Why is the non-#RDBMS becoming so popular?


Where'd non-RDBMS come from and where's it's headed in relation to RDBMS systems. Just about a 4 min view.


Dave Littman:                   Hi, Dave Littman, Truth in IT, joined today by Mike Matchett. Mike is contributing subject matter expert and thought leader here at Truth in IT. Mike, welcome.

Mike Matchett:                  Thanks, Dave.

Dave Littman:                     So hey, today we're talking about non-relational database management systems, and why they are becoming so popular. What do you think, Mike?

Mike Matchett:                  All right, so, what we used to call these things was NoSQL, as in no SQL databases, because the first difference were what came out that you had your relational database that took your highly structured data, had to be forced into a schema, and then that database management system RDBMS, could only handle so much data. And people said, well 10 years ago, "We've got lots of big data and that big data doesn't quite fit into that little relational database system, and it's not structured enough to scrunch in there. So, we're going to make these different databases, and we're going to call them NoSQL, because they're not going to support that structured query language. They're going to have some other attribute, though, that makes them great for handling scale. So, you have graph databases, you have key value stores for just lookup and big things, you have MongoDB, which is more of a document database for JSON documents. You have a couple of really kind of oddball databases, they do different little things. In-memory databases, and so on.

                                                      Then people came back, and of course said, "Hey, we got to add SQL back to these things, 'cause otherwise how would people use them." So, the NoSQL moniker has been losing its popularity. Now we're just calling them either operational databases or non-relational databases. But the truth is, they're just something new and different and usually the main difference today is, you've given up something of the consistency and transactional nature in exchange for speed or scale or something else over here. So, there are many reasons why you want to make that tradeoff and usually they have to do with growing a big web app or something.

Dave Littman:                     Okay, so do you think that relational databases are still useful and in some cases even preferable?

Mike Matchett:                  Oh, absolutely. You're never going to get away from a business application that needs to do a transaction, needs what they call that acid compliance, where you have an atomic transaction that is committed, you know it happened. And if it didn't happen, everything rolls back. So, everything is in a consistent state all the time. You're always going to need that relational database, that transactional database to happen.

                                                      But I think at the same time, people have realized that there's lots more data out there that can be put into some structured way to mine or to navigate. One of my favorite examples is graph databases, where you don't store data as records in a table, you store them as a little web of nodes and links between them, such that when you go to query that, you actually say, "Can you tell me the friend of this node who's got six friends, whose friend is over here, and his friend is over there?" And you can make very easy queries out of these kinds of web questions that you can't do in a relational database without bringing it to its knees. So, they both are going to have their purposes going forward.

Dave Littman:                     Any other thoughts about the trends moving forward towards more non-RDBMS or less?

Mike Matchett:                  So, I think what you're going to see is a little bit of a convergence happening at first, where some non-RDBMS qualities come into the structured database. Then you're going to see, on the other hand, a lot of what were NoSQL databases also now having a SQL layer being laid on top of them so they can support full transactional semantics at scale, too. So, I don't think it becomes one or the other, I think they sort of start to converge a little bit closer to each other and overlap in a lot of important ways.

Dave Littman:                     All right, fabulous. Well hey Mike, thanks again for taking the time to speak with us today.

Mike Matchett:                  Oh, thanks a lot, Dave.

Dave Littman:                     And thanks for watching.