BT

Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ
News Articles Presentations Podcasts Guides

Topics

Choose your language

QCon New York

Find real-world practical inspiration from the world’s most innovative software leaders. Attend in-person.
QCon San Francisco

Learn what's next in software from world-class leaders pushing the boundaries. Attend in-person or get video-only pass to recordings.
The Software Architects' Newsletter

Your monthly guide to all the topics, technologies and techniques that every professional needs to know about. Subscribe for free.

InfoQ Homepage News The RDBMS is not enough.

Building an Efficient Cloud Application Architecture (Live Webinar June 22, 2023) - Save Your Seat

The RDBMS is not enough.

This item in japanese

Bookmarks

Nov 26, 2007 5 min read

Write for InfoQ

Join a community of experts. Increase your visibility.
Grow your career.Learn more
While Relational Databases fit a client-server model, in a world of services new solutions are needed. RDBMS are subject to scalability issues: How to create redundancy, parallelism ?
[Relation Databases] become a single point of failure. In particular, replication is not trivial. To understand why, consider the problem of having two database servers that need to have identical data. Having both servers for reading and writing data makes it difficult to synchronize changes. Having one master server and another slave is bad too, because the master has to take all the heat when users are writing information.
In addition, Assaf Arkin also believes that write consistency is the reason RDBMS are imploding under their own weight.
Features like referential integrity, constraints and atomic updates are really important in the client-server world, but irrelevant in a world of services.
Those are typical issues that Document Oriented Distributed Databases are notably trying to address.
Damien Katz, software engineer at MySQL introduced the four pillars of Data Management:
  • Save: Data saving should be secure (ie ACID), permanent and efficient.
  • See: Data should be available for easy retrieval, integrate simple reporting methods and provide a (fulltext) search.
  • Secure: Compartmentalization of data, allow ssl connection, assign users, groups and roles to data...
  • Share: Be distributed, On and Offline.
With CouchDB, Damien is implementing those four pillars.

What CouchDB is
  • A document database server, accessible via a RESTful JSON API.
  • Ad-hoc and schema-free with a flat address space.
  • Distributed, featuring robust, incremental replication with bi-directional conflict detection and management.
  • Query-able and index-able, featuring a table oriented reporting engine that uses Javascript as a query language.
What CouchDB is not
  • A relational database.
  • A replacement for relational databases.
  • An object-oriented database. More specifically, CouchDB is not meant to function as a seamless persistence layer for an OO programming language.
Inspired by CouchDB and the notion that you insert documents into the database and then define views for querying, Anthony Eden started to write his own Document-Oriented Database: RDDB. An exhaustive review is already available.

The features of RDDB at the moment are:
  • Documents are simply collections of name/value pairs.
  • Views can be defined with Ruby code.
  • A reduce block can be defined to reduce the initial mapped data from a view.
  • Views can be materialized to improve query performance.
  • Datastores/Viewstores/Materialization stores are pluggable. Current implementations are RAM, partitioned files/file system and Amazon S3.
  • Distributed materialization may work, but it's going to be rewritten.
InfoQ had the chance to catch up with Anthony and talk about RDDB, CouchDB and RDBMS.

Firstly what lead you to start working on RDDB, at Rejectconf you were talking about a research project?

I consider RDDB to be a personal research project. Over the last year I've been heavily involved in analytical systems, developing data warehouses and the like. I've also been using Amazon's web services. RDDB will hopefully allow me to put the two together at some point so I can have an analytical database that runs on EC2 and S3. That's my primary goal and the driving factor behind the creation of RDDB.

In your daily job you're exposed to Data integration issues; Do you think Document Oriented Distributed Database are underused today and that they will be adopted more and more?

I'm not sure yet. There is a lot of history behind relational databases and they've had a lot of time to mature. On the one hand this makes them the obvious choice for operational systems since they can be trusted. On the other hand relational databases aren't necessarily the best choice for all types of data storage and lookup, so there are opportunities for new data stores, I'm just not sure of document oriented databases are going to be it - I think it will largely depend on their scalability and their ability to deal with massive amounts of documents without degradation of performance.

Is there still a place for RDBMS in a world of services model? While referential integrity, atomic updates and constraints makes sense in a client-server world, is it still relevant in a world of services?

It's still the standard by which others are judged, so I don't see relational databases going anywhere anytime soon. Ultimately I think we may be able to move past needing atomic updates if we have databases that are temporal in nature, thus removing the need for updates of any sort. Referential integrity can also probably go if we move to an environment where everything that is absolutely necessary is included in a resource and systems become more tolerant to missing links. Constraints will probably always be useful and perhaps they may become even richer with the ability to define logic for constraints.

How do you compare RDDB to CouchDB? (I know you're in early early stage of development, CouchDB is also). What would be the advantage of using RDDB over CouchDB Ruby binding?

I think I can answer these two together. Since CouchDB is written in Erlang and RDDB is written in Ruby, for a Ruby developer RDDB will be much more hackable. CouchDB uses the language features of Erlang for interprocess communication during distributed processing whereas Ruby relies on libraries, such as Rinda and the Ruby SQS library. For a Ruby developer the cost of getting RDDB up and running is going to be significantly less than CouchDB since all you have to do with RDDB is use RubyGems to install. The views in RDDB are written in Ruby whereas the CouchDB views are JSON (at least for now). I think that at the moment RDDB is more pluggable with different implementations for document store, view store and materialization store (all of which support RAM, file system, S3 storage). RDDB also has different implementations for materialization (such as local, Rinda and EC2), as well as threaded and non-threaded materializers.

We wrote an article about ActiveWarehouse a while back, how is the project going? Is it used in the enterprise?

ActiveWarehouse has been quiet lately. I believe that most of the work and usage is on the ETL side with the ActiveWarehouse ETL library. My goal is to release the 1.0 version of ActiveWarehouse ETL in the near future. As for the Rails plugin, it definitely needs more work on the display side before it can get raised up to a 1.0 version. Some people have expressed interest in revising the user interface code so we'll see where that leads.

Rate this Article

Adoption
Style

This content is in the AWS topic

Related Topics:

Hello stranger!

You need to Register an InfoQ account or or login to post comments. But there's so much more behind being registered.

Get the most out of the InfoQ experience.

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Community comments

  • Copyright notice on the slides

    by Jan Lehnardt,

  • RDBMS and Service Layer

    by Alex Popescu,

    • Copyright notice on the slides

      by Jan Lehnardt,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Heya,
      nice article & interview! I'd like to point out that the image above and the linked slides are subject to the creative commons license (specifically creativecommons.org/licenses/by-nc-nd/2.0/). See jan.prima.de/~jan/plok/archives/105-Slides-From... for the original publication and extensive commentary.

      And commenting to Anthony's take on CouchDB. Although he doesn't state it explicitly, I'd like to point out (and I think he implies that) CouchDB is not as Ruby-Hacker-friendly as RDDB, but it might have advantages when scaling up and out over multiple machines and locations.

      Cheers,
      Jan
      --

      Back to top
    • Re: Copyright notice on the slides

      by Werner Schuster,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      @Jan:
      Thanks for the link to the slides and blog entry. I removed the image - it's better to see it in the context of your slides anyway.

      Back to top
    • Re: Copyright notice on the slides

      by Jan Lehnardt,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Oh, a copyright note would have been enough, you can sure use the image! :-) Still, thanks for being quick about it.

      Back to top
    • Re: Copyright notice on the slides

      by Anthony Eden,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Jan,

      CouchDB might have indeed have advantages when scaling up and out over multiple machines and locations at this point and time, especially given the support for interprocess communication both locally and over the net that is built into Erlang, however I would argue that the same thing can be accomplished in Ruby through use of libraries now. The approach in RDDB will be to use EC2 and S3 or possibly Rinda or maybe some other sort of messaging system. I also think it will be interesting to see what Matz does in the next couple of years since he brought up his interest in this subject and RubyConf this year.

      -Anthony

      Back to top
    • Re: Copyright notice on the slides

      by Jan Lehnardt,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Heya Anthony,
      thanks for chiming in! There are advantages for both systems; just like CouchDB does not want to replace the RDBMS, as there's room for both. I'll definitely have a closer look at RDDB now ;-)

      Cheers,
      Jan
      --

      Back to top
    • RDBMS and Service Layer

      by Alex Popescu,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      IMO what Document Oriented Distributed Database are evangelizing is just a set of services that can work with the current RDBMS. For example in Java world most of these service have been standardized through a JSR (JSR-170) on top of physical storage.

      See: Data should be available for easy retrieval, integrate simple reporting methods and provide a (fulltext) search.


      I would say that the relational model is very good fit for reporting. And the most of the current RDBMS are already providing support for fulltext indexing.

      Secure: Compartmentalization of data [...]


      I assume this means veritical/horizontal partitioning, which is another feature that current RDBMS are providing or at least starting to consider.

      Concluding, I wouldn't say that RDBMS are having problems in some of these directions. Indeed, as the requirements of the today apps are very high, we are waiting for the RDBMS providers to try to catch up with the latest requirements on their side.

      ./alex
      --
      .w( the_mindstorm )p.
      Alexandru Popescu
      Senior Software Eng.
      InfoQ Techlead/Co-founder

      Back to top
    • Re: RDBMS and Service Layer

      by Jan Lehnardt,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      > I assume this means veritical/horizontal partitioning, which is another feature that current RDBMS are providing or at least starting to consider.

      No, that means that data you put in in the name of user X is not readable by requests from user Y, for example.

      Also, by no means, traditional RDBM systems are not good at what's mentioned in the four pillars. The pillars are, in fact, a guide to help comparing different data storage systems. When it comes to sharing though, most traditional systems fall flat on the face, sorry :-)

      Again, CouchDB is not here to replace anything, it is just another tool worth considering for certain types of problems.

      Cheers,
      Jan
      --

      Back to top
    • Re: RDBMS and Service Layer

      by Alex Popescu,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Thanks for the clarification Jan. So, "data compartmentization" is in fact value-level ACL. That's an interesting concept. Till now I was thinking about having column level ACL and I am wondering if such a fine grained ACL wouldn't result (if extensively used) in weird reporting results. Lets say you are querying for some data (in the name of user X that might have hidden values), how do you make the difference between a NULL value and a forbidden value? (will you introduce a new NULL-like value?) On a different level, this feature kind of ruines the possibilities to cache things. ... Well, I think it is too fine grained for me :-).

      ./alex
      --
      .w( the_mindstorm )p.
      Alexandru Popescu
      Senior Software Eng.
      InfoQ Techlead/Co-founder

      Back to top
    • There's another cool solution to this problem: Xcalia

      by Matthew Adams,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      You should check out XIC from Xcalia (www.xcalia.com). It's another solution to the problem of heterogeneous data access, only it uses transparent persistence APIs (JDO, JPA, SDO) to solve the problem. Of course, that does mean that you're talking about a Java (or Groovy) solution, but that serves a lot of applications out there.

      You can think of XIC as an extension of object-relational mapping to object-service mapping, only with XIC, there is no explicit mapping. Check out the slides at
      www.xcalia.com/technology/circumvent-SOA-design...
      for more info. Very much worth a look-see.

      -matthew

      Back to top
    • Re: RDBMS and Service Layer

      by Jan Lehnardt,

      Your message is awaiting moderation. Thank you for participating in the discussion.

      Thanks for the clarification Jan. So, "data compartmentization" is in fact value-level ACL.


      Well, not quite! :-) It is document-level ACL if you will. A document is roughly a data-record. They will have ownership and permissions and all that.

      Cheers,
      Jan
      --

      Back to top

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

BT