Skip to Main Content

Breadcrumb

Question and Answer

Tom Kyte

Thanks for the question, oded.

Asked: March 28, 2009 - 1:04 pm UTC

Last updated: March 31, 2009 - 8:12 am UTC

Version: 10.2.0

Viewed 1000+ times

You Asked

Hi Tom,
I would like to read your opinion in regard to:

http://hinchcliffe.org/archive/2009/03/17/16712.aspx

Non-relational databases. Tony Bain over at Read/Write Web recently asked "Is The Relational Database Doomed?" While it's far too soon to declare the demise of the workhorse relational database that's the bedrock of so many application stacks, there a large number of promising alternatives emerging. Why get rid of the traditional relational database? Certain application designs can greatly benefit from the advantages of document or resource-centric storage approaches. Performance in particular can be much higher with non-relational databases; there are often surprisingly low ceilings to the scale of relational databases, even with clustering and grid computing. And then there is abstraction impedance, which not only can create a lot more overhead when programming but also hurts run-time performance by maintaining several different representations of the data at one time during a service request. Promising non-relational solutions include CouchDB, which I'm starting to see in more and more products, as well as Amazon SimpleDB, Drizzle (from the MySql folks), Mongo, and Scalaris. While many applications will continue to get along just fine with relational databases and object-relational mapping, this is the first time that mainstream database alternatives are readily available for those that are increasingly in need of them.

Thanks, Oded.

and Tom said...

I fail to see how a very narrow specific application profile (such as the ones that meet the criteria of "Certain application designs can greatly benefit from the advantages of document or resource-centric storage approaches." would spell the end of relational databases?

Many (I would say most) conventional applications do NOT benefit from the document or resource-centric storage approaches.

When I see the first payroll, finance, order entry, manufacturing, banking, etc etc etc.... application (that works and works well and doesn't lose anything) - then we can have a discussion on this.

Does this model (document, collection oriented - like a shopping cart) make sense for some application profiles? Sure.


Also, ... this is the first time that mainstream database alternatives are readily available for those that are increasingly in need of them. ... Well, you know, Relational was 'invented' in 1977, there were mainstream database alternatives before Relational, there were mainstream database alternatives DURING relational. Relational is neither the first, last or only mainstream database alternative out there.

Just one of the most widely used - due in large part to its inherit flexibility.

Rating

  (5 ratings)

Is this answer out of date? If it is, please let us know via a Comment

Comments

relational

A reader, March 30, 2009 - 12:19 pm UTC


The "invention" happened a little before 1977...

Tomas, March 30, 2009 - 4:51 pm UTC

A slight correction; "relational", as in "relational theory", was invented during the 1960s by E.F. Codd and officially published in the Communications of the ACM in 1970 ("A Relational Model of Data for Large Shared Data Banks", CACM, Vol 13, No. 6, June 1970 - http://portal.acm.org/citation.cfm?doid=362384.362685 ).

That Larry and c/o took advantage of that and started to build something rather spectacular in 1977 is another matter...


Tom Kyte
March 30, 2009 - 6:07 pm UTC

yes, you are correct... I was talking about the physical availability of something to hold data...

deja vu

Alberto Dell'Era, March 31, 2009 - 3:59 am UTC

I still remember reading, back in 1996, a paper called "hitting the relational wall" that was more or less predicting the death of relational database as well; as a young graduate, I was keen on believing its assertions. Of course, its predictions turned out as being far from accurate ... lesson learned.
Tom Kyte
March 31, 2009 - 8:12 am UTC

I remember a paper by M Stonebraker around that time stating the RDBMS is dead, Object Relational is the only way forward. More recently, he wrote, the RDBMS is dead - columnar data stores are the way forward.

People are always calling it dead, the same is true of every 'technology' as soon as it is invented and rolled out - by definition - it is a legacy product. You can see people calling java a 'dead' language, it is not any more dead than cobol. It is just legacy. And it'll be around for a while I thing.

Just like cobol.

Padmanabham Tadepalli, March 31, 2009 - 11:17 pm UTC

Relational Model was developed based on strong
mathematical model i.e. Relational Algebra. It will stay
until an another model of that of calibre superseeds it.

The fact that other models survive doesnt mean the end of
relational model.


Strong connection.

Girish Singhal, May 13, 2009 - 9:05 am UTC

I would like to relate existing relational database model concept that Oracle implements with that of the one some brahmin clans use to record details of deads (humans) and have carried forward the tradition for more than 10s of 1000s of years. These groups are found in cities or towns like Allahabad, Banaras (Varanasi), Gaya etc... (primarily the cities through which rivers like ganges etc flow. If one goes to the designated places on these rivers in these cities to perform some last rites (immerse the ashes or remains that are left after cremation) for someone in family who dies, one can come to know details (names etc) of his/her ancesters for many bygone centuries.

I think that is one of the best example which states that data modelling using relational techniques is here to stay.
And it is a different matter that many other techniques would be developed that may be suitable to some specific type of applications. Now having said that, do we have a data model that we can use to model out the evolution of universe....

It is a sincere comment and I don't intend to offend anybody or their faith.