User: knarf Date: 2012/04/04 07:15 AM

Added: /cactus_intro/ cactus_intro.pdf /hpc_intro/ hpc_intro.pdf /introduction/ introduction.pdf /pics/ supernova_model.pdf

Modified: /cactus_intro/ cactus_intro.tex /hpc_intro/ hpc_intro.tex

Log: some changes to introduction, commit pdfs

Directory Changes:

Directory: /svn:mime-type/ ==========================

+ application/pdf

File Changes:

Directory: /cactus_intro/ =========================

File [added]: cactus_intro.pdf Delta lines: +0 -0 =================================================================== (Binary files differ)

Property changes on: cactus_intro/cactus_intro.pdf ___________________________________________________________________

File [modified]: cactus_intro.tex Delta lines: +68 -16 =================================================================== --- cactus_intro/cactus_intro.tex 2012-04-04 07:50:13 UTC (rev 33) +++ cactus_intro/cactus_intro.tex 2012-04-04 12:15:48 UTC (rev 34) @@ -72,24 +72,60 @@ \end{itemize} }

-% Copied from the web site. -\begin{frame} \frametitle{History} - \begin{itemize} \justifying - \item Cactus is a direct descendant of many years of code development in Ed -Seidel's group of researchers at the National Center for Supercomputing -Applications (NCSA) as the group wrestled to numerically solve Einstein's -Equations for General Relativity and thus model black holes, neutron stars, and -boson stars. - \item In 1995, Ed Seidel and many of his group moved to the Max Planck -Institute for Gravitational Physics (Albert Einstein Institute or AEI) in -Potsdam, Germany. Frustrated by the difficulties of coordinating the -development and use of several different codes across a large collaborative -group, Paul Walker, Joan Massó, Ed Seidel, and John Shalf, designed and wrote -the original version of Cactus, Cactus 1.0, which provided a collaborative -parallel toolkit for numerical relativity. +\frame{\frametitle{Modularization Motivation} +Simulations are complex, may contain several models at once.\ +Example: Long Gamma Ray Burst: + \begin{columns}[t] + \column{.4\textwidth} + \begin{center} + \includegraphics[width=0.9\textwidth]{supernova_model} + \end{center} + \column{.5\textwidth} + \begin{itemize} + \item General Relativity (black hole) + \item Relativistic Hydrodynamics (star) + \item Microphysics, equation of state (shock wave) + \item Neutrino radiation (cooling, heating) + \item Magnetic Fields (Jet formation) + \item Photon radiation (afterglow) + \end{itemize} + + \end{columns} +} + +\frame{\frametitle{Modularization Motivation} + Added Problems: + \begin{itemize} + \item Example: Einstein Toolkit (not untypical) + \item Code 12+ years old + \item Grad students leave after 3 productive years + \item Most original authors not available anymore + \item Developers distributed over many places in several continents + \item Most physicists are not good programmers \end{itemize} -\end{frame} +}

+\frame{\frametitle{Component Architecture} + \begin{itemize} + \item Split program into independent components (as much as possible) + \item \textit{Framework} provides glue between these + \item Components are developed independently by small groups + \item End user assembles all code: no central control, no authorative version + \item Framework itself does no real work + \item Components don't interact directly with each other + \end{itemize} +} + +\frame { \frametitle{Cactus History} + \begin{itemize} + \item Direct descendant of many years of code development in Ed Seidel's group of + researchers at NCSA + \item 1995, Paul Walker, Joan Masso, Ed Seidel, and John Shalf: Cactus~1.0 + \item Originally for numerical relativity + \item Over the years generalized for use by scientists in other domains + \end{itemize} +} + \frame { \frametitle{Current Users and Developers} \centering \includegraphics[height=7cm]{groups_map}\ @@ -224,6 +260,22 @@ \end{columns} \end{frame}

+\frame { \frametitle{Typical list of component tasks} + \abspic{puzzle}{0.0}{0.6}{0.4} + \begin{itemize} + \item Evolution systems + \item Boundary conditions + \item Initial conditions + \item Time stepping method + \item Finite Difference methods + \item Simulation grid (distributed arrays) + \item I/O (more on next slide) + \item Simulation domain specification + \item Termination condition + \item Twitter client + \end{itemize} +} + \frame { \frametitle{I/O Capabilities} Usual I/O and checkpointing in different formats: \begin{columns}[t]

Directory: /hpc_intro/ ======================

File [added]: hpc_intro.pdf Delta lines: +0 -0 =================================================================== (Binary files differ)

Property changes on: hpc_intro/hpc_intro.pdf ___________________________________________________________________

File [modified]: hpc_intro.tex Delta lines: +31 -0 =================================================================== --- hpc_intro/hpc_intro.tex 2012-04-04 07:50:13 UTC (rev 33) +++ hpc_intro/hpc_intro.tex 2012-04-04 12:15:48 UTC (rev 34) @@ -249,4 +249,35 @@ \end{itemize} }

+\frame{\frametitle{Allocations} +\begin{itemize} + \item Supercomputer time is expensive. + \item 1 CPU hour costs about 5 US-cents + \item 100k cores: $5,000/h + \item Allocation: Amount of computer time a group of + users can use within some time (a year typically) + \item How to get allocation: proposal +\end{itemize} +} + +\frame{\frametitle{Batch Processing} +Simulations take too long +\begin{itemize} + \item Cannot be awake at all times + \item Each user error can destroy weeks of data + \item SC Expensive + \begin{itemize} + \item cannot wait for user input + \item should be utilized close to 100% + \end{itemize} +\end{itemize} +$\rightarrow$ +\begin{itemize} + \item Need to plan simulations carefully ahead of time + \item Then let them run automatically + \item Often don't start right away + \item Might need to be \textit{checkpointed} +\end{itemize} +} + \end{document}

Directory: /introduction/ =========================

File [added]: introduction.pdf Delta lines: +0 -0 =================================================================== (Binary files differ)

Property changes on: introduction/introduction.pdf ___________________________________________________________________

Directory: /pics/ =================

File [added]: supernova_model.pdf Delta lines: +0 -0 =================================================================== (Binary files differ)

Property changes on: pics/supernova_model.pdf ___________________________________________________________________