related_work.inc

\section{Related Work}

In situ processing has become increasingly popular in recent years.
%
Bauer et al. surveyed key research and infrastructures and we direct interested readers
to their survey~\cite{Bauer:InSitu-STAR:2016}.
%
Many of the design decisions made by ALPINE are similar to LibSim~\cite{LibSim} and Catalyst~\cite{Catalyst}: 
linking into the simulation binary, alternating execution between simulation and visualization,
and sharing the same memory space.
%
This contrasts with an approach like ADIOS~\cite{Lofstead2008}, which takes a more loosely-coupled
approach.

\fix{SENSEI~\cite{AyachitSENSEI} is a generic in situ interface that can be characterized as a ``write once, use everywhere'' approach that supports Catalyst, Libsim, and ADIOS back-ends. The ALPINE in situ infrastructure differs in that we have an execution model and our own visualization routines. Additionally, our execution model allows us to perform hybrid in situ, where we can perform on-node visualization then send the results off-node (e.g., ADIOS). }

The ALPINE in situ infrastructure is heavily based on Strawman~\cite{LarsenStrawman}, 
with Strawman effectively serving as an early prototype for ALPINE.
%
The key differences between ALPINE and Strawman are 
(1) ALPINE is intended for end users while Strawman was a mini-app,  % ALPINE includes a new interface intended for end users ?
(2) ALPINE has extended distributed-memory support, including usage of DIY~\cite{DIY}, 
(3) ALPINE uses VTK-m~\cite{Moreland:CGA2016} where Strawman used EAVL~\cite{EAVL}, and
(4) ALPINE has extended support for using other packages and libraries.