[Swift-commit] r7119 - SwiftApps/Swift-MapRed/Paper

[ketan at ci.uchicago.edu]

Author: ketan
Date: 2013-10-02 12:09:06 -0500 (Wed, 02 Oct 2013)
New Revision: 7119

Modified:
   SwiftApps/Swift-MapRed/Paper/.latex.out
   SwiftApps/Swift-MapRed/Paper/swifthadoop.tex
Log:
adding notes

Modified: SwiftApps/Swift-MapRed/Paper/.latex.out
===================================================================
--- SwiftApps/Swift-MapRed/Paper/.latex.out	2013-10-02 15:25:48 UTC (rev 7118)
+++ SwiftApps/Swift-MapRed/Paper/.latex.out	2013-10-02 17:09:06 UTC (rev 7119)
@@ -151,12 +151,13 @@
  uses only Type 1 fonts and that every step in the generation
  process uses the appropriate paper size.
 
-[5 <./fig/combiner.png>] (./swifthadoop.aux) )
+[5 <./fig/combiner.png>] [6] (./swifthadoop.aux) )
 (see the transcript file for additional information){/opt/local/share/texmf-tex
 live/fonts/enc/dvips/base/8r.enc}</opt/local/share/texmf-texlive/fonts/type1/pu
-blic/amsfonts/cm/cmsy8.pfb></opt/local/share/texmf-texlive/fonts/type1/urw/time
-s/utmb8a.pfb></opt/local/share/texmf-texlive/fonts/type1/urw/times/utmbi8a.pfb>
-</opt/local/share/texmf-texlive/fonts/type1/urw/times/utmr8a.pfb></opt/local/sh
-are/texmf-texlive/fonts/type1/urw/times/utmri8a.pfb>
-Output written on swifthadoop.pdf (5 pages, 210267 bytes).
+blic/amsfonts/cm/cmsy10.pfb></opt/local/share/texmf-texlive/fonts/type1/public/
+amsfonts/cm/cmsy8.pfb></opt/local/share/texmf-texlive/fonts/type1/urw/times/utm
+b8a.pfb></opt/local/share/texmf-texlive/fonts/type1/urw/times/utmbi8a.pfb></opt
+/local/share/texmf-texlive/fonts/type1/urw/times/utmr8a.pfb></opt/local/share/t
+exmf-texlive/fonts/type1/urw/times/utmri8a.pfb>
+Output written on swifthadoop.pdf (6 pages, 219156 bytes).
 Transcript written on swifthadoop.log.

Modified: SwiftApps/Swift-MapRed/Paper/swifthadoop.tex
===================================================================
--- SwiftApps/Swift-MapRed/Paper/swifthadoop.tex	2013-10-02 15:25:48 UTC (rev 7118)
+++ SwiftApps/Swift-MapRed/Paper/swifthadoop.tex	2013-10-02 17:09:06 UTC (rev 7119)
@@ -452,6 +452,29 @@
 
 \bibliographystyle{IEEEtran}
 \bibliography{ref}
+\ketanote{
+NOTES
+\begin{itemize}
+\item HDFS on RAM will provide built-in replication.
+\item How to work both with and without name/value pairs: Swift has no intrinsic name/value concept, and one can find good use cases both with and without keys.  Note that the separate project to add associative arrays to Swift is one way to integrate the concept of keys
+\item How to do a reduction trees (especially for non-key-based workflows) in a manner that reduces the amount of data and avoids the requirement of sending the output of every map operation back to a single site for reduction
+\item The idea of file-based vs. record-based distributed computing. Distinct advantages of each from the point of view of:
+\item Usability of systems
+\item Adaptation to different kinds of fabric
+\item Suitability to application classes
+\item Performance
+\item Portability of solution
+\item Parallel find is a nice example especially on distributed Chirp servers
+\item Experiment platforms: + Cray systems: Raven, Blue Waters, Beagle + Bluegene/p
+\item Example workflow apps:
+    + DSSAT : 120k jobs, real science, high priority
+    + Synthetic workflow patterns: SISO, SIMO, MISO, MIMO
+    + Parallel and distributed find off Chirp servers
+\item Expressivity | Programmability | Interoperability
+\item Why a scientist program one way or the other
+\item Compare with Spark model of computation
+\end{itemize}    
+}
 \end{document}
 NOTES
 -- HDFS on RAM will provide built-in replication.