MongoDB Released as Open Source — The Document-Oriented Symbol of the NoSQL Movement

On 11 February 2009, the New York startup 10gen made the first public release of MongoDB. Instead of rows and columns, it stored JSON-style documents—a document-oriented database. Then unknown, this design would within a few years become the public face of the whole "NoSQL movement". Version 1.0 shipped that August; what appeared on this day was the first pre-1.0 public build.

10gen's Origins — Escaping DoubleClick

10gen was founded by three DoubleClick alumni: Dwight Merriman (co-founder and CTO of the ad-tech company), Kevin P. Ryan (its CEO), and Eliot Horowitz (an engineer there, later co-founder and CTO of ShopWiki). At DoubleClick they had wrestled with web-scale real-time data processing—aggregating bid and serving logs—and, by the company's own account, felt the scaling ceiling of conventional relational databases firsthand.

They founded 10gen in 2007. The original goal was a PaaS (Platform-as-a-Service) called Babble, closer to Google App Engine than to a general cloud, and assembled entirely from open-source components. The market for Babble never appeared; the database layer built underneath it did better than the platform above. In 2009 they dropped the PaaS plan, refocused as a database company, and released MongoDB.

The company name comes from "Tenth Generation"; "MongoDB" from "humongous". Together: a tenth-generation DB for huge data.

The Document Model — Nested Documents, not Tables

MongoDB's defining feature is that the unit of data is not a row but a BSON document. BSON is a binary form of JSON, and a single document can contain nested objects, arrays, and typed values (date, ObjectID, and so on).

{
  _id: ObjectId("..."),
  name: "Alice",
  orders: [
    { item: "Book", price: 1500, date: ISODate("2024-01-15") },
    { item: "Pen",  price: 200,  date: ISODate("2024-02-03") }
  ],
  address: { city: "Tokyo", zip: "100-0001" }
}

What an RDBMS would model as four tables joined together (user, order, order_item, address) fits into one document. The design choice is to align the application's object structure with the database's storage structure, attacking the object-relational impedance mismatch that makes ORM frameworks so awkward.

MongoDB is also schemaless: no need to declare a table layout in advance, and documents of varying shape can coexist in the same collection. For agile teams whose schemas shift week to week, this is a strong feature.

The NoSQL Movement — A Convergence in 2009

2009 was also the year "NoSQL" entered industry vocabulary. In June, Johan Oskarsson, then a developer at Last.fm, organised a San Francisco meetup to discuss "open-source distributed, non-relational databases" and reached for NoSQL as its name. MongoDB, Cassandra (open-sourced from Facebook in 2008), CouchDB (2005-), Redis (2009-), and other newcomers were in the room. The word itself was not new: Carlo Strozzi had used it in 1998 for something else entirely—a relational database that simply did not expose a SQL interface.

The context was a sense that RDBMS were running out of headroom for web-scale services. Google's Bigtable paper (2006) showed that a non-relational distributed store could carry production load, and Amazon's Dynamo paper (2007) argued publicly for eventual consistency in place of strong consistency in distributed environments. The view that "ACID and the relational model are not absolutes" became NoSQL's theoretical backbone.

Inside that wave, MongoDB took the role of the document-oriented flagbearer. Alongside Cassandra (column-family), Redis (key-value), and Neo4j (graph), it became one of the canonical "not-an-RDBMS" choices.

Rapid Adoption and Sharp Criticism — The ACID Argument

In the first half of the 2010s, MongoDB spread fast. Foursquare, Craigslist, The New York Times, eBay, Shutterstock—adopters ranged from startups to enterprises.

But the critiques were equally sharp. Early MongoDB defaulted to fire-and-forget writes: db.users.insert() could return success while the data evaporated. In November 2013 Sarah Mei published "Why You Should Never Use MongoDB", drawing on the distributed social network Diaspora: because the same user turns up as friend, commenter, and liker, denormalising an activity stream into a single document forces you to duplicate or to cross-reference, and either way the document model stops paying for itself. Her conclusion was that wherever an implicit schema exists—wherever you expect particular things inside the JSON—MongoDB is the wrong choice. The post was read everywhere.

10gen (renamed MongoDB Inc. in August 2013) answered the criticism with persistent engineering. MongoDB 3.0 in 2015 made the WiredTiger storage engine (B-tree with compression, and an optimistic protocol that gives document-level locking) the default. Version 4.0 in June 2018 added multi-document ACID transactions within a replica set, extended across sharded clusters in 4.2 (2019). The flagbearer of an anti-ACID movement ended up adopting ACID—a fitting microcosm of NoSQL's evolution overall.

Changing the Licence — AGPL to SSPL

The other reversal was the licence. MongoDB Community Server had long shipped under the GNU AGPLv3. On 16 October 2018 MongoDB Inc. announced a move to the Server Side Public License (SSPL), a licence it drafted itself; every release issued on or after that date, including patch releases of older versions, fell under it. The target was the pattern in which a cloud vendor that did not build the software offers it as a service and captures the value. Eliot Horowitz, then CTO, put it this way in the press release:

"Unfortunately, once an open source project becomes interesting, it is too easy for cloud vendors who have not developed the software to capture all of the value while contributing little back to the community."

The SSPL has never been approved by the Open Source Initiative, and Debian, Red Hat Enterprise Linux, and Fedora dropped MongoDB from their distributions. It opened the sequence of open-source companies tightening licences against cloud providers that Redis, Elastic, and HashiCorp would later join.

2017 IPO and Today

MongoDB Inc. began trading on NASDAQ as MDB on 19 October 2017. Priced at US$24, the stock opened at US$33 and closed day one at US$32.07—about US$1.2 billion of market capitalisation at the offering price, roughly US$1.6 billion at the close—raising US$192 million. For the fiscal year ended January 2026 the company reported US$2.46 billion of revenue, up 23 per cent year over year, with more than 65,200 customers.

MongoDB Atlas (2016-), the multi-cloud managed service running on AWS, Azure, and GCP, accounted for US$1.81 billion of that—comfortably over 70 per cent of revenue. Open-source upstream plus paid managed downstream: this is now one of the standard revenue models for open-source companies, echoed by Snowflake, Confluent, and Databricks.

A Second Axiom: Data Is a Document

The release of MongoDB was more than a product launch. It was the moment when the idea "let the database schema match the shape of the application's data" moved out of research papers and Google's internal systems and into an OSS download anyone could install in five minutes.

The 40-year axiom "data is a table", set by the relational model, was joined by another axiom: "data is a document". 11 February 2009 was one of the days that made that pairing real. Firestore, DynamoDB Document Mode, the MongoDB API in Cosmos DB, PostgreSQL JSONB—document-oriented thinking has settled into the database world as a permanent option.