Tag: parallel

Parallel Programming using R in Windows

Ashamed at my lack of parallel programming, I decided to learn some R Parallel Programming (after all parallel blogging is not really respect worthy in tech-geek-ninja circles).

So I did the usual Google- CRAN- search like a dog thing only to find some obstacles.

Obstacles-

Some Parallel Programming Packages like doMC are not available in Windows

http://cran.r-project.org/web/packages/doMC/index.html

Some Parallel Programming Packages like doSMP depend on Revolution’s Enterprise R (like –

http://blog.revolutionanalytics.com/2009/07/simple-scalable-parallel-computing-in-r.html

and http://www.r-statistics.com/2010/04/parallel-multicore-processing-with-r-on-windows/ (No the latest hack didnt work)

or are in testing like multicore (for Windows) so not available on CRAN

http://cran.r-project.org/web/packages/multicore/index.html

fortunately available on RForge

http://www.rforge.net/multicore/files/

Revolution did make DoSnow AND foreach available on CRAN

see http://blog.revolutionanalytics.com/2009/08/parallel-programming-with-foreach-and-snow.html

but the documentation in SNOW is overwhelming (hint- I use Windows , what does that tell you about my tech acumen)

http://sekhon.berkeley.edu/snow/html/makeCluster.html and

http://www.stat.uiowa.edu/~luke/R/cluster/cluster.html

what is a PVM or MPI? and SOCKS are for wearing or getting lost in washers till I encountered them in SNOW


Finally I did the following-and made the parallel programming work in Windows using R

require(doSNOW)
cl<-makeCluster(2) # I have two cores
registerDoSNOW(cl)
# create a function to run in each itteration of the loop

check <-function(n) {

+ for(i in 1:1000)

+ {

+ sme <- matrix(rnorm(100), 10,10)

+ solve(sme)

+ }

+ }
times <- 100     # times to run the loop
system.time(x <- foreach(j=1:times ) %dopar% check(j))
user  system elapsed
0.16    0.02   19.17
system.time(for(j in 1:times ) x <- check(j))
user  system elapsed</pre>
39.66    0.00   40.46

stopCluster(cl)

And it works!

AsterData gets $30 mill in funding

From the press release, the maker of Map Reduce based BI software gets 30 mill $ as Series C funding. Given the valuation recently by IBM to Netezza, AsterData seems set to cross the Billion Dollar valuation within the next 18-24 months IMO

Aster Data Closes $30 Million Series C Financing

Explosive Growth and Market Leadership Attracts New and Existing Investors

San Carlos, CA – September 22, 2010 – Aster Data, a market leader in big data management and advanced analytics, today announced that it has closed a $30 million Series C round of financing led by both new and existing investors. The company will use the new funding to accelerate growth, scale operations, and expand its global market share in the $20 billion database market – a market that is experiencing rapid growth as a result of both the explosion in data volumes across organizations and the urgent need to deliver a new class of analytics and data-driven applications. The Series C round of funding includes previous investors Sequoia Capital, JAFCO Ventures, Institutional Venture Partners, Cambrian Ventures, as well as an additional new strategic investor.  Also investing in this round is early investor David Cheriton, who previously backed high-growth companies including Google and VMware, and co-founded several successful technology companies.

Today’s Series C funding announcement underscores a year of strong innovation, execution, and overall momentum for the analytic database company. Key milestones include:

Strong sales growth: Since 2008, Aster Data has doubled revenue year-over-year and secured key customers that leverage Aster Data’s platform to address the big data management problem including MySpace, comScore, Barnes & Noble, and Akamai. Like so many organizations today,
Aster Data’s customers are experiencing explosive data growth across their organizations and recognize the need for rich, advanced analytics that give them deeper insights from their data.

Key executive hires: Quentin Gallivan, former CEO of both PivotLink and Postini and EVP of worldwide sales at Verisign, recently joined the company as Chief Executive Officer. In addition, earlier this year, John Calonico, previously at Interwoven, BEA, and Autodesk, joined as Chief Financial Officer; and Nitin Donde, formerly an executive at EMC and 3PAR, joined as Executive Vice President Engineering.  The strength and experience of Aster Data’s management team helps further establish a strong operational foundation for growth in 2010 and beyond.

Industry recognition: Aster Data was positioned in the “Visionaries” Quadrant of Gartner, Inc.’s

Data Warehouse Database Management Systems Magic Quadrant, published 2010 *; was recently named 2011 Tech Pioneer by the World Economic Forum; was named “Company to Watch” in the Information Management category of TechWeb’s Intelligent Enterprise 2010 Editors’ Choice Awards; and was awarded the 2010 San Francisco Business Times Technology and Innovation Award in the Best Product and Services Category.

Product Innovation: Aster Data continues to deliver ground-breaking capabilities to address the big data management and advanced analytics market need. Its recent announcement of
Aster Data nCluster 4.6 includes a column data store, making it the first hybrid row and column MPP DBMS with a unified SQL and MapReduce analytic framework for advanced analytics on large data sets. This year, Aster Data also delivered the most extensive library of pre-packaged MapReduce analytics totaling over 1000 functions, to ease and accelerate delivery of highly advanced analytic applications.

Aster Data’s analytic database, also called a ‘Data-Analytics Server’ is specifically designed to enable organizations to cost effectively store and analyze massive volumes of data. Aster Data leverages the power of commodity, general-purpose hardware, to reduce the cost to scale to support large data volumes and uniquely allows analysis of all data ‘in-database’ enabling richer and faster processing of large data sets. Aster Data’s in-database analytics engine uses the power of MapReduce, a parallel processing framework created by Google.

”The funding we received in our Series C round is a strong endorsement of Aster Data’s market leadership position and the high growth potential of the big data market,” said Quentin Gallivan, Chief Executive Officer, Aster Data. “The Aster Data team has executed exceptionally well to-date and I am excited to have the resources to accelerate the growth of the company as we expand our operations and execute aggressively across all fronts.”

Matlab-Mathematica-R and GPU Computing

Matlab announced they have a parallel computing toolbox- specially to enable GPU computing as well

http://www.mathworks.com/products/parallel-computing/

Parallel Computing Toolbox™ lets you solve computationally and data-intensive problems using multicore processors, GPUs, and computer clusters. High-level constructs—parallel for-loops, special array types, and parallelized numerical algorithms—let you parallelize MATLAB® applications without CUDA or MPI programming. You can use the toolbox with Simulink® to run multiple simulations of a model in parallel.

MATLAB GPU Support

The toolbox provides eight workers (MATLAB computational engines) to execute applications locally on a multicore desktop. Without changing the code, you can run the same application on a computer cluster or a grid computing service (using MATLAB Distributed Computing Server™). You can run parallel applications interactively or in batch.

Parallel Computing with MATLAB on Amazon Elastic Compute Cloud (EC2)

Also a video of using Mathematica and GPU

Also R has many packages for GPU computing

Parallel computing: GPUs

from http://cran.r-project.org/web/views/HighPerformanceComputing.html

  • The gputools package by Buckner provides several common data-mining algorithms which are implemented using a mixture of nVidia‘s CUDA langauge and cublas library. Given a computer with an nVidia GPU these functions may be substantially more efficient than native R routines. The rpud package provides an optimised distance metric for NVidia-based GPUs.
  • The cudaBayesreg package by da Silva implements the rhierLinearModel from the bayesm package using nVidia’s CUDA langauge and tools to provide high-performance statistical analysis of fMRI voxels.
  • The rgpu package (see below for link) aims to speed up bioinformatics analysis by using the GPU.
  • The magma package provides an interface to the hybrid GPU/CPU library Magma (see below for link).
  • The gcbd package implements a benchmarking framework for BLAS and GPUs (using gputools).

I tried to search for SAS and GPU and SPSS and GPU but got nothing. Maybe they would do well to atleast test these alternative hardwares-

Also see Matlab on GPU comparison for the product Jacket vs Parallel Computing Toolbox

http://www.accelereyes.com/products/compare

Where is Waldo? Webcast on Network Intelligence

From the good folks at AsterData, a webcast on a slightly interesting analytics topic

Enterprises and government agencies can become overwhelmed with information. The value of all that data lies in the insights it can reveal. To get the maximum value, you need an analytic platform that lets you analyze terabytes of information rapidly for immediate actionable insights.

Aster Data’s massively parallel database with an integrated analytics engine can quickly reveal hard-to-recognize trends on huge datasets which other systems miss. The secret? A patent-pending SQL-MapReduce framework that enables business analysts and business intelligence (BI) tools to iteratively analyze big data more quickly. This allows you to find anomalies more quickly and stop disasters before they happen.

Discover how you can improve:

  • Network intelligence via graph analysis to understand connectivity among suspects, information propagation, and the flow of goods
  • Security analysis to prevent fraud, bot attacks, and other breaches
  • Geospatial analytics to quickly uncover details about regions and subsets within those communities
  • Visual analytics to derive deeper insights more quickly

SAS/Blades/Servers/ GPU Benchmarks

Just checked out cool new series from NVidia servers.

Now though SAS Inc/ Jim Goodnight thinks HP Blade Servers are the cool thing- the GPU takes hardware high performance computing to another level. It would be interesting to see GPU based cloud computers as well – say for the on Demand SAS (free for academics and students) but which has had some complaints of being slow.

See this for SAS and Blade Servers-

http://www.sas.com/success/ncsu_analytics.html

To give users hands-on experience, the program is underpinned by a virtual computing lab (VCL), a remote access service that allows users to reserve a computer configured with a desired set of applications and operating system and then access that computer over the Internet. The lab is powered by an IBM BladeCenter infrastructure, which includes more than 500 blade servers, distributed between two locations. The assignment of the blade servers can be changed to meet shifts in the balance of demand among the various groups of users. Laura Ladrie, MSA Classroom Coordinator and Technical Support Specialist, says, “The virtual computing lab chose IBM hardware because of its quality, reliability and performance. IBM hardware is also energy efficient and lends itself well to high performance/low overhead computing.

Thats interesting since IBM now competes (as owner of SPSS) and also cooperates with SAS Institute

And

http://www.theaustralian.com.au/australian-it/the-world-according-to-jim-goodnight-blade-switch-slashes-job-times/story-e6frgakx-1225888236107

You’re effectively turbo-charging through deployment of many processors within the blade servers?

Yes. We’ve got machines with 192 blades on them. One of them has 202 or 203 blades. We’re using Hewlett-Packard blades with 12 CP cores on each, so it’s a total 2300 CPU cores doing the computation.

Our idea was to give every one of those cores a little piece of work to do, and we came up with a solution. It involved a very small change to the algorithm we were using, and it’s just incredible how fast we can do things now.

I don’t think of it as a grid, I think of it as essentially one computer. Most people will take a blade and make a grid out of it, where everything’s a separate computer running separate jobs.

We just look at it as one big machine that has memory and processors all over the place, so it’s a totally different concept.

GPU servers can be faster than CPU servers, though , Professor G.




Source-

http://www.nvidia.com/object/preconfigured_clusters.html

TESLA GPU COMPUTING SOLUTIONS FOR DATA CENTERS
Supercharge your cluster with the Tesla family of GPU computing solutions. Deploy 1U systems from NVIDIA or hybrid CPU-GPU servers from OEMs that integrate NVIDIA® Tesla™ GPU computing processors.

When compared to the latest quad-core CPU, Tesla 20-series GPU computing processors deliver equivalent performance at 1/20th the power consumption and 1/10th the cost. Each Tesla GPU features hundreds of parallel CUDA cores and is based on the revolutionary NVIDIA® CUDA™ parallel computing architecture with a rich set of developer tools (compilers, profilers, debuggers) for popular programming languages APIs like C, C++, Fortran, and driver APIs like OpenCL and DirectCompute.

NVIDIA’s partners provide turnkey easy-to-deploy Preconfigured Tesla GPU clusters that are customizable to your needs. For 3D cloud computing applications, our partners offer the Tesla RS clusters that are optimized for running RealityServer with iray.

Available Tesla Products for Data Centers:
– Tesla S2050
– Tesla M2050/M2070
– Tesla S1070
– Tesla M1060

Also I liked the hybrid GPU and CPU

And from a paper on comparing GPU and CPU using Benchmark tests on BLAS from a Debian- Dirk E’s excellent blog

http://dirk.eddelbuettel.com/blog/

Usage of accelerated BLAS libraries seems to shrouded in some mystery, judging from somewhat regularly recurring requests for help on lists such as r-sig-hpc(gmane version), the R list dedicated to High-Performance Computing. Yet it doesn’t have to be; installation can be really simple (on appropriate systems).

Another issue that I felt needed addressing was a comparison between the different alternatives available, quite possibly including GPU computing. So a few weeks ago I sat down and wrote a small package to run, collect, analyse and visualize some benchmarks. That package, called gcbd (more about the name below) is now onCRAN as of this morning. The package both facilitates the data collection for the paper it also contains (in the vignette form common among R packages) and provides code to analyse the data—which is also included as a SQLite database. All this is done in the Debian and Ubuntu context by transparently installing and removing suitable packages providing BLAS implementations: that we can fully automate data collection over several competing implementations via a single script (which is also included). Contributions of benchmark results is encouraged—that is the idea of the package.

And from his paper on the same-

Analysts are often eager to reap the maximum performance from their computing platforms.

A popular suggestion in recent years has been to consider optimised basic linear algebra subprograms (BLAS). Optimised BLAS libraries have been included with some (commercial) analysis platforms for a decade (Moler 2000), and have also been available for (at least some) Linux distributions for an equally long time (Maguire 1999). Setting BLAS up can be daunting: the R language and environment devotes a detailed discussion to the topic in its Installation and Administration manual (R Development Core Team 2010b, appendix A.3.1). Among the available BLAS implementations, several popular choices have emerged. Atlas (an acronym for Automatically Tuned Linear Algebra System) is popular as it has shown very good performance due to its automated and CPU-speci c tuning (Whaley and Dongarra 1999; Whaley and Petitet 2005). It is also licensed in such a way that it permits redistribution leading to fairly wide availability of Atlas.1 We deploy Atlas in both a single-threaded and a multi-threaded con guration. Another popular BLAS implementation is Goto BLAS which is named after its main developer, Kazushige Goto (Goto and Van De Geijn 2008). While `free to use’, its license does not permit redistribution putting the onus of con guration, compilation and installation on the end-user. Lastly, the Intel Math Kernel Library (MKL), a commercial product, also includes an optimised BLAS library. A recent addition to the tool chain of high-performance computing are graphical processing units (GPUs). Originally designed for optimised single-precision arithmetic to accelerate computing as performed by graphics cards, these devices are increasingly used in numerical analysis. Earlier criticism of insucient floating-point precision or severe performance penalties for double-precision calculation are being addressed by the newest models. Dependence on particular vendors remains a concern with NVidia’s CUDA toolkit (NVidia 2010) currently still the preferred development choice whereas the newer OpenCL standard (Khronos Group 2008) may become a more generic alternative that is independent of hardware vendors. Brodtkorb et al. (2010) provide an excellent recent survey. But what has been lacking is a comparison of the e ective performance of these alternatives. This paper works towards answering this question. By analysing performance across ve di erent BLAS implementations|as well as a GPU-based solution|we are able to provide a reasonably broad comparison.

Performance is measured as an end-user would experience it: we record computing times from launching commands in the interactive R environment (R Development Core Team 2010a) to their completion.

And

Basic Linear Algebra Subprograms (BLAS) provide an Application Programming Interface
(API) for linear algebra. For a given task such as, say, a multiplication of two conformant
matrices, an interface is described via a function declaration, in this case sgemm for single
precision and dgemm for double precision. The actual implementation becomes interchangeable
thanks to the API de nition and can be supplied by di erent approaches or algorithms. This
is one of the fundamental code design features we are using here to benchmark the di erence
in performance from di erent implementations.
A second key aspect is the di erence between static and shared linking. In static linking,
object code is taken from the underlying library and copied into the resulting executable.
This has several key implications. First, the executable becomes larger due to the copy of
the binary code. Second, it makes it marginally faster as the library code is present and
no additional look-up and subsequent redirection has to be performed. The actual amount
of this performance penalty is the subject of near-endless debate. We should also note that
this usually amounts to only a small load-time penalty combined with a function pointer
redirection|the actual computation e ort is unchanged as the actual object code is identi-
cal. Third, it makes the program more robust as fewer external dependencies are required.
However, this last point also has a downside: no changes in the underlying library will be
reected in the binary unless a new build is executed. Shared library builds, on the other
hand, result in smaller binaries that may run marginally slower|but which can make use of
di erent libraries without a rebuild.

Basic Linear Algebra Subprograms (BLAS) provide an Application Programming Interface(API) for linear algebra. For a given task such as, say, a multiplication of two conformantmatrices, an interface is described via a function declaration, in this case sgemm for singleprecision and dgemm for double precision. The actual implementation becomes interchangeablethanks to the API de nition and can be supplied by di erent approaches or algorithms. Thisis one of the fundamental code design features we are using here to benchmark the di erencein performance from di erent implementations.A second key aspect is the di erence between static and shared linking. In static linking,object code is taken from the underlying library and copied into the resulting executable.This has several key implications. First, the executable becomes larger due to the copy ofthe binary code. Second, it makes it marginally faster as the library code is present andno additional look-up and subsequent redirection has to be performed. The actual amountof this performance penalty is the subject of near-endless debate. We should also note thatthis usually amounts to only a small load-time penalty combined with a function pointerredirection|the actual computation e ort is unchanged as the actual object code is identi-cal. Third, it makes the program more robust as fewer external dependencies are required.However, this last point also has a downside: no changes in the underlying library will bereected in the binary unless a new build is executed. Shared library builds, on the otherhand, result in smaller binaries that may run marginally slower|but which can make use ofdi erent libraries without a rebuild.

And summing up,

reference BLAS to be dominated in all cases. Single-threaded Atlas BLAS improves on the reference BLAS but loses to multi-threaded BLAS. For multi-threaded BLAS we nd the Goto BLAS dominate the Intel MKL, with a single exception of the QR decomposition on the xeon-based system which may reveal an error. The development version of Atlas, when compiled in multi-threaded mode is competitive with both Goto BLAS and the MKL. GPU computing is found to be compelling only for very large matrix sizes. Our benchmarking framework in the gcbd package can be employed by others through the R packaging system which could lead to a wider set of benchmark results. These results could be helpful for next-generation systems which may need to make heuristic choices about when to compute on the CPU and when to compute on the GPU.

Source – DirkE’paper and blog http://dirk.eddelbuettel.com/papers/gcbd.pdf

Quite appropriately-,

Hardware solutions or atleast need to be a part of Revolution Analytic’s thinking as well. SPSS does not have any choice anymore though 😉

It would be interesting to see how the new SAS Cloud Computing/ Server Farm/ Time Sharing facility is benchmarking CPU and GPU for SAS analytics performance – if being done already it would be nice to see a SUGI paper on the same at http://sascommunity.org.

Multi threading needs to be taken care automatically by statistical software to optimize current local computing (including for New R)

Acceptable benchmarks for testing hardware as well as software need to be reinforced and published across vendors, academics  and companies.

What do you think?


AsterData releases nCluster 4.6

From the press release

Aster Data nCluster 4.6, which includes a column data store, making Aster Data nCluster 4.6 the first platform with a unified SQL-MapReduce analytic framework on a hybrid row and column massively parallel processing (MPP) database management system (DBMS). The unified SQL-MapReduce analytic framework and Aster Data’s suite of 1000+ MapReduce-ready analytic functions, delivers a substantial breakthrough in richer, high performance analytics on large data volumes where data can be stored in either a row or column format.

With Aster Data nCluster 4.6, customers can choose the data format best suited to their needs and benefit from the power of Aster Data’s SQL-MapReduce analytic capabilities, providing maximum query performance by leveraging row-only, column-only, or hybrid storage strategies. Aster Data makes selection of the appropriate storage strategy easy with the new Data Model Express tool that determines the optimal data model based on a customer’s query workloads.  Both row and column stores in Aster Data nCluster 4.6 benefit from platform-level services including Online Precision Scaling™ on commodity hardware, dynamic workload management, and always-on availability, all of which now operate on both row and column stores. All 1000+ MapReduce-ready analytic functions released previously through Aster Data Analytic Foundation — a powerful suite of pre-built MapReduce analytic software building blocks — now run on a hybrid row and column architecture.  Aster Data nCluster 4.6 also includes new pre-built analytic functions, including decision trees and histograms. For custom analytic application development, the Aster Data IDE, Aster Data Developer Express, also fully and seamlessly supports the hybrid row and column store in Aster DatanCluster 4.6.

More advanced analytics infrastructure.