단일 필드 및 복합 인덱스 만들기

A single-field index is the simplest index type: it tracks the values of exactly one document field in a sorted B-tree. You create one with db.collection.createIndex({ field: 1 }), where 1 means ascending order and -1 means descending. Single-field indexes are ideal for queries that filter or sort on just one field.

MongoDB auto-generates an index name like email_1 or age_-1 from the field name and direction. You can override this with the name option to give meaningful labels to your indexes—especially helpful when managing many indexes in production or when the auto-generated name would exceed the 127-character limit imposed on compound indexes with many fields.

A compound index tracks multiple fields together in a single B-tree. The entries are sorted first by the first field, then by the second field within each group of the first, and so on. This makes compound indexes far more selective and versatile than multiple single-field indexes for queries that filter or sort on several fields at once.

A compound index on { a, b, c } can serve queries on { a }, { a, b }, and { a, b, c }—these are called index prefixes. It cannot serve a query on just { b } or { c } alone because the B-tree is ordered by the first field first. Understanding the prefix rule helps you avoid creating redundant single-field indexes when a compound index already covers them.

You create a compound index by passing an object with multiple fields to createIndex. Field order matters: put equality fields first (fields filtered with $eq or exact values), then range fields, then sort fields. This arrangement ensures the index is used for both filtering and sorting in one traversal, following the ESR (Equality, Sort, Range) principle.

Always verify that your new index is actually being used with .explain('executionStats'). Look for winningPlan.stage: 'IXSCAN' to confirm index use, and check totalDocsExamined vs nReturned—a well-indexed query should examine roughly the same number of documents it returns.

In MongoDB 4.2+, all index builds are non-blocking by default: they hold an exclusive lock only briefly at the start and end of the build, allowing reads and writes to continue during the lengthy build phase. On older versions you had to specify { background: true } explicitly. Building a large index can still consume significant CPU and I/O resources, so schedule builds during low-traffic periods in production.

You can remove an index with db.collection.dropIndex(), passing either the index name or the key specification. Dropping unused indexes reduces write overhead and frees memory. The _id index cannot be dropped. Use dropIndexes() (plural) to drop all non-_id indexes at once—useful when rebuilding an index strategy from scratch.

$indexStats is an aggregation stage that shows how many times each index has been used since the mongod process last started. An index with zero accesses is a prime candidate for removal. Combine this data with the index size from db.collection.stats() to build a complete cost/benefit picture for each index in your collection.

Multiple single-field indexes can sometimes be combined via index intersection, but MongoDB's planner prefers compound indexes when they exist. A single well-designed compound index is almost always faster and more predictable than relying on the planner to intersect two separate indexes. Create compound indexes for your most frequent multi-field queries rather than hoping intersection will save you.

MongoDB allows a maximum of 64 indexes per collection. This is more than enough for well-designed schemas, but if you approach this limit, it is a strong signal that something is wrong—perhaps you have many redundant indexes or a table-like design that doesn't fit the document model. Audit your index usage regularly and prune indexes that have zero $indexStats accesses.

In this lesson you learned: single-field indexes track one field in a sorted B-tree and support equality, range, and sort queries, compound indexes track multiple fields and must be queried using their prefixes, and ESR ordering (Equality, Sort, Range) maximises compound index effectiveness. Next up we explore special index properties like unique, sparse, partial, and TTL.