question

Timothy - good question! +1 I am sure that NoSQL has it's place for certain applications like Twitter and the like where data integrity is of much less importance than what a SQL Server DBA may see as a "real" database. As I understand it NoSQL implementations drop performance killing features like referential integrity in order to make themselves more flexible and scalable - again, something that detracts from it being a "real" database and more a temporary/dispersed data store. For those reasons, and many more, I don't think that NoSQL will move into a position where it could displace SQL Server. That is if they stay on their current path of development. I still believe that learning a NoSQL variant can be valuable. You would have an insight into that type of system and it gives you another weapon in your arsenal. On top of that, it is just an interesting technology, so why not! As to which one is the right one to start with, no idea. Tell me once you have found out! I spend so much time trying to learn SQL Server already, I couldn't find the time for something that (currently) doesn't have any bearing at work.

yeah, great question Timothy. my, largely uneducated, thoughts are: 1 - yes 2 - yes 3 - no idea To expand on that, I would say that as a DBA it is my duty to keep on top of data technologies that might be of use to my employer and to that effect I need to understand NoSQL tech. Not to run it but to know what it can/cant do and whether it can provide a solution to any business need we have. It wont ever replace our accounts or payoll system as we want to make sure that I dont get the CEO's salary rise and so on due to failed/inconsistent transactions but it might do for a website feature or something like that. In essence I can see it beside SQL and/or Oracle in some places but not moving in and taking over. At least not in the short - med term. Other than understanding what they do and roughly how it happens I know none of the products so couldnt advise on one to pick up to learn in depth.

If you want to position yourself for better and challenging career 10 years from now, go ahead and start learning it. Even though its a big gamble I can see a scenario where companies deal with enormous volume of data and want to run horizontally scalable databases on commodity hardwares. So keep an eye on it and you might help in the future heavily data driven startups who might be the next facebook, twitter, google, amazon and the likes. As for me I will stick with RDBMS for now:-)

I think it's useful to distinguish between NoSQL as an idea versus the collection of products that currently go under that name. In principle replacing SQL with something better ought to have a lot of benefits. It's bound to happen sooner or later and probably sooner than many people would expect. However, none of the current crop of software commonly called NoSQL looks like a serious SQL-beater (unfortunately). Mainly because they mostly eschew the relational model in favour of less useful alternatives.

According to this tiny news article, Cassandra may end up being capable of [sitting on top of SQL Server][1] & Oracle: > Cassandra itself is not a database. But it does allow someone to use a variety of open source databases and deliver a consistent, high volume write function to a Web page. The project is not ruling out working with a variety of proprietary databases, including Oracle. If that's the case, then Cassandra is not a direct competetor with SQL Server. But I'd be willing to bet that SQL DBA's will be asked to work with it! [1]: http://www.zdnet.com/blog/open-source/cassandra-becoming-the-nosql-alternative/5971

My thoughts are: 1. NoSQL databases wouldn't replace SQL Server any more than aeroplanes will replace trains. They are different things, and have different strengths and weaknesses. 2. I wouldn't say there was huge benefit in learning a NoSQL variant for a SQL DBA - I mean, to be honest, there's so much to learn about SQL Server and relational theory anyway... But it can't hurt, if you're interested. 3. Like others - no idea!

1) Am I correct in understanding that NoSQL is not going to displace products like SQL Server? Initially yes, however this is primarily due to the momentum of existing companies that have invested staff, equipment, and software licensing costs that they need to recoup. Startups will be free (from legacy apps and SQL-specialized engineers) to try the technology, and because the cost of setting up a small cluster is relatively small, it may be a good option for them. Column-oriented data stores don't require much (if any) in the way of ORM either. Persons can concentrate on doing things with their objects as opposed to developing relational schema, or writing SQL to split and reassemble their objects to/from tables that match that schema. 2) Is there substantial value in learning a NoSQL variant for a SQL Server DBA? Yes. You should at the very least understand the kinds of problems that can be solved more easily with distributed key-stores or graphs as opposed to relations (and vice-versa). Learning a new approach to working with data will expand your options when facing data related concerns. 3) If there is value in a SQL Server DBA learning NoSQL, which one forms a logical start? Cassandra? Or something more exotic like Map Reduce? Start with the fundamentals, Read the BigTable and Dynamo papers (just google them). Learn about the map-reduce or map-fold paradigm. Try a few examples with Hadoop MapReduce or Erlang/OTP. Then try using a bit of Hadoop Hive (it has a syntax very similar to SQL). That should bridge the gap.