tauroast.tex

\section{High Energy Physics Applications}

We study applications taken from two experiments of the CERN Large Hadron Collider, namely the ATLAS experiment and the CMS experiment. 
In LHC, the ATLAS and CMS experiments are distinct, 
developed independently by two entirely separate physics communities. Consequently, their applications  
have very different software distribution and data management frameworks, raising the question of whether common reproducibility frameworks and tools work across the two communities.  
%making it interesting to examine if common reproducibility frameworks and 
%tools work across the two communities. 
One of the applications of the ATLAS experiment is the \emph{Athena} application, which is a general purpose processing framework including algorithms
for event reconstruction and data reduction \cite{calafiura2005athena}. The CMS experiment is conducted through an application termed  \emph{TauRoast}, which searches for specific 
cases where the Higgs boson decays to two tau leptons~\cite{chatrchyan2013search}. 

Code and data in \emph{TauRoast} are available through five different networked filesystems which are mounted locally, an HDFS cluster for the CMS dataset, 
some configuration files were stored on CVMFS~\cite{blomer2011cernvm}, and a variety of software tools were on an NFS, PanFS and AFS systems.
In addition, code may exist in version control systems such as Git, CVS, and CMS Software Distribution (CMSSW). %~\cite{cms2006cms}. 

\begin{wrapfigure}{r}{0.4\textwidth}
\small
\centering
\includegraphics[width=.4\textwidth]{data-code-size.eps}
\caption{Inputs to Tau Roast}
\label{fig:data-code-size}
\end{wrapfigure}

Data that is input to \emph{TauRoast} is obtained by reducing it through a pipeline, as shown in Figure \ref{fig:data-code-size}. Consequently, the real input data may 
vary depending upon the topic of research. Similarly the software may name many possible components but the used components are
smaller than the named ones. 

Data in \emph{Athena} is obtained through an external Dropbox-like system called the FaxBox, but does not pass through any reduction steps. Code is obtained through 
CVMFS, which provides the analysis routines. 
The invoked configuration will change, however, depending upon the input data code.  Thus in Athena
 the used code and configuration are dynamic depending upon input data, whereas in TauRoast the code and data are static, but the amount of data and code to include changes depending on the science involved.