Programming with Big Data in R: Difference between revisions

Revision as of 15:25, 27 June 2013

Revision as of 15:27, 27 June 2013

Line 19:

* The pbdR built on [http://cran.r-project.org/package=pbdMPI pbdMPI] uses [[SPMD|SPMD parallelism]] where every processors are considered as workers and own parts of data. The [[SPMD|SPMD parallelism]]<ref name=spmd/><ref name=spmd_ostrouchov/> introduced in mid 1980 is particularly efficient in homogeneous computing environments for large data, for example, performing [[Singular value decomposition|singular value decomposition]]<ref>{{Cite book | last1=Golub | first1=Gene H. | author1-link=Gene H. Golub | last2=Van Loan | first2=Charles F. | author2-link=Charles F. Van Loan | title=Matrix Computations | publisher=Johns Hopkins | edition=3rd | isbn=978-0-8018-5414-9 | year=1996 }}

</ref> on a large matrix, or performing [[Mixture model|clustering analysis]] on high-dimensional large data. On the other hand, there is no restriction to use [[Master/slave (technology)|manager/workers parallelism]] in [[SPMD|SPMD parallelism]] environment.

* The [http://cran.r-project.org/package=Rmpi Rmpi]<ref name=rmpi/> uses [[Master/slave (technology)|manager/workers parallelism]] where one main processor (manager) servers as the control of all other processors (workers). The [[Master/slave (technology)|manager/workers parallelism]]<ref>[http://userpages.uni-koblenz.de/~laemmel/MapReduce/paper.pdf "Google's MapReduce Programming Model -- Revisited"] — paper by Ralf Lämmel; from [[Microsoft]]</ref> introduced inaround midearly 2000 is particularly efficient for large tasks in small [[Computer cluster|clusters]], for example, [[Bootstrapping (statistics)|bootstrap method]] and [[Monte Carlo method|Monte Carlo simulation]] in applied statistics since [[Independent and identically distributed random variables|i.i.d.]] assumption is commonly used in most [[Statistics|statistical analysis]]. In particular, [http://math.acadiau.ca/ACMMaC/Rmpi/structure.html|task pull] parallelism has better performance for Rmpi in heterogeneous computing environments.

The idea of [[SPMD|SPMD parallelism]] is to let every processors do the same works but on different parts of large data, for example, modern [[Graphics processing unit|GPU]] is a large collection of slower co-processors which can simply apply the same computation on different parts of (relatively smaller) data, but it ends up an efficient way to obtain final solution.<ref>{{cite web | url = http://www.engadget.com/2006/09/29/stanford-university-tailors-folding-home-to-gpus/ | title = Stanford University tailors Folding@home to GPUs | author = Darren Murph | accessdate = 2007-10-04 }}</ref><ref>{{cite web | url = http://graphics.stanford.edu/~mhouston/ | title = Folding@Home - GPGPU | author = Mike Houston | accessdate = 2007-10-04 }}</ref> It is clearly that pbdR is not only suitable for small [[Computer cluster|clusters]], but also is stabler for analyzing [[Big data]] and is more scalable for [[Supercomputer|supercomputers]].<ref>{{cite journal|author=Schmidt, D., Ostrouchov, G., Chen, W.-C., and Patel, P.|title=Tight Coupling of R and Distributed Linear Algebra for High-Level Programming with Big Data|year=2012|pages=811-815|journal=High Performance Computing, Networking, Storage and Analysis (SCC), 2012 SC Companion:|url=http://dl.acm.org/citation.cfm?id=2477156}}</ref>