Pricing examples

This page includes examples of how billing units are calculated in some of the most common scenarios. Note that each query might differ in data processed based on factors such as the query plan, the shape of the data, and the indexes available.

We recommend using the Query Explain feature to better understand the cost and performance of your specific queries.

Read operations

Most read operations entail either performing a point read of a specific document or scanning a range of data based on an identifier. Read operations consume read units. See the following examples.

Full-text Search

Charges for full-text search queries are distinguished between the cost to perform a search on a text index and the cost to retrieve the documents. The cost to perform a search on a text index is determined by the complexity of the query and the amount of data scanned in the index to find the results. Factors influencing this cost include the number and nature of search terms (e.g. terms, phrases), whether relevance scoring is applied, and the number of index entries that need to be examined. The cost for retrieving relevant documents is based on the number and size of documents fetched after the search phase is complete.

The following table delineates the charges associated with executing full-text search queries for an ecommerce product catalog for a plant nursery. This example assumes a collection of 200,000 product listings in a collection, with each product listing document being 2KiB, and a corresponding text search index size of approximately 100MiB.

db.products.find(
  {$text: {$search: '"dry water" stack flat'}}
);

db.products.find(
  {$text: {$search: '"indoor" "citrus" "tree"'}}
);

db.products.find(
  {$text: {$search: 'balcony planter'}}
).sort({score: {$meta: "textScore"}}).limit(50);

These examples illustrate how different query patterns impact costs:

• The "dry water" query involves a phrase ("dry water") and additional terms. Phrase searches are more complex as the query engine needs to check token proximity and order, contributing to the 7 read units for the search phase. Since very few documents match, document retrieval cost is low.
• The "indoor" "citrus" "tree" query searches for three separate terms. While slightly less complex per term than a phrase search, it still requires looking up multiple tokens. It has a slightly lower search index scan cost of 6 read units, than the "dry water" example, but 36 documents are returned, increasing the total cost.
• The balcony planter query has the highest full-text search cost of 25 units by a significant margin. This is primarily because it requests results to be sorted by relevance (.sort({score: {$meta: "textScore"}}) ) on terms that match many documents. To determine the top 50 most relevant results, the query engine must find all potential matches for "balcony" or "planter", calculate a relevance score for each, and then sort them. This scoring and sorting process across a large number of candidates within the index is computationally intensive, even though only 50 documents are ultimately retrieved. The broadness of the terms combined with the overhead of relevance scoring makes the search phase more expensive.

Geospatial Queries

Charges for geospatial queries are determined by the complexity of the query and the specific conditions and sorting utilized, as well as the resources required to scan all the data in the relevant geospatial index.

The following table delineates the charges associated with executing geospatial queries for a maps application. This example assumes a collection of 1,000,000 documents representing points of interest, with each point of interest document being 1KiB, and a corresponding geospatial index size of approximately 60MiB.

db.pois.find(
  {location: {
    $near: {
      $geometry: <point>,
      $maxDistance: 10000
    }}})

db.pois.find(
  {location: {
    $near: {
      $geometry: <point>,
      $maxDistance: 100000
    }}}).limit(10)

These examples illustrate how different query patterns impact costs:

• 10km Radius Query: This query scans a smaller geographic area. Since it finds only 10 matches within this 10km radius, the query engine has to do less work to identify and order these points by distance (as $near implicitly does). This results in a lower index scan cost of 6 read units.
• 100km Radius Query: This query covers a much larger area. The query finds 1,000 potential matches within the 100km radius. To fulfill the request for the top 10 closest points, the query engine needs to calculate distances and sort a much larger set of candidates (up to 1,000 points) within the index. This additional computational work of sifting through and ranking many more points is why the geospatial query cost is higher at 9 read units, compared to the 10km query, despite also returning only 10 documents.

Point reads

Example billing for point reads:

• Point read of a single 1 KiB document. Consumes: 1 read unit
• Point read of a single 4 KiB document. Consumes: 1 read unit
• Point read of a single 1 MiB document. Consumes: 256 read units
• Point read of 100 documents, 1 KiB each. Consumes: 100 read units

Scanning

The following examples include scenarios that scan documents or index entries.

Scanning Documents

• Query which scans 100 documents, 1 KiB each. Consumes: 25 read units

Scanning indexes

The scanning cost, in terms of bytes, is the same regardless of whether it is a document or index being scanned. However, index entries are often smaller in size. As a result, they can often provide a more cost effective way of scanning data.

• Query which scans 100 index entries, 1 KiB each. Consumes: 25 read units.
• Query which scans 100 index entries, 128 bytes each. Consumes: 4 read units.

Minimum document or index entry size

In certain situations it may not be necessary to read the contents of a document or index entry to satisfy a query. This includes simple count queries like counting the total number of documents in a collection. In these situations, a minimum cost of 32 bytes applies per item scanned.

• Count the number of documents in a collection. The query scans 1000 items in the collection. Consumes: 8 read units.

Combination of scanning and point reads

Many queries perform a combination of scanning and point reads to satisfy an operation.

• Query which scans 128 index entries, 256 bytes each and performs a point read of 128 documents, 4 KiB each. Consumes: 136 read units, comprised of:
  • 128 read units for point reads
  • 8 read units for index scans

Query Explain

Query Explain helps you understand how the database executes your queries. The details provided can help you optimize your queries.

The following costs apply when using Query Explain:

• Query Explain which executes the query: Query cost applies.
• Query Explain using plan only option. Consumes: 1 read unit (minimum cost of a query)

Write operations

Write operations (creates, updates and deletes) are charged based on the size of the documents and indexes being created, modified, or deleted during the operation. Write operations consume write units. Write units are calculated in 1 KiB tranches.

Simple write operations, like update by document ID, only incur the cost of the writes. Write operations which require querying to satisfy the operation will additionally incur the read costs associated with the query.

See the following examples.

Creates

• Create a new 10 KiB document with no indexes. Consumes: 10 write units
• Create a 1 KiB document with 1 index entry of 256 bytes on the collection. Consumes: 2 write units

Updates

• Find a 10 KiB document by document ID and update with no indexes on the collection. Consumes: 10 write units
• Find a 1 KiB document by document ID and update 1 field with 1 index entry of 256 bytes. Consumes: 3 write units. Note: Updating an index entry in this situation consumes 2 write units one to delete and one to recreate the index entry.
• Find a 1 KiB document by document ID and update nothing (no changes). Consumes: 1 write units (the minimum write costs)
• Query all 1 KiB documents in a collection, which scans 1,000 documents, and insert a new 256 bytes field with no indexes on the collection: 1000 read units and 1000 writes units.

Deletes

• Delete a 1 KiB document, which has 1 index on the collection. Consumes: 2 write units
• Delete a 1 KiB document, which has no indexes on the collection. Consumes: 1 write unit

Index builds

Index builds charge for the index entries created or modified during the build operation. These costs are incurred anytime an index definition is added or removed. The index entries are billed identically to writes incurring 1 write unit per 1KiB.

• Create a new index for a collection containing 500 documents, index entries created are 1 KiB each. Consumes 500 write units.
• Delete an existing index for a collection containing 500 documents, index entries deleted are 1KiB each. Consumes 500 write units.