Professors and Patches: For a Betterrrr R

Professors sometime throw out provocative statements to ensure intellectual debate. I have had almost 1500+ hits in less than 2 days ( and I am glad I am on , my old beloved server would have crashed))

The remarks from Ross Ihaka, covered before and also at Xian’s blog at

Note most of his remarks are techie- and only a single line refers to Revlution Analytics.

Other senior members of community (read- professors are silent, though brobably some thought may have been ignited behind scenes)

Ross Ihaka Says:
September 12, 2010 at 1:23 pm

Since (something like) my name has been taken in vain here, let me
chip in.

I’ve been worried for some time that R isn’t going to provide the base
that we’re going to need for statistical computation in the
future. (It may well be that the future is already upon us.) There
are certainly efficiency problems (speed and memory use), but there
are more fundamental issues too. Some of these were inherited from S
and some are peculiar to R.

One of the worst problems is scoping. Consider the following little

f =
function() {
if (runif(1) > .5)
x = 10

The x being returned by this function is randomly local or global.
There are other examples where variables alternate between local and
non-local throughout the body of a function. No sensible language
would allow this. It’s ugly and it makes optimisation really
difficult. This isn’t the only problem, even weirder things happen
because of interactions between scoping and lazy evaluation.

In light of this, I’ve come to the conclusion that rather than
“fixing” R, it would be much more productive to simply start over and
build something better. I think the best you could hope for by fixing
the efficiency problems in R would be to boost performance by a small
multiple, or perhaps as much as an order of magnitude. This probably
isn’t enough to justify the effort (Luke Tierney has been working on R
compilation for over a decade now).

To try to get an idea of how much speedup is possible, a number of us
have been carrying out some experiments to see how much better we
could do with something new. Based on prototyping we’ve been doing at
Auckland, it looks like it should be straightforward to get two orders
of magnitude speedup over R, at least for those computations which are
currently bottle-necked. There are a couple of ways to make this

First, scalar computations in R are very slow. This in part because
the R interpreter is very slow, but also because there are a no scalar
types. By introducing scalars and using compilation it looks like its
possible to get a speedup by a factor of several hundred for scalar
computations. This is important because it means that many ghastly
uses of array operations and the apply functions could be replaced by
simple loops. The cost of these improvements is that scope
declarations become mandatory and (optional) type declarations are
necessary to help the compiler.

As a side-effect of compilation and the use of type-hinting it should
be possible to eliminate dispatch overhead for certain (sealed)
classes (scalars and arrays in particular). This won’t bring huge
benefits across the board, but it will mean that you won’t have to do
foreign language calls to get efficiency.

A second big problem is that computations on aggregates (data frames
in particular) run at glacial rates. This is entirely down to
unnecessary copying because of the call-by-value semantics.
Preserving call-by-value semantics while eliminating the extra copying
is hard. The best we can probably do is to take a conservative
approach. R already tries to avoid copying where it can, but fails in
an epic fashion. The alternative is to abandon call-by-value and move
to reference semantics. Again, prototyping indicates that several
hundredfold speedup is possible (for data frames in particular).

The changes in semantics mentioned above mean that the new language
will not be R. However, it won’t be all that far from R and it should
be easy to port R code to the new system, perhaps using some form of
automatic translation.

If we’re smart about building the new system, it should be possible to
make use of multi-cores and parallelism. Adding this to the mix might just
make it possible to get a three order-of-magnitude performance boost
with just a fraction of the memory that R uses. I think it’s something
really worth putting some effort into.

I also think one other change is necessary. The license will need to a
better job of protecting work donated to the commons than GPL2 seems
to have done. I’m not willing to have any more of my work purloined by
the likes of Revolution Analytics, so I’ll be looking for better
protection from the license (and being a lot more careful about who I
work with).

The discussion spilled over to Stack Overflow as well

n the past week I’ve been following a discussion where Ross Ihaka wrote (here ):

I’ve been worried for some time that R isn’t going to provide the base that we’re going to need for statistical computation in the future. (It may well be that the future is already upon us.) There are certainly efficiency problems (speed and memory use), but there are more fundamental issues too. Some of these were inherited from S and some are peculiar to R.

He then continued explaining. This discussion started from this post, and was then followed by commentsherehereherehereherehere and maybe some more places I don’t know of.

We all know the problem now.

R can be improved substantially in terms of speed.

For some solutions, here are the patches by Radford-


    Speeds up access to lists, pairlists, and environments using the
    $ operator.  The speedup comes mainly from avoiding the overhead of 
    calling DispatchOrEval if there are no complexities, from passing
    on the field to extract as a symbol, or a name, or both, as available,
    and then converting only as necessary, from simplifying and inlining
    the pstrmatch procedure, and from not translating string multiple

    Relevant timing test script:  test-dollar.r 

    This test shows about a 40% decrease in the time needed to extract
    elements of lists and environments.

    Changes unrelated to speed improvement:

    A small error-reporting bug is fixed, illustrated by the following
    output with r52822:

    > options(warnPartialMatchDollar=TRUE)
    > pl <- pairlist(abc=1,def=2)
    > pl$ab
    [1] 1
    Warning message:
    In pl$ab : partial match of 'ab' to ''

    Some code is changed at the end of R_subset3_dflt because it seems 
    to be more correct, as discussed in code comments. 


    Speeds up a large number of operations by avoiding allocation of
    an extra CONS cell in the procedures for evaluating argument lists.

    Relevant timing test scripts:  all of them, but will look at test-em.r 

    On test-em.r, the speedup from this patch is about 5%.


    Speeds up lookup of symbols defined in the base environment, by
    flagging symbols that have a base environment definition recorded
    in the global cache.  This allows the definition to be retrieved
    quickly without looking in the hash table.  

    Relevant timing test scripts:  all of them, but will look at test-em.r 

    On test-em.r, the speedup from this patch is about 3%.

    Issue:  This patch uses the "spare" bit for the flag.  This bit is
    misnamed, since it is already used elsewhere (for closures).  It is
    possible that one of the "gp" bits should be used instead.  The
    "gp" bits should really be divided up for faster access, and so that
    their present use is apparent in the code.

    In case this use of the "spare" bit proves unwise, the patch code is 
    conditional on FAST_BASE_CACHE_LOOKUP being defined at the start of


    Speeds up lookup of function symbols that begin with a character
    other than a letter or ".", by allowing fast bypass of non-global
    environments that do not contain (and have never contained) symbols 
    of this sort.  Since it is expected that only functions will be
    given names of this sort, the check is done only in findFun, though
    it could also be done in findVar.

    Relevant timing test scripts:  all of them, but will look at test-em.r 

    On test-em.r, the speedup from this patch is about 8%.    

    Issue:  This patch uses the "spare" bit to flag environments known
    to not have symbols starting with a special character.  See remarks
    on patch-fast-base.

    In case this use of the "spare" bit proves unwise, the patch code is 
    conditional on FAST_SPEC_BYPASS being defined at the start of envir.r.


    Speeds up for loops by not allocating new space for the loop
    variable every iteration, unless necessary.  

    Relevant timing test script:  test-for.r

    This test shows a speedup of about 5%.  

    Change unrelated to speed improvement:

    Fixes what I consider to be a bug, in which the loop clobbers a
    global variable, as demonstrated by the following output with r52822:

    > i <- 99
    > f <- function () for (i in 1:3) { print(i); if (i==2) rm(i); }
    > f()
    [1] 1
    [1] 2
    [1] 3
    > print(i)
    [1] 3


    Speeds up matrix products, including vector dot products.  The
    speed issue here is that the R code checks for any NAs, and 
    does the multiply in the matprod procedure (in array.c) if so,
    since BLAS isn't trusted with NAs.  If this check takes about
    as long as just doing the multiply in matprod, calling a BLAS
    routine makes no sense.  

    Relevant time test script:  test-matprod.r

    With no external BLAS, this patch speeds up long vector-vector 
    products by a factor of about six, matrix-vector products by a
    factor of about three, and some matrix-matrix products by a 
    factor of about two.

    Issue:  The matrix multiply code in matprod using an LDOUBLE
    (long double) variable to accumulate sums, for improved accuracy.  
    On a SPARC system I tested on, operations on long doubles are 
    vastly slower than on doubles, so that the patch produces a 
    large slowdown rather than an improvement.  This is also an issue 
    for the "sum" function, which also uses an LDOUBLE to accumulate
    the sum.  Perhaps an ordinarly double should be used in these
    places, or perhaps the configuration script should define LDOUBLE 
    as double on architectures where long doubles are extraordinarily 

    Due to this issue, not defining MATPROD_CAN_BE_DONE_HERE at the
    start of array.c will disable this patch.


    Speeds up parentheses by making "(" a special operator whose
    argument is not evaluated, thereby bypassing the overhead of
    evalList.  Also slightly speeds up curly brackets by inlining
    a function that is stylistically better inline anyway.

    Relevant test script:  test-parens.r

    In the parens part of test-parens.r, the speedup is about 9%.


    Speeds up numerous operations by making PROTECT, UNPROTECT, etc.
    be mostly macros in the files in src/main.  This takes effect
    only for files that include Defn.h after defining the symbol
    USE_FAST_PROTECT_MACROS.  With these macros, code of the form
    v = PROTECT(...) must be replaced by PROTECT(v = ...).  

    Relevant timing test scripts:  all of them, but will look at test-em.r 

    On test-em.r, the speedup from this patch is about 9%.


    Speeds up some binary and unary arithmetic operations by, when
    possible, using the space holding one of the operands to hold
    the result, rather than allocating new space.  Though primarily
    a speed improvement, for very long vectors avoiding this allocation 
    could avoid running out of space.

    Relevant test script:  test-complex-expr.r

    On this test, the speedup is about 5% for scalar operands and about
    8% for vector operands.

    Issues:  There are some tricky issues with attributes, but I think
    I got them right.  This patch relies on NAMED being set correctly 
    in the rest of the code.  In case it isn't, the patch can be disabled 
    by not defining AVOID_ALLOC_IF_POSSIBLE at the top of arithmetic.c.


    Speeds up a^2 when a is a long vector by not checking for the
    special case of an exponent of 2 over and over again for every 
    vector element.

    Relevant test script:  test-square.r

    The time for squaring a long vector is reduced in this test by a
    factor of more than five.


    Speeds up the "sum" and "prod" functions by not checking for NA
    when na.rm=FALSE, and other detailed code improvements.

    Relevant test script:  test-sum-prod.r

    For sum, the improvement is about a factor of 2.5 when na.rm=FALSE,
    and about 10% when na.rm=TRUE.

    Issue:  See the discussion of patch-matprod regarding LDOUBLE.
    There is no change regarding this issue due to this patch, however.


    Speeds up the transpose operation (the "t" function) from detailed
    code improvements.

    Relevant test script:  test-transpose.r

    The improvement for 200x60 matrices is about a factor of two.
    There is little or no improvement for long row or column vectors.


    Speeds up arithmetic on vectors of the same length, or when on
    vector is of length one.  This is done with detailed code improvements.

    Relevant test script:  test-vec-arith.r

    On long vectors, the +, -, and * operators are sped up by about     
    20% when operands are the same length or one operand is of length one.

    Rather mysteriously, when the operands are not length one or the
    same length, there is about a 20% increase in time required, though
    this may be due to some strange C optimizer peculiarity or some 
    strange cache effect, since the C code for this is the same as before,
    with negligible additional overhead getting to it.  Regardless, this 
    case is much less common than equal lengths or length one.

    There is little change for the / operator, which is much slower than
    +, -, or *.


    Speeds up extraction of subsets of vectors or matrices (eg, v[10:20]
    or M[1:10,101:110]).  This is done with detailed code improvements.

    Relevant test script:  test-vec-subset.r

    There are lots of tests in this script.  The most dramatic improvement
    is for extracting many rows and columns of a large array, where the 
    improvement is by about a factor of four.  Extracting many rows from
    one column of a matrix is sped up by about 30%. 

    Changes unrelated to speed improvement:

    Fixes two latent bugs where the code incorrectly refers to NA_LOGICAL
    when NA_INTEGER is appropriate and where LOGICAL and INTEGER types
    are treated as interchangeable.  These cause no problems at the moment,
    but would if representations were changed.


    (Formerly part of patch-vec-subset)  This patch also speeds up
    extraction, and also replacement, of subsets of vectors or
    matrices, but focuses on the creation of the indexes rather than
    the copy operations.  Often avoids a duplication (see below) and
    eliminates a second scan of the subscript vector for zero
    subscripts, folding it into a previous scan at no additional cost.

    Relevant test script:  test-vec-subset.r

    Speeds up some operations with scalar or short vector indexes by
    about 10%.  Speeds up subscripting with a longer vector of
    positive indexes by about 20%.

    Issues:  The current code duplicates a vector of indexes when it
    seems unnecessary.  Duplication is for two reasons:  to handle
    the situation where the index vector is itself being modified in
    a replace operation, and so that any attributes can be removed, which 
    is helpful only for string subscripts, given how the routine to handle 
    them returns information via an attribute.  Duplication for the
    second reasons can easily be avoided, so I avoided it.  The first
    reason for duplication is sometimes valid, but can usually be avoided
    by first only doing it if the subscript is to be used for replacement
    rather than extraction, and second only doing it if the NAMED field
    for the subscript isn't zero.

    I also removed two layers of procedure call overhead (passing seven
    arguments, so not trivial) that seemed to be doing nothing.  Probably 
    it used to do something, but no longer does, but if instead it is 
    preparation for some future use, then removing it might be a mistake.

Software problems are best solved by writing code or patches in my opinion rather than discussing endlessly
Some other solutions to a BETTERRRR R
1) Complete Code Design Review
2) Version 3 - Tuneup
3) Better Documentation
4) Suing Revolution Analytics for the code - Hand over da code pardner

KXEN Update

Update from a very good data mining software company, KXEN –

  1. Longtime Chairman and founder Roger Haddad is retiring but would be a Board Member. See his interview with Decisionstats here (note images were hidden due to migration from .com to )
  2. New Members of Leadership are as-
John Ball, CEOJohn Ball
Chief Executive Officer

John Ball brings 20 years of experience in enterprise software, deep expertise in business intelligence and CRM applications, and a proven track record of success driving rapid growth at highly innovative companies.

Prior to joining KXEN, Mr. Ball served in several executive roles at, the leading provider of SaaS applications. Most recently, John served as VP & General Manager, Analytics and Reporting Products, where he spearheaded’s foray into CRM analytics and business intelligence. John also served as VP & General Manager, Service and Support Applications at, where he successfully grew the business to become the second largest and fastest growing product line at Before, Ball was founder and CEO of Netonomy, the leading provider of customer self-service solutions for the telecommunications industry. Ball also held a number of executive roles at Business Objects, including General Manager, Web Products, where delivered to market the first 3 versions of WebIntelligence. Ball has a master’s degree in electrical engineering from Georgia Tech and a master’s degree in electric

I hope John atleast helps build a KXEN application- there are only 2 data mining apps there on App Exchange. Also on the wish list  more social media presence, a Web SaaS/Amazon API for KXEN, greater presence in American/Asian conferences, and a solution for SME’s (which cannot afford the premium pricing of the flagship solution. An alliance with bigger BI vendors like Oracle, SAP or IBM  for selling the great social network analysis.

Bill Russell as Non Executive Chairman-

Bill Russell as Non-executive Chairman of the Board, effective July 16 2010. Russell has 30 years of operational experience in enterprise software, with a special focus on business intelligence, analytics, and databases.Russell held a number of senior-level positions in his more than 20 years at Hewlett-Packard, including Vice President and General Manager of the multi-billion dollar Enterprise Systems Group. He has served as Non-executive Chairman of the Board for Sylantro Systems Corporation, webMethods Inc., and Network Physics, Inc. and has served as a board director for Cognos Inc. In addition to KXEN, Russell currently serves on the boards of Saba, PROS Holdings Inc., Global 360, ParAccel Inc., and B.T. Mancini Company.

Xavier Haffreingue as senior vice president, worldwide professional services and solutions.
He has almost 20 years of international enterprise software experience gained in the CRM, BI, Web and database sectors. Haffreingue joins KXEN from software provider Axway where he was VP global support operations. Prior to Axway, he held various leadership roles in the software industry, including VP self service solutions at Comverse Technologies and VP professional services and support at Netonomy, where he successfully delivered multi-million dollar projects across Europe, Asia-Pacific and Africa. Before that he was with Business Objects and Sybase, where he ran support and services in southern Europe managing over 2,500 customers in more than 20 countries.

David Guercio  as senior vice president, Americas field operations. Guercio brings to the role more than 25 years experience of building and managing high-achieving sales teams in the data mining, business intelligence and CRM markets. Guercio comes to KXEN from product lifecycle management vendor Centric Software, where he was EVP sales and client services. Prior to Centric, he was SVP worldwide sales and client services at Inxight Software, where he was also Chairman and CEO of the company’s Federal Systems Group, a subsidiary of Inxight that saw success in the US Federal Government intelligence market. The success in sales growth and penetration into the federal government led to the acquisition of Inxight by Business Objects in 2007, where Guercio then led the Inxight sales organization until Business Objects was acquired by SAP. Guercio was also a key member of the management team and a co-founder at Neovista, an early pioneer in data mining and predictive analytics. Additionally, he held the positions of director of sales and VP of professional services at Metaphor Computer Systems, one of the first data extraction solutions companies, which was acquired by IBM. During his career, Guercio also held executive positions at Resonate and SiGen.

3) Venture Capital funding to fund expansion-

It has closed $8 million in series D funding to further accelerate its growth and international expansion. The round was led by NextStage and included participation from existing investors XAnge Capital, Sofinnova Ventures, Saints Capital and Motorola Ventures.

This was done after John Ball had joined as CEO.

4) Continued kudos from analysts and customers for it’s technical excellence.

KXEN was named a leader in predictive analytics and data mining by Forrester Research (1) and was rated highest for commercial deployments of social network analytics by Frost & Sullivan (2)

Also it became an alliance partner of Accenture- which is also a prominent SAS partner as well.

In Database Optimization-

In KXEN V5.1, a new data manipulation module (ADM) is provided in conjunction with scoring to optimize database workloads and provide full in-database model deployment. Some leading data mining vendors are only now beginning to offer this kind of functionality, and then with only one or two selected databases, giving KXEN a more than five-year head start. Some other vendors are only offering generic SQL generation, not optimized for each database, and do not provide the wealth of possible outputs for their scoring equations: For example, real operational applications require not only to generate scores, but decision probabilities, error bars, individual input contributions – used to derive reasons of decision and more, which are available in KXEN in-database scoring modules.

Since 2005, KXEN has leveraged databases as the data manipulation engine for analytical dataset generation. In 2008, the ADM (Analytical Data Management) module delivered a major enhancement by providing a very easy to use data manipulation environment with unmatched productivity and efficiency. ADM works as a generator of optimized database-specific SQL code and comes with an integrated layer for the management of meta-data for analytics.

KXEN Modeling Factory- (similar to SAS’s recent product Rapid Predictive Modeler and

KXEN Modeling Factory (KMF) has been designed to automate the development and maintenance of predictive analytics-intensive systems, especially systems that include large numbers of models, vast amounts of data or require frequent model refreshes. Information about each project and model is monitored and disseminated to ensure complete management and oversight and to facilitate continual improvement in business performance.

Main Functions

Schedule: creation of the Analytic Data Set (ADS), setup of how and when to score, setup of when and how to perform model retraining and refreshes …

: Monitormodel execution over time, Track changes in model quality over time, see how useful one variable is by considering its multiple instance in models …

: Rather than having to wade through pages of event logs, KMF Department allows users to manage by exception through notifications.

Other products from KXEN have been covered here before , including Structural Risk Minimization-

Thats all for the KXEN update- all the best to the new management team and a splendid job done by Roger Haddad in creating what is France and Europe’s best known data mining company.

Note- Source –

Event: Predictive analytics with R, PMML and ADAPA


The September meeting is at the Oracle campus. (This is next door to the Oracle towers, so there is plenty of free parking.) The featured talk is from Alex Guazzelli (Vice President – Analytics, Zementis Inc.) who will talk about “Predictive analytics with R, PMML and ADAPA”.

* 6:15 – 7:00 Networking and Pizza (with thanks to Revolution Analytics)
* 7:00 – 8:00 Talk: Predictive analytics with R, PMML and ADAPA
* 8:00 – 8:30 General discussion

Talk overview:

The rule in the past was that whenever a model was built in a particular development environment, it remained in that environment forever, unless it was manually recoded to work somewhere else. This rule has been shattered with the advent of PMML (Predictive Modeling Markup Language). By providing a uniform standard to represent predictive models, PMML allows for the exchange of predictive solutions between different applications and various vendors.

Once exported as PMML files, models are readily available for deployment into an execution engine for scoring or classification. ADAPA is one example of such an engine. It takes in models expressed in PMML and transforms them into web-services. Models can be executed either remotely by using web-services calls, or via a web console. Users can also use an Excel add-in to score data from inside Excel using models built in R.

R models have been exported into PMML and uploaded in ADAPA for many different purposes. Use cases where clients have used the flexibility of R to develop and the PMML standard combined with ADAPA to deploy range from financial applications (e.g., risk, compliance, fraud) to energy applications for the smart grid. The ability to easily transition solutions developed in R to the operational IT production environment helps eliminate the traditional limitations of R, e.g. performance for high volume or real-time transactional systems and memory constraints associated with large data sets.

Speaker Bio:

Dr. Alex Guazzelli has co-authored the first book on PMML, the Predictive Model Markup Language which is the de facto standard used to represent predictive models. The book, entitled PMML in Action: Unleashing the Power of Open Standards for Data Mining and Predictive Analytics, is available on As the Vice President of Analytics at Zementis, Inc., Dr. Guazzelli is responsible for developing core technology and analytical solutions under ADAPA, a PMML-based predictive decisioning platform that combines predictive analytics and business rules. ADAPA is the first system of its kind to be offered as a service on the cloud.
Prior to joining Zementis, Dr. Guazzelli was involved in not only building but also deploying predictive solutions for large financial and telecommunication institutions around the globe. In academia, Dr. Guazzelli worked with data mining, neural networks, expert systems and brain theory. His work in brain theory and computational neuroscience has appeared in many peer reviewed publications. At Zementis, Dr. Guazzelli and his team have been involved in a myriad of modeling projects for financial, health-care, gaming, chemical, and manufacturing industries.

Dr. Guazzelli holds a Ph.D. in Computer Science from the University of Southern California and a M.S and B.S. in Computer Science from the Federal University of Rio Grande do Sul, Brazil.

Interview Stephanie McReynolds Director Product Marketing, AsterData

Here is an interview with Stephanie McReynolds who works as as Director of Product Marketing with AsterData. I asked her a couple of questions about the new product releases from AsterData in analytics and MapReduce.

Ajay – How does the new Eclipse Plugin help people who are already working with huge datasets but are new to AsterData’s platform?

Stephanie- Aster Data Developer Express, our new SQL-MapReduce development plug-in for Eclipse, makes MapReduce applications easy to develop. With Aster Data Developer Express, developers can develop, test and deploy a complete SQL-MapReduce application in under an hour. This is a significant increase in productivity over the traditional analytic application development process for Big Data applications, which requires significant time coding applications in low-level code and testing applications on sample data.

Ajay – What are the various analytical functions that are introduced by you recently- list say the top 10.

Stephanie- At Aster Data, we have an intense focus on making the development process easier for SQL-MapReduce applications. Aster Developer Express is a part of this initiative, as is the release of pre-defined analytic functions. We recently launched both a suite of analytic modules and a partnership program dedicated to delivering pre-defined analytic functions for the Aster Data nCluster platform. Pre-defined analytic functions delivered by Aster Data’s engineering team are delivered as modules within the Aster Data Analytic Foundation offering and include analytics in the areas of pattern matching, clustering, statistics, and text analysis– just to name a few areas. Partners like Fuzzy Logix and Cobi Systems are extending this library by delivering industry-focused analytics like Monte Carlo Simulations for Financial Services and geospatial analytics for Public Sector– to give you a few examples.

Ajay – So okay I want to do a K Means Cluster on say a million rows (and say 200 columns) using the Aster method. How do I go about it using the new plug-in as well as your product.

Stephanie- The power of the Aster Data environment for analytic application development is in SQL-MapReduce. SQL is a powerful analytic query standard because it is a declarative language. MapReduce is a powerful programming framework because it can support high performance parallel processing of Big Data and extreme expressiveness, by supporting a wide variety of programming languages, including Java, C/C#/C++, .Net, Python, etc. Aster Data has taken the performance and expressiveness of MapReduce and combined it with the familiar declarativeness of SQL. This unique combination ensures that anyone who knows standard SQL can access advanced analytic functions programmed for Big Data analysis using MapReduce techniques.

kMeans is a good example of an analytic function that we pre-package for developers as part of the Aster Data Analytic Foundation. What does that mean? It means that the MapReduce portion of the development cycle has been completed for you. Each pre-packaged Aster Data function can be called using standard SQL, and executes the defined analytic in a fully parallelized manner in the Aster Data database using MapReduce techniques. The result? High performance analytics with the expressiveness of low-level languages accessed through declarative SQL.

Ajay – I see an an increasing focus on Analytics. Is this part of your product strategy and how do you see yourself competing with pure analytics vendors.

Stephanie – Aster Data is an infrastructure provider. Our core product is a massively parallel processing database called nCluster that performs at or beyond the capabilities of any other analytic database in the market today. We developed our analytics strategy as a response to demand from our customers who were looking beyond the price/performance wars being fought today and wanted support for richer analytics from their database provider. Aster Data analytics are delivered in nCluster to enable analytic applications that are not possible in more traditional database architectures.

Ajay – Name some recent case studies in Analytics of implementation of MR-SQL with Analytical functions

Stephanie – There are three new classes of applications that Aster Data Express and Aster Analytic Foundation support: iterative analytics, prediction and optimization, and ad hoc analysis.

Aster Data customers are uncovering critical business patterns in Big Data by performing hypothesis-driven, iterative analytics. They are exploring interactively massive volumes of data—terabytes to petabytes—in a top-down deductive manner. ComScore, an Aster Data customer that performs website experience analysis is a good example of an Aster Data customer performing this type of analysis.

Other Aster Data customers are building applications for prediction and optimization that discover trends, patterns, and outliers in data sets. Examples of these types of applications are propensity to churn in telecommunications, proactive product and service recommendations in retail, and pricing and retention strategies in financial services. Full Tilt Poker, who is using Aster Data for fraud prevention is a good example of a customer in this space.

The final class of application that I would like to highlight is ad hoc analysis. Examples of ad hoc analysis that can be performed includes social network analysis, advanced click stream analysis, graph analysis, cluster analysis and a wide variety of mathematical, trigonometry, and statistical functions. LinkedIn, whose analysts and data scientists have access to all of their customer data in Aster Data are a good example of a customer using the system in this manner.

While Aster Data customers are using nCluster in a number of other ways, these three new classes of applications are areas in which we are seeing particularly innovative application development.


Stephanie McReynolds is Director of Product Marketing at Aster Data, where she is an evangelist for Aster Data’s massively parallel data-analytics server product. Stephanie has over a decade of experience in product management and marketing for business intelligence, data warehouse, and complex event processing products at companies such as Oracle, Peoplesoft, and Business Objects. She holds both a master’s and undergraduate degree from Stanford University.

Dryad- Microsoft's answer to MR

While reading across the internet I came across Microsoft’s version to MapReduce called Dryad- which has been around for some time, but has not generated quite the buzz that Hadoop or MapReduce are doing.


DryadLINQ is a simple, powerful, and elegant programming environment for writing large-scale data parallel applications running on large PC clusters.


New! An academic release of Dryad/DryadLINQ is now available for public download.

The goal of DryadLINQ is to make distributed computing on large compute cluster simple enough for every programmers. DryadLINQ combines two important pieces of Microsoft technology: the Dryad distributed execution engine and the .NET Language Integrated Query (LINQ).

Dryad provides reliable, distributed computing on thousands of servers for large-scale data parallel applications. LINQ enables developers to write and debug their applications in a SQL-like query language, relying on the entire .NET library and using Visual Studio.

DryadLINQ translates LINQ programs into distributed Dryad computations:

  • C# and LINQ data objects become distributed partitioned files.
  • LINQ queries become distributed Dryad jobs.
  • C# methods become code running on the vertices of a Dryad job.

DryadLINQ has the following features:

  • Declarative programming: computations are expressed in a high-level language similar to SQL
  • Automatic parallelization: from sequential declarative code the DryadLINQ compiler generates highly parallel query plans spanning large computer clusters. For exploiting multi-core parallelism on each machine DryadLINQ relies on the PLINQ parallelization framework.
  • Integration with Visual Studio: programmers in DryadLINQ take advantage of the comprehensive VS set of tools: Intellisense, code refactoring, integrated debugging, build, source code management.
  • Integration with .Net: all .Net libraries, including Visual Basic, and dynamic languages are available.
  • and
  • Conciseness: the following line of code is a complete implementation of the Map-Reduce computation framework in DryadLINQ:
    • public static IQueryable<R>
      MapReduce<S,M,K,R>(this IQueryable<S> source,
      Expression<Func<S,IEnumerable<M>>> mapper,
      Expression<Func<M,K>> keySelector,
      Expression<Func<K,IEnumerable<M>,R>> reducer)
      return source.SelectMany(mapper).GroupBy(keySelector, reducer);



    The Dryad Project is investigating programming models for writing parallel and distributed programs to scale from a small cluster to a large data-center.


    New! An academic release of DryadLINQ is now available for public download.

    Dryad is an infrastructure which allows a programmer to use the resources of a computer cluster or a data center for running data-parallel programs. A Dryad programmer can use thousands of machines, each of them with multiple processors or cores, without knowing anything about concurrent programming.

    The Structure of Dryad Jobs

    A Dryad programmer writes several sequential programs and connects them using one-way channels. The computation is structured as a directed graph: programs are graph vertices, while the channels are graph edges. A Dryad job is a graph generator which can synthesize any directed acyclic graph. These graphs can even change during execution, in response to important events in the computation.

    Dryad is quite expressive. It completely subsumes other computation frameworks, such as Google’s map-reduce, or the relational algebra. Moreover, Dryad handles job creation and management, resource management, job monitoring and visualization, fault tolerance, re-execution, scheduling, and accounting.

    The Dryad Software Stack

    As a proof of Dryad’s versatility, a rich software ecosystem has been built on top Dryad:

    • SSIS on Dryad executes many instances of SQL server, each in a separate Dryad vertex, taking advantage of Dryad’s fault tolerance and scheduling. This system is currently deployed in a live production system as part of one of Microsoft’s AdCenter log processing pipelines.
    • DryadLINQ generates Dryad computations from the LINQ Language-Integrated Query extensions to C#.
    • The distributed shell is a generalization of the pipe concept from the Unix shell in three ways. If Unix pipes allow the construction of one-dimensional (1-D) process structures, the distributed shell allows the programmer to build 2-D structures in a scripting language. The distributed shell generalizes Unix pipes in three ways:
      1. It allows processes to easily connect multiple file descriptors of each process — hence the 2-D aspect.
      2. It allows the construction of pipes spanning multiple machines, across a cluster.
      3. It virtualizes the pipelines, allowing the execution of pipelines with many more processes than available machines, by time-multiplexing processors and buffering results.
    • Several languages are compiled to distributed shell processes. PSQL is an early version, recently replaced with Scope.


    Dryad: Distributed Data-Parallel Programs from Sequential Building Blocks
    Michael Isard, Mihai Budiu, Yuan Yu, Andrew Birrell, and Dennis Fetterly
    European Conference on Computer Systems (EuroSys), Lisbon, Portugal, March 21-23, 2007

    Video of a presentation on Dryad at the Google Campus, given by Michael Isard, Nov 1, 2007.

    Also interesting to read-

    Why does Dryad use a DAG?

    he basic computational model we decided to adopt for Dryad is the directed-acyclic graph (DAG). Each node in the graph is a computation, and each edge in the graph is a stream of data traveling in the direction of the edge. The amount of data on any given edge is assumed to be finite, the computations are assumed to be deterministic, and the inputs are assumed to be immutable. This isn’t by any means a new way of structuring a distributed computation (for example Condor had DAGMan long before Dryad came along), but it seemed like a sweet spot in the design space given our other constraints.

    So, why is this a sweet spot? A DAG is very convenient because it induces an ordering on the nodes in the graph. That makes it easy to design scheduling policies, since you can define a node to be ready when its inputs are available, and at any time you can choose to schedule as many ready nodes as you like in whatever order you like, and as long as you always have at least one scheduled you will continue to make progress and never deadlock. It also makes fault-tolerance easy, since given our determinism and immutability assumptions you can backtrack as far as you want in the DAG and re-execute as many nodes as you like to regenerate intermediate data that has been lost or is unavailable due to cluster failures.


      MapReduce Analytics Apps- AsterData's Developer Express Plugin

      AsterData continues to wow with it’s efforts on bridging MapReduce and Analytics, with it’s new Developer Express plug-in for Eclipse. As any Eclipse user knows, that greatly improves ability to write code or develop ( similar to creating Android apps if you have tried to). I did my winter internship at AsterData last December last year in San Carlos, and its an amazing place with giga-level bright people.

      Here are some details ( Note I plan to play a bit more on the plugin on my currently downUbuntu on this and let you know)

      Aster Data Developer Express provides an integrated set of tools for development of SQL and MapReduce analytics for Aster Data nCluster, a massively parallel database with an integrated analytics engine.

      The Aster Data Developer Express plug-in for Eclipse enables developers to easily create new analytic application projects with the help of an intuitive set of wizards, immediately test their applications on their desktop, and push down their applications into the nCluster database with a single click.

      Using Developer Express, analysts can significantly reduce the complexity and time needed to create advanced analytic applications so that they can more rapidly deliver deeper and richer analytic insights from their data.

      and from the Press Release

      Now, any developer or analyst that is familiar with the Java programming language can complete a rich analytic application in under an hour using the simple yet powerful Aster Data Developer Express environment in Eclipse. Aster Data Developer Express delivers both rapid development and local testing of advanced analytic applications for any project, regardless of size.

      The free, downloadable Aster Data Developer Express IDE now brings the power of SQL-MapReduce to any organization that is looking to build richer analytic applications that can leverage massive data volumes. Much of the MapReduce coding, including programming concepts like parallelization and distributed data analysis, is addressed by the IDE without the developer or analyst needing to have expertise in these areas. This simplification makes it much easier for developers to be successful quickly and eliminates the need for them to have any deep knowledge of the MapReduce parallel processing framework. Google first published MapReduce in 2004 for parallel processing of big data sets. Aster Data has coupled SQL with MapReduce and brought SQL-MapReduce to market, making it significantly easier for any organization to leverage the power of MapReduce. The Aster Developer Express IDE simplifies application development even further with an intuitive point-and-click development environment that speeds development of rich analytic applications. Applications can be validated locally on the desktop or ultimately within Aster Data nCluster, a massive parallel processing (MPP) database with a fully integrated analytics engine that is powered by MapReduce—known as a data-analytics server.

      Rich analytic applications that can be easily built with Aster Data’s downloadable IDE include:

      Iterative Analytics: Uncovering critical business patterns in your data requires hypothesis-driven, iterative analysis.  This class of applications is defined by the exploratory navigation of massive volumes of data in a top-down, deductive manner.  Aster Data’s IDE makes this easy to develop and to validate the algorithms and functions required to deliver these advanced analytic applications.

      Prediction and Optimization: For this class of applications, the process is inductive. Rather than starting with a hypothesis, developers and analysts can easily build analytic applications that discover the trends, patterns, and outliers in data sets.  Examples include propensity to churn in telecommunications, proactive product and service recommendations in retail, and pricing and retention strategies in financial services.

      Ad Hoc Analysis: Examples of ad hoc analysis that can be performed includes social network analysis, advanced click stream analysis, graph analysis, cluster analysis, and a wide variety of mathematical, trigonometry, and statistical functions.

      “Aster Data’s IDE and SQL-MapReduce significantly eases development of advanced analytic applications on big data. We have now built over 350 analytic functions in SQL-MapReduce on Aster Data nCluster that are available for customers to purchase,” said Partha Sen, CEO and Founder of Fuzzy Logix. “Aster Data’s implementation of MapReduce with SQL-MapReduce goes beyond the capabilities of general analytic development APIs and provides us with the excellent control and flexibility needed to implement even the most complex analytic algorithms.”

      Richer analytics on big data volumes is the new competitive frontier. Organizations have always generated reports to guide their decision-making. Although reports are important, they are historical sets of information generally arranged around predefined metrics and generated on a periodic basis.

      Advanced analytics begins where reporting leaves off. Reporting often answers historical questions such as “what happened?” However, analytics addresses “why it happened” and, increasingly, “what will happen next?” To that end, solutions like Aster Data Developer Express ease the development of powerful ad hoc, predictive analytics and enables analysts to quickly and deeply explore terabytes to petabytes of data.
      “We are in the midst of a new age in analytics. Organizations today can harness the power of big data regardless of scale or complexity”, said Don Watters, Chief Data Architect for MySpace. “Solutions like the Aster Data Developer Express visual development environment make it even easier by enabling us to automate aspects of development that currently take days, allowing us to build rich analytic applications significantly faster. Making Developer Express openly available for download opens the power of MapReduce to a broader audience, making big data analytics much faster and easier than ever before.”

      “Our delivery of SQL coupled with MapReduce has clearly made it easier for customers to build highly advanced analytic applications that leverage the power of MapReduce. The visual IDE, Aster Data Developer Express, introduced earlier this year, made application development even easier and the great response we have had to it has driven us to make this open and freely available to any organization looking to build rich analytic applications,” said Tasso Argyros, Founder and CTO, Aster Data. “We are excited about today’s announcement as it allows companies of all sizes who need richer analytics to easily build powerful analytic applications and experience the power of MapReduce without having to learn any new skills.”

      You can have a look here at

      R Oracle Data Mining

      Here is a new package called R ODM and it is an interface to do Data Mining via Oracle Tables through R. You can read more here and here . Also there is a contest for creative use of R and ODM.

      R Interface to Oracle Data Mining

      The R Interface to Oracle Data Mining ( R-ODM) allows R users to access the power of Oracle Data Mining’s in-database functions using the familiar R syntax. R-ODM provides a powerful environment for prototyping data analysis and data mining methodologies.

      R-ODM is especially useful for:

      • Quick prototyping of vertical or domain-based applications where the Oracle Database supports the application
      • Scripting of “production” data mining methodologies
      • Customizing graphics of ODM data mining results (examples: classificationregressionanomaly detection)

      The R-ODM interface allows R users to mine data using Oracle Data Mining from the R programming environment. It consists of a set of function wrappers written in source R language that pass data and parameters from the R environment to the Oracle RDBMS enterprise edition as standard user PL/SQL queries via an ODBC interface. The R-ODM interface code is a thin layer of logic and SQL that calls through an ODBC interface. R-ODM does not use or expose any Oracle product code as it is completely an external interface and not part of any Oracle product. R-ODM is similar to the example scripts (e.g., the PL/SQL demo code) that illustrates the use of Oracle Data Mining, for example, how to create Data Mining models, pass arguments, retrieve results etc.

      R-ODM is packaged as a standard R source package and is distributed freely as part of the R environment’s Comprehensive R Archive Network ( CRAN). For information about the R environment, R packages and CRAN, see


      Present and win an Apple iPod Touch!
      The BI, Warehousing and Analytics (BIWA) SIG is giving an Apple iPOD Touch to the best new presenter. Be part of the TechCast series and get a chance to win!

      Consider highlighting a creative use of R and ODM.

      BIWA invites all Oracle professionals (experts, end users, managers, DBAs, developers, data analysts, ISVs, partners, etc.) to submit abstracts for 45 minute technical webcasts to our Oracle BIWA (IOUG SIG) Community in our Wednesday TechCast series. Note that the contest is limited to new presenters to encourage fresh participation by the BIWA community.

      Also an interview with Oracle Data Mining head, Charlie Berger