MongoDBbeginner

MongoDB Overview

What MongoDB is, how replica sets and the document model shape design, and the workloads it fits.

2 min readBeginnerUpdated 31 Jul 2026Edit this page

MongoDB is a document database. Data is stored as BSON documents in collections, indexed by field paths, and distributed through replica sets and optional sharding.

Architecture

WiredTiger storage engine. Document-level concurrency, MVCC, snapshot isolation and compression by default.
Replica set. One primary accepts writes; secondaries replicate the oplog and can serve reads. Members hold elections when the primary becomes unreachable. See Replica Sets.
Oplog. A capped collection of idempotent operations, used for replication, change streams and point-in-time recovery.
Sharding. Optional horizontal partitioning by a shard key, coordinated by config servers and routed through mongos. See Sharding.

MongoDB three-member replica set

Best use cases

Aggregate-oriented data where a document is naturally read and written as a unit — orders with line items, product catalogues, user profiles with nested preferences.
Schemas that vary between records or evolve frequently, where an ALTER TABLE per change would be painful.
Workloads that scale by sharding on a key that appears in most queries.
Applications that benefit from documents matching their in-memory object model.

When not to use it

When the data is highly relational and queries join many entities. Joins exist ($lookup) and are expensive; a relational engine does this better.
When you need cross-document invariants enforced by the database. Multi-document transactions exist, but they are more costly than a single-document update and should be the exception.
For analytical scans over huge collections — a column store is the right tool.

Data model

{
  _id: ObjectId("66aa1f2b9c4e2f0012ab34cd"),
  customerId: ObjectId("66aa1e119c4e2f0012ab1122"),
  status: "paid",
  totalCents: 19900,
  items: [
    { sku: "TS-001", qty: 2, priceCents: 4950 },
    { sku: "MG-014", qty: 1, priceCents: 10000 }
  ],
  createdAt: ISODate("2026-07-31T10:22:00Z")
}

Documents have a 16 MB limit, which is a design constraint rather than a technicality: it means unbounded arrays inside a document are a modelling error. See Document Modeling.

Schemaless does not mean structure-free

Use schema validation to enforce the shape you actually rely on:

db.createCollection("orders", {
  validator: { $jsonSchema: {
    bsonType: "object",
    required: ["customerId", "status", "totalCents", "createdAt"],
    properties: {
      status: { enum: ["pending", "paid", "shipped", "cancelled"] },
      totalCents: { bsonType: "long", minimum: 0 }
    }
  }},
  validationLevel: "strict"
});

Consistency and transactions

Single-document writes are atomic. Multi-document transactions are available on replica sets and sharded clusters, with a default 60-second limit and higher cost — model to avoid needing them.

Consistency is tunable per operation through read and write concerns, which is where most of the durability decisions actually live.