I'd wanted to lift the burden of writing ORM layers and dealing with relational algebra and focus on persisting my intrinsic application models. I'd studied XML and other semi-structured web data at the graduate level, but was frustrated there wasn't a technology that was meeting that need. Companies like Microsoft and Oracle were adding XML functionality to their databases, but because it was bolted on an already constrained structure, it failed to address the mangling that relational structures impose.
I saw a demo of MongoDB in 2010 and was thoroughly impressed. The developer experience out of the box was exquisite. During a 1 hour presentation, I was able to download the binary, connect to the shell, create records in a familiar form (JSON), query the data from my preferred language, and all within a often second-class platform, Windows.
There was a certain _joy_ to using MongoDB, but I had my doubts, too. If this form of database interaction was so easy and fun to use, what was the catch?
Well, the catch was that MongoDB was re-thinking the issue from the ground up, re-envisioning the way we work with data, in an attempt to transcend the inertia that held back existing offerings.
When I started working with MongoDB, it felt like the same transcendence I experienced moving from C++ to Python in 1998--the approach was simpler, more intuitive, and powerful.
Sure, there were some growing pains. The concept came first without acknowledged writes or even journaling. Replication came in stages. Sharding came later.
But our operation was able to go from startup-scale to a global enterprise-scale operation backed by MongoDB, performing admirably, and giving developers the ability to rapidly innovate in a low-impedance environment. We've since developed dozens of applications against MongoDB.
I've considered emergent competing offerings like PostgreSQL with its JSON store, but when I did a few years ago, they were still dramatically behind in replication and sharding support, requiring extensive administration for aspects that MongoDB handled out of the box.
And in the meantime, MongoDB has continued to innovate, bringing in pluggable storage backends, advancing the query engine and aggregation framework, adding native geo-location support, and formalizing the concept of shard zones (a key feature for global scalability). And with Atlas, they've made server administration a near zero-cost proposition.
I'm a fan because when it comes to boots-on-the-ground experience, I see nothing else in the industry that even comes close. The most pressing deficiencies have been addressed and without compromising the key innovations. As a developer and technology leader, I choose MongoDB as my default database and would only consider a traditional database if the use case was intrinsically highly constrained and tabular in nature.