Tag: people

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?


Linux= Who did what and how much?

A report distributed under Creative Commons 3 and available at

That shows Canonical — the commercial arm of Ubuntu — has contributed only about one percent of the code to the GNOME desktop for Linux. while Red Hat accounts for 17 percent of the code and Novell developers are responsible for about 11 percent. That prompted some heartburn from Mark, creator- founder Cannonical/ Ubuntu at http://www.markshuttleworth.com/archives/517

And it would be a very different story if it weren’t for the Mozilla folks and Netscape before them, and GNOME and KDE, and Google and everyone else who have exercised that stack in so many different ways, making it better along the way. There are tens of thousands of people who are not in any way shape or form associated with Ubuntu, who make this story real. Many of them have been working at it for more than a decade – it takes a long time to make an overnight success :) while Ubuntu has only been on the scene six years. So Ubuntu cannot be credited solely for the delight of its users.

Nevertheless, the Ubuntu Project does bring something unique, special and important to free software: a total commitment to everyday users and use cases, the idea that free software should be “for everyone” both economically and in ease of use, and a willingness to chase down the problems that stand between here and there. I feel that commitment is a gift back to the people who built every one of those packages. If we can bring free software to ten times the audience, we have amplified the value of your generosity by a factor of ten, we have made every hour spent fixing an issue or making something amazing, ten times as valuable. I’m very proud to be spending the time and energy on Ubuntu that I do. Yes, I could do many other things, but I can’t think of another course which would have the same impact on the world.

I recognize that not everybody will feel the same way. Bringing their work to ten times the audience without contributing features might just feel like leeching, or increasing the flow of bug reports 10x. I suppose you could say that no matter how generous we are to downstream users, if upstream is only measuring code, then any generosity other than code won’t be registered. I don’t really know what to do about that – I didn’t found Ubuntu as a vehicle for getting lots of code written, that didn’t seem to me to be what the world needed.

Open source communities work like democracies with all noise whereas R and D within corporates have a stricter hierarchy. Still for all that – Ubuntu and Android have made Linux mainstream just as R has made statistical software available to all.

And Ubuntu also has great support for R (particularly the single click R Commander Install and Icon) available at http://packages.ubuntu.com/lucid/math/r-cran-rcmdr

John M. Chambers Statistical Software Award – 2011

Write code, win cash, and the glory. Deep bow to Father John M Chambers, inventor of S ,for endowing this award for statistical software creation by grads and undergrads.

An effort to be matched by companies like SAS, SPSS which after all came from grad school work. Now back to the competition, I gotta get my homies from U Tenn in a team ( I was a grad student last year though taking this year off due to medico- financial reasons)

John M. Chambers Statistical Software Award – 2011
Statistical Computing Section
American Statistical Association

The Statistical Computing Section of the American Statistical
Association announces the competition for the John M.  Chambers
Statistical Software Award. In 1998 the Association for Computing
Machinery presented its Software System Award to John Chambers for the
design and development of S. Dr. Chambers generously donated his award
to the Statistical Computing Section to endow an annual prize for
statistical software written by an undergraduate or graduate student.
The prize carries with it a cash award of $1000, plus a substantial
allowance for travel to the annual Joint Statistical Meetings where
the award will be presented.

Teams of up to 3 people can participate in the competition, with the
cash award being split among team members. The travel allowance will
be given to just one individual in the team, who will be presented the
award at JSM.  To be eligible, the team must have designed and
implemented a piece of statistical software. The individual within
the team indicated to receive the travel allowance must have begun the
development while a student, and must either currently be a student,
or have completed all requirements for her/his last degree after
January 1, 2009.  To apply for the award, teams must provide the
following materials:

Current CV’s of all team members.

A letter from a faculty mentor at the academic institution of the
individual indicated to receive the travel award.  The letter
should confirm that the individual had substantial participation in
the development of the software, certify her/his student status
when the software began to be developed (and either the current
student status or the date of degree completion), and briefly
discuss the importance of the software to statistical practice.

A brief, one to two page description of the software, summarizing
what it does, how it does it, and why it is an important
contribution.  If the team member competing for the travel
allowance has continued developing the software after finishing
her/his studies, the description should indicate what was developed
when the individual was a student and what has been added since.

An installable software package with its source code for use by the
award committee. It should be accompanied by enough information to allow
the judges to effectively use and evaluate the software (including
its design considerations.)  This information can be provided in a
variety of ways, including but not limited to a user manual (paper
or electronic), a paper, a URL, and online help to the system.

All materials must be in English.  We prefer that electronic text be
submitted in Postscript or PDF.  The entries will be judged on a
variety of dimensions, including the importance and relevance for
statistical practice of the tasks performed by the software, ease of
use, clarity of description, elegance and availability for use by the
statistical community. Preference will be given to those entries that
are grounded in software design rather than calculation.  The decision
of the award committee is final.

All application materials must be received by 5:00pm EST, Monday,
February 21, 2011 at the address below.  The winner will be announced
in May and the award will be given at the 2011 Joint Statistical
Meetings.

Information on the competition can also be accessed on the website of
the Statistical Computing Section (www.statcomputing.org or see the
ASA website, www.amstat.org for a pointer), including the names and
contributions of previous winners.  Inquiries and application
materials should be emailed or mailed to:

Chambers Software Award
c/o Fei Chen
Avaya Labs
233 Mt Airy Rd.
Basking Ridge, NJ 07920
feic@avaya.com

Kill R? Wait a sec

1) Is R efficient? (scripting wise, and performance wise) _ Depends on how you code it- some Packages like foreach can help but basic efficiency come from programmer. XDF formats from Revoscalar -the non open R package further improve programming efficiency

2) Should R be written from scratch?

You got to be kidding- It depends on how you define scratch after 2 million users

This has been done with S, then S Plus and now R.

3) What should be the license of R (if it was made a new)?

GPL license is fine. You need to do a better job of executing the license. Currently interfaces to R exist from SPSS, SAS, KXEN , other companies as well. To my knowledge royalty payments as well as formal code sharing does not agree.

R core needs to do a better job of protecting the work of 2500 package-creators rather than settling for a few snacks at events, sponsorships, Corporate Board Membership for Prof Gentleman, and 4-5 packages donated to it. The only way R developers can currently support their research is write a book (ny Springer mostly)

Eg GGplot and Hmisc are likely to be used more by average corporate user. Do their creators deserve royalty if creators of RevoScalar are getting it?

If some of 2 million users gave 1 $ to R core (compared to 9 million in last round of funding in Revolution Analytics)- you would have enough money to create a 64 bit optimized R for Linux (missing in Enterprise R), Amazon R APIs (like Karim Chine’s efforts), R GUIs (like Rattle’s commercial version) etc etc

The developments are not surprising given that Microsoft and Intel are funding Revolution Analytics http://www.dudeofdata.com/?p=1967

R controversies come and go (this has happened before including the NYT article and shakeup at Revo)

An interesting debate on whether R should be killed to make an upgrade to a more efficient language.

From Tal (creator R Bloggers) and on R help list-

There is currently a (very !) lively discussions happening around the web, surrounding the following topics:
1) Is R efficient? (scripting wise, and performance wise)
2) Should R be written from scratch?
3) What should be the license of R (if it was made a new)?

Very serious people have taken part in the debates so far.  I hope to let you know of the places I came by, so you might be able to follow/participate
in these (IMHO) important discussions.

The discussions started in the response for the following blog post on
Xi’An’s blog:
http://xianblog.wordpress.com/2010/09/06/insane/


Followed by the (short) response post by Ross Ihaka:
http://xianblog.wordpress.com/2010/09/13/simply-start-over-and-build-something-better/


Other discussions started to appear on Andrew Gelman’s blog:
http://www.stat.columbia.edu/~cook/movabletype/archives/2010/09/ross_ihaka_to_r.html

And (many) more responses started to appear in the hackers news website:
http://news.ycombinator.com/item?id=1687054

I hope these discussions will have fruitful results for our community,
Tal

—————-Contact
Details:——————————————————-
Contact me: Tal.Galili@gmail.com |  972-52-7275845
Read me: www.talgalili.com (Hebrew) | www.biostatistics.co.il (Hebrew) |
www.r-statistics.com (English)

My 0 cents ( see it would 2 cents but it;s free)

Trrrouble in land of R…and Open Source Suggestions

Recently some comments by Ross Ihake , founder of R Statistical Software on Revolution Analytics, leading commercial vendor of R….. came to my attention-

http://www.stat.auckland.ac.nz/mail/archive/r-downunder/2010-May/000529.html

[R-downunder] Article on Revolution Analytics

Ross Ihaka ihaka at stat.auckland.ac.nz
Mon May 10 14:27:42 NZST 2010

On 09/05/10 09:52, Murray Jorgensen wrote:
> Perhaps of interest:
>
> http://www.theregister.co.uk/2010/05/06/revolution_commercial_r/

Please note that R is "free software" not "open source".  These guys
are selling a GPLed work without disclosing the source to their part
of the work. I have complained to them and so far they have given me
the brush off. I am now considering my options.

Don't support these guys by buying their product. The are not feeding
back to the rights holders (the University of Auckland and I are rights
holders and they didn't even have the courtesy to contact us).

--
Ross Ihaka                         Email:  ihaka at stat.auckland.ac.nz
Department of Statistics           Phone:  (64-9) 373-7599 x 85054
University of Auckland             Fax:    (64-9) 373-7018
Private Bag 92019, Auckland
New Zealand
and from http://www.theregister.co.uk/2010/05/06/revolution_commercial_r/
Open source purists probably won't be all too happy to learn that Revolution is going to be employing an "open core" strategy, which means the core R programs will remain open source and be given tech support under a license model, but the key add-ons that make R more scalable will be closed source and sold under a separate license fee. Because most of those 2,500 add-ons for R were built by academics and Revolution wants to supplant SPSS and SAS as the tools used by students, Revolution will be giving the full single-user version of the R Enterprise stack away for free to academics. 
Conclusion-
So one co-founder of R is advocating not to buy from Revolution Analytics , which has the other co-founder of R, Gentleman on its board. 
Source- http://www.revolutionanalytics.com/aboutus/leadership.php

2) If Revolution Analytics is using 2500 packages for free but insisting on getting paid AND closing source of it’s packages (which is a technical point- how exactly can you prevent source code of a R package from being seen)

Maybe there can be a PACKAGE marketplace just like Android Apps, Facebook Apps, and Salesforce.com Apps – so atleast some of the thousands of R package developers can earn – sorry but email lists do not pay mortgages and no one is disputing the NEED for commercializing R or rewarding developers.

Though Barr created SAS, he gave up control to Goodnight and Sall https://decisionstats.wordpress.com/2010/06/02/sas-early-days/

and Goodnight and Sall do pay their developers well- to the envy of not so well paid counterparts.

3) I really liked the innovation of Revolution Analytics RevoScalar, and I wish that the default R dataset be converted to XDF dataset so that it basically kills

off the R criticism of being slow on bigger datasets. But I also realize the need for creating an analytics marketplace for R developers and R students- so academic version of R being free and Revolution R being paid seems like a trade off.

Note- You can still get a job faster as a stats student if you mention SAS and not R as a statistical skill- not all stats students go into academics.

4) There can be more elegant ways of handling this than calling for ignoring each other as REVOLUTION and Ihake seem to be doing to each other.

I can almost hear people in Cary, NC chuckling at Norman Nie, long time SPSS opponent and now REVOLUTION CEO, and his antagonizing R’s academicians within 1 year of taking over- so I hope this ends well for all. The road to hell is paved with good intentions- so if REVOLUTION can share some source code with say R Core members (even Microsoft shares source code with partners)- and R Core and Revolution agree on a licensing royalty from each other, they can actually speed up R package creation rather than allow this 2 decade effort to end up like S and S plus and TIBCO did.

Maybe Richard Stallman can help-or maybe Ihaka has a better sense of where things will go down in a couple of years-he must know something-he invented it, didnt he

On 09/05/10 09:52, Murray Jorgensen wrote:
> Perhaps of interest:
>
> http://www.theregister.co.uk/2010/05/06/revolution_commercial_r/

Please note that R is "free software" not "open source".  These guys
are selling a GPLed work without disclosing the source to their part
of the work. I have complained to them and so far they have given me
the brush off. I am now considering my options.

Don't support these guys by buying their product. The are not feeding
back to the rights holders (the University of Auckland and I are rights
holders and they didn't even have the courtesy to contact us).

--
Ross Ihaka                         Email:  ihaka at stat.auckland.ac.nz
Department of Statistics           Phone:  (64-9) 373-7599 x 85054
University of Auckland             Fax:    (64-9) 373-7018
Private Bag 92019, Auckland
New Zealand

Data Mining 2010:SAS Conference in Vegas

An interesting conference which I attended last year, this year one of the main guests is an ex professor of mine at UTenn. I am India bound this year though for family reasons.

http://www.sas.com/events/dmconf/over.html

Latest News

Early Bird Special
Register for M2010 before Sept. 17 and save $200 on conference fees!

Additional Data Mining Resources
Find additional data mining resouces including links to whitepapers, webinars, audio seminars, videos, blogs and online communities.

Location
Caesars Palace
Las Vegas, NV

Conference: October 25-26
Pre-conference workshops: October 24
Post-conference training: October 27-29

The M2010 Data Mining Conference is an international educational conference and exhibition for data mining practitioners including analysts, statisticians, programmers, consultants and anyone involved with data management within their organization, Hosted by SAS, M2010 is now in its 13th year and has become the world’s largest data mining conference, attracting over 600 people from various industries including Financial Services, Retail, Insurance, Technology, Education, Healthcare, Pharmaceutical, Government and more.

This conference is the top-choice for serious education and career networking. Conference highlights include

  • 6 keynotes
  • 36 sessions
  • 6 session tracks
  • exhibit hall
  • poster session
  • SAS software training
  • educational workshops
  • special events
  • networking opportunities
  • predictive modeling certification testing event.

Session Topics

  • Business applications
  • Data augmentation
  • Perspectives from the financial services industry
  • Fraud detection
  • Perspectives from the healthcare industry
  • New and emerging technologies
  • Perspectives from the retail industry
  • Data mining in marketing
  • Retention and Life Cycle Analysis
  • Text mining
  • And more! (View session abstracts.)