Additional NoSQL databses supported by SAS

[Samka]

Are there plans for SAS to support additional  data  vendors?   I am particularly curious about MongoDb, which is one of the most ubiquitous  NoSQL databases, with perhaps the largest reservoir of programmers for a Hadoop solution.

Is this strictly a technical question, or is it primarily a business question, from both the side of SAS and outside vendors?

[SASKiwi]

A quick search for MongoDb indicates there are ODBC drivers available. This means SAS/ACCESS to ODBC is a possibility. You could check with SAS Tech Support to see if they are aware of anyone using ODBC with MongoDb. 

[Samka]

I assume the ODBC drivers are developed by SAS, right?  Is that a technical or business decision, do you think?  

When I go to:
http://support.sas.com/documentation/cdl/en/hadoopov/68100/PDF/default/hadoopov.pdf


I found on page 9:

A commercial Hadoop distribution is the collection of Hadoop components (such as HDFS, Hive, and MapReduce) that is provided by a vendor. Many commercial Hadoop distributions include additional proprietary software. SAS supports commercial Hadoop distributions from Cloudera, Hortonworks, IBM InfoSphere BigInsights, MapR, and Pivota

[LinusH]

No, SAS only develops ODBC drivers for its own data sources.

ODBC drivers are usually developed by the DB vendor, or a 3rd party.

[SASKiwi]

No by third parties as @LinusH says. A Google search popped up several possibilities, for example:

http://www.simba.com/drivers/mongodb-odbc-jdbc/?gclid=CJfS3-61t80CFVMAvAodnoMKNw

[Samka]

I appreciate your replies--at the risk my appearing too dense, what does this mean in real life?  Does it mean that other DBs have not developed drivers for SAS, which makes little sense to me from a business standpoint, unless there is something about certain databases that makes such development not technically feasible?   Or, is it the absence of some sort of business relationship with SAS that would be a necessary prerequisite to that kind of development?

In other words, how does this statement from SAS fit with your answer?

SAS supports commercial Hadoop distributions from Cloudera, Hortonworks, IBM InfoSphere BigInsights, MapR, and Pivota

[SASKiwi]

Well I guess it is impossible for SAS to develop custom interfaces to all databases, hence they support just those for which there is sufficient commercial demand. For those databases where a custom interface does not exist a generic interface like ODBC (Open Database Connectivity) is still possible.

We use SAS/ACCESS to ODBC to work with SQL Server. There is nothing wrong with this approach and many ODBC solutions work very well. 

[Samka]

You have described the interface with a RDB rather than NoSQL. While MongoDB is not on the scale of SQL Server in market penetration, (big data probably is not there for any platform yet) it is one of the most widely used and one of the easier to find developers for. So you have gotten to the nub of my question: why all the other NoSQL but not the largest? As an analyst that question intrigues me--and I am still not sure whether the answer is a technical or a business one--perhaps if I had better technical skills I would know. 🙂

[SASKiwi]

That is more of a business question that should be directed at your SAS Account Manager.

[LinusH]

What is your business case?
That's the one question SAS need to know about their customers to be confident in investing in a tailored access engine.
I don't have much knowledge about MongoDB as such, bet generally speaking, the wording of NoSQL implies the the SW is not (only) relational. SAS on the other hand a typical relational SW so any direct connections will loose in translation. Perhaps there are other non-SQL based ways of interacting?

[Samka]

Linus, SAS lists several NoSQL DB that or supports. I posted the list from SAS documents. It does not list the most common NoSQL DB with the most developers available and trained. I am simply in my naïveté asking why:
1. Is it a business decision because the two companies have not agreed to work on it together?
2. Is there a technical reason why not?

It would seem to me, ignorant of both business arrangements and technical details, that the dominant NoSQL DB and the dominant statistical package would collaborate -- what am I missing?

Seems pretty straightforward but I must not be expressing it well.

[TomKari]

Interesting conversation!

My two cents worth is that SAS has to make resource allocation decisions on which third-party data management products to provide additional support for, beyond "generic" ODBC interfaces. I'm not familiar with the NoSQL domain, but if you feel that SAS should invest in MondoDB, you can certainly make a case to your SAS office.

In my opinion, if you can round up a bunch of SAS clients who also use MongoDB, and who would benefit from additional features that might be provided by an extended SAS interface, I think that SAS would be very receptive. Basically, you simply have to demonstrate that there's a customer base large enough to make it worth their while.

On the other hand, if MongoDB is terrifically popular, but all of the people who use it except for you have no interest in SAS, that will make it a very tough sell.

Good luck,

  Tom

[Samka]

Tom, so your conclusion is that it is a business decision by SAS rather than a technical
obstacle. Thanks

[LinusH]

Of course it's a business decision. SAS is a commercial enterprise.
And I don't really know what list you refer to. SAS/ACCESS interfaces to Hadoop distributions targets the relational parts, like Hive.

[Samka]

Page 9 of the attached document is what I am basing it on. Thanks.