{One core and Multi-cores per node results:} +\begin{frame}{One core and Multi-cores per node results} + %\textcolor{blue}{One core and Multi-cores per node results:} \begin{figure}[h!] \includegraphics[width=.48\textwidth]{c2/eng_s_mc.eps} @@ -854,7 +884,20 @@ for a warehouse-sized computer. \end{frame} - +%%%%%%%%%%%%%%%%%%%% +%% SLIDE 34 %% +%%%%%%%%%%%%%%%%%%%% +\begin{frame}{Comparing the objective function to EDP} + + EDP is the products between the energy consumption and the delay. + \vspace{-5 mm} + \begin{figure}[!t] + \centering + \includegraphics[width=.6\textwidth]{c2/edp_dist.eps} + + + \end{figure} +\end{frame} %\begin{frame}{Summary} %\begin{itemize} % \small @@ -894,7 +937,7 @@ for a warehouse-sized computer. \textcolor{blue}{The execution of a synchronous parallel iterative application over a grid } \vspace{-8 mm} \begin{figure} - \animategraphics[autopause,controls,scale=0.25,buttonsize=0.2cm]{10}{syn/a-}{0}{503} + \animategraphics[autopause,controls,scale=0.26,buttonsize=0.2cm]{10}{syn/a-}{0}{647} %\includegraphics[width=0.6\textwidth]{syn/a-503} \end{figure} \end{frame} @@ -908,7 +951,7 @@ for a warehouse-sized computer. \textcolor{blue}{The execution of an asynchronous parallel iterative application over a grid } \vspace{-8 mm} \begin{figure} - \animategraphics[autopause,controls,scale=0.25,buttonsize=0.2cm]{10}{asyn/a-}{0}{440} + \animategraphics[autopause,controls,scale=0.26,buttonsize=0.2cm]{10}{asyn/a-}{0}{556} %\includegraphics[width=0.6\textwidth]{asyn/a-440} \end{figure} \end{frame} @@ -922,7 +965,7 @@ for a warehouse-sized computer. \textcolor{blue}{Using asynchronous communications with DVFS } \vspace{-8 mm} \begin{figure} - \animategraphics[autopause,controls,scale=0.25,buttonsize=0.2cm]{10}{asyn+dvfs/a-}{0}{314} + \animategraphics[autopause,controls,scale=0.26,buttonsize=0.2cm]{10}{asyn+dvfs/a-}{0}{344} %\includegraphics[width=0.6\textwidth]{asyn+dvfs/a-314} \end{figure} \end{frame} @@ -1028,27 +1071,27 @@ for a warehouse-sized computer. %%%%%%%%%%%%%%%%%%%% %% SLIDE 48 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The simulation results} -\centering \small \textcolor{blue}{The best scenario in terms of energy and performance is the Async. MS with Sync. DVFS} +%\begin{frame}{The simulation results} +%\centering \small \textcolor{blue}{The best scenario in terms of energy and performance is %the Async. MS with Sync. DVFS} -\centering - \includegraphics[scale=0.42]{c3/energy_saving.eps} +%\centering + % \includegraphics[scale=0.42]{c3/energy_saving.eps} - \centering The average energy saving = \textcolor{red}{22\%} -\end{frame} + %\centering The average energy saving = \textcolor{red}{22\%} +%\end{frame} %%%%%%%%%%%%%%%%%%%% %% SLIDE 49 %% %%%%%%%%%%%%%%%%%%%% -\begin{frame}{The simulation results} -\centering +%\begin{frame}{The simulation results} +%\centering - \includegraphics[scale=0.42]{c3/perf_degra.eps} + % \includegraphics[scale=0.42]{c3/perf_degra.eps} - \centering The average speed-up = \textcolor{red}{5.72\%} -\end{frame} +%\centering The average speed-up = \textcolor{red}{5.72\%} +%\end{frame} @@ -1056,7 +1099,7 @@ for a warehouse-sized computer. %% SLIDE 50 %% %%%%%%%%%%%%%%%%%%%% \begin{frame}{The Grid'5000 results} - \vspace{-20 mm} + \vspace{-10 mm} \begin{figure}[!t] \centering \hspace{-8 mm} @@ -1065,6 +1108,9 @@ for a warehouse-sized computer. \end{figure} \vspace{-5 mm} \centering \footnotesize + + %\small \textcolor{blue}{The best scenario in terms of energy and performance is the Async. MS with Sync. DVFS} + The average energy saving = \textcolor{red}{26.93\%}, the average speed-up = \textcolor{red}{21.48\%} \end{frame}