Often it’s said the NoSQL databases will inevitably prevail because they are cheaper. But that is wrong and it misses the much bigger points at issue in the transition from relational databases to NoSQL, some of which are addressed in this webinar.
Money absolutely matters. The NoSQL databases in widespread use are open source and, right away, there are sizable savings involved in not paying the licensing fees exacted by the proprietary SQL database companies. Those fees can be mammoth.
But this is about a lot more than licensing dollars. In fact, it’s about a lot more than money. When NoSQL is implemented, what happens with developers is dazzling. When they work with relational databases, developers - by necessity - spend lengthy hours contemplating schema and they also have to worry, a lot, about not breaking the app and then of course there is the constant focus on trying to keep data sets down to sizes the relational database and the associated and expensive hardware can digest.
This is hard work. It is kind of like being told to find ways to fit an NFL lineman into a 42 regular suit without ripping the seams. [[Matt to replace with soccer or tennis reference :)]]
Switch into a NoSQL world and, suddenly, the size of the data sets doesn’t matter - but what is particularly beautiful is how it all runs on just about any hardware that happens to be around.
Need more capacity? There is no need to put in an order for namebrand gear and wait weeks, maybe months, for its delivery. In a NoSQL, just spin up some more servers in your private or public cloud, or build your own servers using commodity hardware. It probably is good enough and, yes, there are documented cases where NoSQL users were able to switch out of dependence on multi-million dollar servers and into commodity servers costing a few thousand bucks apiece, if that.
Performance increased, even though data grew, and that is because of the nimble and swift analysis delivered by NoSQL tools.
Craigslist offers another example that shows why TCO, alone, is not the right analysis. When it wanted to add a simple row to its apartment listings - are pets allowed: yes or no? -- it took a staggering three months to insert that without breaking the existing schema. Three months, in the 21st century. With a NoSQL database that could be done in a few hours.
In NoSQL, developers focus on agile development where schema are there to serve, not to rule. Data, too, can be digested in polymorphic formats, there is no need in NoSQL to shoehorn data into formats that aren’t natural.
There is vast literature documenting the extraordinary TCO advantage owned by NoSQL databases but the main point is simpler still. Relational databases are 20th century tools that are exceptional in handling 20th century data sets and, yes, a lot of those still exist.
Scenarios are plentiful where NoSQL costa a fraction -- perhaps one-third or less -- of a relational database architected to handle identical data sets. But sometimes left out is that, in most cases, the analytics are more robust and vastly more flexible in the NoSQL for instances.
The reason: NoSQL is built for 21st century data sets where the ability to scale is a must-have, frequently to sizes that would have been unimaginable just a few years ago. When yet another petabyte of data is hurled into the mix the only proper response today is, "thank you, sir, may we have another?"
Because there will be another...and another. That is the way it goes in the 21st century and that is why this is the age of NoSQL.