[SERVER-55575] Optimize $group on meta $min/$max on measurements for time-series collections - MongoDB Jira

XML

Word

Printable

For a time-series collection, we automatically bucket fields and have a control block with the min and max of each, like so:

{ _id: ObjectId(...),

  control: {

    version: 1, // in case we want to change the format in a later release

    min: { _id: ..., time: ISODate("2020-09-09T16:54:04.174Z"), field0: 0.01, ... },

    max: { _id: ..., time: ISODate("2020-09-09T16:54:04.450Z"), field0: 0.99, ... },

},

  meta: ...,

  data: { _id: { 0:..., 1: ..., ...},

          time: { 0: ISODate("2020-09-09T16:54:04.174Z"),

                  1: ISODate("2020-09-09T16:54:04.450Z"), ...},

          field0: { 0: 0.99, 1: 0.01,...},

          field1: { 2: "high", 42: "low",...}, // Missing fields are omitted

          ...}

For a query over that collection like this:

db.example.aggregate([{$group: {_id: "$meta.x", field0: {$max: "$field0"}}}])

We should be able to avoid unpacking or looking at "data" at all.

is related to

SERVER-55624 Track materialized counts and sums for time-series measurements