monarch-bench

Benchmarks for comparing MongoDB operations across:

• Monarch ORM
• Mongoose
• Native MongoDB driver

The suite uses tinybench and runs against a real MongoDB instance or falls back to an in-memory server via mongodb-memory-server when no MONGO_URI is set.

Requirements

• Node.js 22+
• pnpm

Install

pnpm install

Run Benchmarks

pnpm bench

Set MONGO_URI in .env to run against a real MongoDB instance. Without it, an in-memory server is started automatically.

This runs:

• Initialization benchmarks (initialization, connection)
• Method benchmarks for schema sizes: small, medium, large
• Virtuals benchmarks (computed fields via findOne / findMany)
• Relations benchmarks (findOne, findMany) with and without indexes, across populate styles (populateOne, populateMany, populateBoth) and multi-field variants

Project Structure

• src/index.ts: benchmark entrypoint and suite orchestration
• src/setup/: benchmark harness, runner, and schema-sized data builders
• src/db/: driver/ORM connection and model/schema setup (monarch.ts, mongoose.ts, native.ts)
• src/bench/init.ts: initialization benchmarks
• src/bench/methods/: per-method benchmarks (find, updateOne, aggregate, etc.)
• src/bench/virtuals.ts: virtuals (computed field) benchmarks
• src/bench/relations/: relation populate benchmarks (findOnePopulate*, findManyPopulate*)

Scripts

• pnpm bench: run benchmark suite
• pnpm check: TypeScript check + formatting check
• pnpm format: check formatting with Prettier
• pnpm format:fix: apply Prettier formatting

Results

All figures are median throughput (ops/s) — higher is better.

Environment: M2 MacBook Air, MongoDB v7.0.8 (local, Homebrew)

---

Initialization

Benchmark	Monarch	Mongoose
initialization	1,362	737
connection	187	146

Monarch bootstraps ~85% faster than Mongoose. Connection setup is ~28% faster. This is a fixed cost per process, so the difference matters most in short-lived or serverless environments.

---

Methods — Small Schema

Method	Monarch	Mongoose	Native
distinct	12,645	12,164	13,058
find	10,073	7,913	10,603
findById	12,251	11,651	12,533
findByIdAndUpdate	12,397	9,864	12,931
findByIdAndDelete	11,911	8,882	12,201
findOne	12,390	10,174	12,632
findOneAndReplace	10,777	6,021	11,337
findOneAndUpdate	12,579	9,958	13,001
findOneAndDelete	12,429	9,717	17,937
insertOne	16,997	5,587	13,238
insertMany	1,647	695	6,032
bulkWrite	11,199	10,979	11,385
replaceOne	11,511	6,936	11,887
updateOne	12,685	11,696	12,938
updateMany	7,542	7,391	7,884
deleteOne	12,987	11,713	13,086
deleteMany	12,973	11,940	13,108
aggregate	4,348	4,578	4,747
countDocuments	9,934	9,438	10,101
estimatedDocumentCount	16,021	14,414	16,032

Monarch leads on almost every operation and closely tracks native. The biggest gaps are on writes: insertOne is ~3x faster than Mongoose, insertMany ~2.4x. aggregate, bulkWrite, distinct, and updateMany are roughly equal across all three.

---

Methods — Medium Schema

Method	Monarch	Mongoose	Native
distinct	12,516	12,134	13,230
find	9,581	7,288	9,893
findById	11,934	8,972	12,390
findByIdAndUpdate	12,220	8,807	12,672
findByIdAndDelete	11,511	7,637	11,881
findOne	11,550	8,558	12,320
findOneAndReplace	9,441	4,802	10,017
findOneAndUpdate	12,257	8,801	12,632
findOneAndDelete	11,656	7,824	12,078
insertOne	11,550	4,259	12,552
insertMany	1,099	358	1,504
bulkWrite	11,096	10,934	11,147
replaceOne	10,213	5,600	11,178
updateOne	12,605	11,231	12,889
updateMany	7,762	7,315	7,871
deleteOne	12,931	11,045	13,216
deleteMany	12,841	11,057	12,980
aggregate	877	796	723
countDocuments	10,174	9,494	9,654
estimatedDocumentCount	15,989	14,109	16,260

The write advantage grows with schema size. insertOne is now ~2.7x faster than Mongoose, insertMany ~3x. Read operations maintain a 30–50% lead.

---

Methods — Large Schema

Method	Monarch	Mongoose	Native
distinct	12,882	11,639	13,333
find	9,167	5,916	9,360
findById	10,830	7,069	11,401
findByIdAndUpdate	11,326	6,369	11,184
findByIdAndDelete	10,811	5,636	11,050
findOne	11,289	7,143	11,348
findOneAndReplace	7,733	2,424	8,259
findOneAndUpdate	11,204	6,014	11,116
findOneAndDelete	10,148	5,643	14,679
insertOne	10,714	2,527	12,952
insertMany	523	222	1,057
bulkWrite	14,193	14,406	14,590
replaceOne	11,486	3,304	12,546
updateOne	16,925	10,652	12,821
updateMany	7,777	6,743	7,466
deleteOne	12,384	8,097	13,205
deleteMany	12,948	9,379	12,308
aggregate	710	711	660
countDocuments	9,852	9,136	9,543
estimatedDocumentCount	15,784	13,552	15,748

At large schema the write gap is most pronounced: insertOne is ~4x faster than Mongoose, insertMany ~2.4x, findOneAndReplace ~3.2x. aggregate and bulkWrite are tied across all three drivers.

---

Virtuals (computed fields, 100 comments/post)

Benchmark	Monarch	Mongoose
findOne	12,546	11,310
findMany (n=10)	9,007	4,991
findMany (n=100)	881	481
findMany (n=1000)	168	69

Monarch's virtual computation scales better per document. The gap grows with document count: +11% at findOne, +80% at n=10, +83% at n=100, and +143% at n=1000.

---

Relations — findOne (no index, 100 comments/post)

Populate style	Monarch	Mongoose
populateOne	8,236	4,116
populateMany	228	168
populateBoth	247	187
populateOneMulti (3 refs)	4,707	3,147
populateManyMulti (3 FKs)	76	147
populateBothMulti (6 fields)	80	147

populateOne is ~100% faster in Monarch. populateMany and populateBoth (single FK) are ~35% faster. The multi-field many/both cases reverse: Monarch is ~2x slower when resolving 3 foreign-key fields without indexes, as each field requires a full collection scan.

---

Relations — findOne (indexed, 100 comments/post)

Populate style	Monarch	Mongoose
populateOne	7,308	4,421
populateMany	3,958	1,477
populateBoth	2,755	1,368
populateOneMulti (3 refs)	4,705	2,885
populateManyMulti (3 FKs)	847	486
populateBothMulti (6 fields)	866	458

Indexes unlock Monarch's populate advantage fully. populateOne is ~65% faster. populateMany and populateBoth are ~2–2.7x faster. Multi-field many/both are ~1.7–1.9x faster.

---

Relations — findMany (no index, 100 comments/post)

Populate style	n	Monarch	Mongoose
populateOne	10	850	1,419
populateOne	100	105	180
populateMany	10	94	81
populateMany	100	~1	~8
populateBoth	10	88	83
populateBoth	100	~1	~9
populateOneMulti	10	227	463
populateOneMulti	100	63	105
populateManyMulti	10	31	28
populateManyMulti	100	~0	~2
populateBothMulti	10	27	27
populateBothMulti	100	~0	~2

Without indexes, Mongoose wins all findMany scenarios. At n=100 with populateMany or populateBoth, Monarch reaches ~780ms per call vs ~117ms in Mongoose. The multi-field variants at n=100 are worse still (~2.75s vs ~402ms). populateOneMulti also shows Mongoose ~2x faster at both scales.

Warning: populateMany and populateBoth at n=100 without indexes reaches ~780ms–2.75s latency per operation in Monarch (vs ~117–405ms in Mongoose). Always use indexed relations when calling findMany with foreign-key populate.

---

Relations — findMany (indexed, 100 comments/post)

Populate style	n	Monarch	Mongoose
populateOne	10	349	641
populateOne	100	83	179
populateMany	10	168	61
populateMany	100	33	8
populateBoth	10	140	64
populateBoth	100	29	7
populateOneMulti	10	374	534
populateOneMulti	100	62	104
populateManyMulti	10	81	26
populateManyMulti	100	8	2
populateBothMulti	10	75	26
populateBothMulti	100	7	2

With indexes the winner depends on populate direction. For populateOne and populateOneMulti (parent holds the FK), Mongoose is ~2x faster. For populateMany, populateBoth, and their multi variants (child holds the FK), Monarch is ~2–4x faster at n=100.

Note: populateOne and populateOneMulti are actually slower in the indexed collections than in the unindexed ones (for both ORMs). This is because these populate styles resolve posts.pinnedCommentId → comments._id, and _id is always indexed by MongoDB — the extra indexes on the indexed collections provide no benefit for this lookup direction, but do add overhead to the collection.