# Pipeline Optimization > Depth-first rabbithole dossier for Pipeline Optimization; source-anchored research pack. Parent: [MongoDB Aggregation Pipeline](https://llms-explorer.com/tree/mongodb-aggregation-pipeline/) · 4 facets · 12 facts · page: https://llms-explorer.com/tree/pipeline-optimization/ ## How it works - [low] If a pipeline has a $group stage or a $sort followed by $group/$sort that causes the pipeline to divide, avoid specifying allowDiskUse:true if possible to prevent increased network transfer of intermediate records. Sources: https://www.practical-mongodb-aggregations.com/guides/sharding.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [high] To optimize MongoDB aggregation pipelines, filter documents with a $match stage as early as possible so that later stages process fewer documents. Sources: https://www.mongodb.com/docs/manual/core/aggregation-pipeline-limits.md, https://studio3t.com/knowledge-base/articles/mongodb-aggregation-framework/ — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#patterns` - [high] Use an index to satisfy the sort operation so that MongoDB does not sort documents in memory, and move the $sort stage to near the start of the pipeline if it does not depend on preceding $unwind, $group, or $project stages. Sources: https://www.mongodb.com/docs/manual/core/aggregation-pipeline-limits.md, https://www.practical-mongodb-aggregations.com/guides/performance.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#patterns` ## Problems, failure modes and limitations - [low] At runtime, the aggregation engine collapses a $sort and $limit into a single special internal sort stage and attempts to move $match stages to the top of the pipeline, using top-level $match content as the initial query filter. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [low] The aggregation pipeline's 100 MB RAM limit for blocking stages applies to $group and $sort stages; if exceeded, the database throws an error unless allowDiskUse is true, which allows spilling to disk but significantly increases latency. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [low] The $sort + $limit coalescence optimization still applies when allowDiskUse is true and the n items exceed the aggregation memory limit. Sources: https://www.mongodb.com/docs/manual/core/aggregation-pipeline-optimization.md — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` ## Comparisons and alternatives - [low] The MongoDB aggregation pipeline automatically reshapes queries to improve performance, with no limit to the number of stages or how they are combined, unlike the more complex map-reduce framework. Sources: https://studio3t.com/knowledge-base/articles/mongodb-aggregation-framework/ — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [medium] Using array operators instead of $unwind and $group stages avoids introducing unnecessary blocking stages, which is significantly more efficient for large data sets and prevents the runtime from splitting the pipeline. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html, https://www.practical-mongodb-aggregations.com/guides/sharding.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#patterns` ## Facts and statements - [low] Filtering documents before a $lookup stage reduces the work done by later stages, and the optimizer can move independent filters to the beginning of the pipeline to allow index usage. Sources: https://www.mongodb.com/docs/manual/core/aggregation-pipeline-optimization.md — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [low] The $sort stage may not be the first stage in the pipeline if a $match stage also takes advantage of the same index. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#core` - [high] To take advantage of indexes, use $match or $sort stages in the first stage of the pipeline, and always inspect the explain plan to ensure the intended behavior is induced. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html, https://studio3t.com/knowledge-base/articles/mongodb-aggregation-framework/ — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#methodology` - [low] If a $sort stage must remain later in the pipeline, move it as late as possible to ensure earlier stages significantly reduce the number of records streaming into the blocking sort stage. Sources: https://www.practical-mongodb-aggregations.com/guides/performance.html — source: `~/.global-ai-hub/research/.pipeline/mongodb-2026-10-14/frb/pipeline-optimization-c282113b1a/rabbithole-synthesis.md#patterns` ## Related concepts - Pipeline — is a part of Pipeline Optimization - Optimization — is a part of Pipeline Optimization