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?


Protected: Analyzing SAS Institute-WPS Lawsuit

This content is password protected. To view it please enter your password below:

Not just a Cloud

While browsing the rather content heavy site of Oracle, I came across this interesting white paper on cloud computing.

Platform-as-a-Service Private Cloud with Oracle Fusion Middleware

at http://www.oracle.com/us/technologies/036500.pdf

It basically says that Oracle has the following offerings for PaaS-

  • Application grid
  • Oracle SOA Suite and Oracle Business Process Management Suite
  • Oracle WebCenter Suite
  • Oracle Identity Management

Here is why traditional software licensing model can be threatened by Cloud Computing. These are very basic and conservative costs. If you have a software budget you can run the numbers yourself.

Suppose you pay $10,000 for an annual license and say an extra $5,000 for hardware costs for it.Assume you are using in house resources (employees) which cost you another $50,000/year.

The per hour cost of this very basic resource is Total Cost/ Number of hours utilized.

Assuming a 100 % utilization at work hours ( which is not possible) but still .

That’s a 40 hour week * 48 weeks ( including holidays).

or 33.85 $ per hour.

That’s the cut off point for you deciding to offshore work to contractors or outsourcing.

Assuming say a more realistic 80% utilization the per hour cost is= $42.31/hour.

Now assume we cant outsource because of data hygiene or some reason- so we take the same people costs/ exclude them and calculate only the total cost of ownership ( software and hardware).

thats $15,000 per 0.8 per 40*48 hours.

That’s still an astonishing 9.76 $ per hour.

Compare this cost with the cost of running a virtual instance of R on an Amazon Ec2.

Eg. http://biocep-distrib.r-forge.r-project.org/

or using http://www.zementis.com (which is now introducing an Excel add in as well at http://www.zementis.com/Excel-Ai.htm)

The per hour costs are not going to be more than 3.5 $ per hour. Thats much much better than ANY stats software licensed today on ANY desktop /Server configuration.

See the math. Thats why cloud is much more than time sharing, Dr G 😉

First of all, I don’t see anything greatly new and wonderful and different about cloud computing. It was timesharing way back in ’60. It’s not a whole lot different. I certainly have issues asking a bank to send us all their data and we’re going to put it up on a cloud. They’re going to say, ‘What about security? How will I know who else is up there in that cloud?’ I don’t know, it’s just a cloud.-

Dr Jim Goodnight, SAS Institute.

 

Portrait of a Lady

Thats a screenshot of Daneese Cooper’s Wikepedia page. Danese was fired without severance by the Intel Capital Series B investors at http://www.reolution-computing.com If this is what you get after a lifetime of working in open Source, maybe I should recommend

people get job with Prof Jim Goodnight, Phd who rarely fires people and has managed to steer his company profitably without an IPO or Series Z funding.

On the other hand I kind of admire ladies trying to work in software companies. They are so few. and look up to people like Daneese to say that yes they can make it big too.

Good bye Daneese. May your big heart rest in piece on your blog  http://danesecooper.blogs.com/.

Screenshot-28

Mergers and Acqusitions: Analyzing them

Valuation of future cash flows is an inexact science- too often it relies either on flat historical numbers (we grew by 5% last year so next year we will grow by 10%)

To add to the fun is the agency conflict, manager’s priorities (in terms of stock options encashment) is different from owner’s priorities.

These are some ways you can track companies for analysis-

1) Make a Google Alert on Company Name

2) Track if there is sudden and sustained spike in activity – it may be that company may be on road show seeking like minded partners, investors or mergers.

3) Watch for sudden drop in news alerts- it may mean radio silence or company may be in negotiations

4) Watch how company starts behaving with traditional antagonists…….

The easiest word thrown in the melee is ethics, copyright violations or payments delayed.

I am pasting an extract by a noted and renowned analyst in the business intelligence field-

Curt Monash

His Professional opinion on SAP

SAP’s NetWeaver Business Warehouse software will soon run natively on Teradata’s database for high-end data warehousing and BI (business intelligence), the vendors announced Monday.

SAP and its BusinessObjects BI subsidiary already had partnerships and product integrations with Teradata. But the vendors’ many joint customers have been clamoring for more, and native Business Warehouse support is the answer, said Tim Lang, vice president of product management for Business Objects.

SAP expects the new capability to enter beta testing in the fourth quarter of this year, with general availability in the first quarter of 2010, according to a spokesman.

Under the partnership, SAP will be handling first-line support, according to Lang. Pricing was not available.

The announcement drew a skeptical response from analyst Curt Monash of Monash Research, who questioned how deeply SAP will be committed to selling its customers on Teradata versus rival platforms.

“Business Objects has long been an extremely important partner for Teradata. But SAP’s most important DBMS partner is and will long be IBM, simply because [IBM] DB2 is not Oracle,” Monash said.”

Credit-

http://www.infoworld.com/d/data-management/sap-and-teradata-deepen-data-warehousing-ties-088

and here are some words from Curt Monash’s personal views on SAP

Typical nonsense from SAP

Below, essentially in its entirety, is an e-mail I just received from SAP, today, January 3. (Emphasis mine.)

Thank you for attending SAPs 4th Annual Analyst Summit in Las Vegas. We hope you found the time to be valuable. To ensure that we continue meeting your informational needs, please take a few moments to complete our online survey by using the link below. We ask that you please complete the survey before December 20. We look forward to receiving your feedback.

What makes this typical piece of SAP over-organization particularly amusing is that I didnt actually attend the event. I was planning to, but after considerable effort I think I finally made it clear to VP of Analyst Relations Don Bulmer that I was fed up with being lied to* by him and his colleagues. In connection with that, we came to a mutual agreement, as it were, that I wouldnt go.

*and lied about

Obviously, administrative ineptitude and dishonesty are two very different matters, united only by the fact that they both are characteristics of SAP, particularly its analyst relations group. Having said that, I should hasten to add that there are plenty of people at SAP I still trust. If Peter Zencke or Lothar Schubert tells me something, I expect it to be true. And its not just Germans; I feel the same way about Dan Rosenberg or Andrew Cabanski-Dunning, to name just a couple non-German SAP guys.

But I have to say this both SAPs ethics and its internal business processes are sufficiently screwed up as to cast doubt on SAPs qualifications to run the worlds best-run businesses.

Source:

http://www.monashreport.com/2007/01/03/sap-nonsense-ethics/

Journalism ethics off course makes sure that journalists don’t get renumerance or have to compulsorily declare benefits openly.This is not true for online journalism as it is still evolving.

Curt Monash is the grand daddy of all Business Intelligence Journalists- he has been doing this and seen it all since 1981 ( I was 4 years old then).

Almost incorruptible and therefore much respected his Monash report remains closely watched.

Some techniques to thwart Business Intelligence journalists is off course tactics of

1) Fear

2) Uncertainity

3) Doubt

by planting false leaks, or favoring more pliable journalists than the ones who ask difficult questions.

Another way is to use Search Engine Optimization so the Google search is rendered ineffective for diificult journalists for people to read them.

Why did I start this thread?

Well it seems the Business Intelligence world is coming to a round of consolidations and mergers. So will the trend of mega vendors first mentioned by M Fauschette here lead to a trend of mega journalist agencies as well- like a Fox News for all business intelligence journalists to report and get a share of the booty.

The Business Intelligence companies have long viewed analyst relationships as an unnecessary and uncontrollable marketing channel which they would like to see evolve.

Television Ratings can be manipulated for advertising similarly can you manipulate views, page views, clicks on a website for website advertisement.The catch is Google Trends may just give you the actual picture, but you can lie low by choosing not to submit or ping google during initial days and then we the website is big enough in terms of viewers or contributing bloggers can then safely ping Google as the momentum would be inertial in terms of getting bigger and bigger.

http://www.mfauscette.com/software_technology_partn/2009/05/the-emergence-of-the-mega-tech-vendor-economy.html

Here are some facts as per companies-

1) For SAS Institute

a) WPS is launching its Desktop software which enables SAS language users to migrate seamlessly at 1/10 th of the cost of SAS Base and SAS Stat. It will include Proc Reg and Proc Logistic in this and have a huge documentation.

b) R – open source software is increasingly powerful to manipulate data. SAS/IML tried offering a peace hand but they would need to reconcile with the GPL conditions for R- so if it is a plugin the source code is open and so on

c) Inference of R may be acquired by SAS to get a limited liability stake in a R based user platform.

d) Traditional Rival SPSS ( the two have dunked it out in analytics since 40 years) has a much better GUI and launched a revamped brand PASW. They are no longer distracted with a lawsuit which curiously accused them of stock manipulation and were found innocent.

e) Jim Goodnight has been dominating the industry since 1975 and has managed to stay private despite three recessions and huge inducements ( a wise miove given the mess in the markets in 2008). After Jim who will lead SAS with as much wisdom is an open question. Jim has refused Microsoft some years back, and is still very much in command despite being isolated in terms of industry alliances he remains respected. Pressure on him to rush into a merger would may just backfire.

f) The politics of envy- SAS is hated by many analytics people just as in some corners people hate America- it is because it is number 1, and been there too long.Did you mention anti-trust investigations . Well WPS is based out of UK and the European Union takes competition much more seriously.

g) Long time grudges – SAS is disliked despite its substantial R and D investments, the care it takes of its employees, and local community. Naturally people who are excluded or were excluded at some point of time have resentments.

h) SAS ambitions in Business Intelligence where curiously it is not that expensive and is actually more efficient than other players. The recent salvo fired by Jim Davis declaring business analytics as better than business intelligence- a remark much resented by cricket loving British journalist, Peter J Thomas

http://peterthomas.wordpress.com/category/business-intelligence/sas-bi-ba-controversy/

Intellectuals can carry huge grudges for decades ( Newton and Liebnitz) or Me with people who delay my interviews.

Teradata

1) Teradata has been a big partner with both SAS and SAP. It has also been losing ground recently in the same scenario SAS will shortly face.

It was also spun off in 2007-8 by the parent company NCR

http://it.toolbox.com/blogs/infosphere/against-the-flow-ncr-unacquires-teradata-13842

So will SAS buy Teradata

Will SAP Buy Teradata

Will SAS merge with Teradata and acquired by SAP while reaching a compromise with both WPS and R Project.

Will SAS call the bluff, make sincere efforts with the GPL and academic community to reconcile, give away multiple SAS Base and SAS Stat licenses in colleges and universities (like Asia, India, China) by expanding their academic program globally, start offering more coverage to JMP at a reduced price, make a trust for succession.

I dont know. All I know is I like writing code and poetry. Any code that gets the job done.

Any poem that I want to write ( see scribd books on the right)