2011 Mongo FR - Indexing in MongoDB

MongoDB Indexing and Query Optimizer Details Antoine Girbal Mongo FR March 23, 2011

What will we cover?
Many details of how indexing and the query optimizer work
A full understanding of these details is not required to use mongo, but this knowledge can be helpful when making optimizations.
We’ll discuss functionality of Mongo 1.8 (for our purposes pretty similar to 1.6 and almost identical to 1.7 edge).
Much of the material will be presented through examples.
Diagrams are to aid understanding – some details will be left out.

Btree (conceptual diagram) 1 2 3 4 5 6 7 8 9 {_id:4,x:6}

Find One Document
db.c.find( {x:6} ).limit( 1 )
Index {x:1}

Find One Document 1 2 3 4 5 6 7 8 9 6 ? {_id:4,x:6}

Find One Document > db.c.find( {x:6} ).limit( 1 ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 1, "nscannedObjects" : 1, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 6, 6 ] ] } } Uses a btree cursor to find the object. Index ranges are around a single value.

Find One Document 1 2 3 4 5 6 6 6 9 6 ? {_id:4,x:6} Now we have duplicate x values

Equality Match
db.c.find( {x:6} )
Index {x:1}
Several documents to be returned

Equality Match 9 1 2 3 4 5 6 6 6 6 ? {_id:4,x:6} {_id:5,x:6} {_id:1,x:6}

Equality Match > db.c.find( {x:6} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 3, "nscannedObjects" : 3, "n" : 3, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 6, 6 ] ] } }

Equality Match 1 2 3 4 5 6 6 6 9 6 ?

Full Document Matcher
db.c.find( {x:6,y:1} )
Index {x:1}
Object content needs to be checked

Full Document Matcher 9 1 2 3 4 5 6 6 6 6 ? {y:4,x:6} {y:5,x:6} {y:1,x:6}

Full Document Matcher > db.c.find( {x:6,y:1} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 3, "nscannedObjects" : 3, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 6, 6 ] ] } } Documents for all matching index keys are scanned, but only one document matched on non index keys.

Range Match
db.c.find( {x:{$gte:4,$lte:7}} )
Index {x:1}

Range Match 8 1 2 3 4 5 6 7 9 4 <= ? <= 7

Range Match > db.c.find( {x:{$gte:4,$lte:7}} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 4, "nscannedObjects" : 4, "n" : 4, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 4, 7 ] ] } }

Range Match 1 2 3 4 5 6 7 8 9

Exclusive Range Match
db.c.find( {x:{$gt:4,$lt:7}} )
Index {x:1}
Range of index is same as inclusive range match
but boundaries are not scanned nor returned

Multikeys
db.c.find( {x:{$gt:7}} )
Index {x:1}
documents contain lists with several values like [8,9].

Multikeys 1 2 3 4 5 6 7 9 ? > 7 {_id:4,x:[8,9]} 8

Multikeys > db.c.find( {x:{$gt:7}} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 2, "nscannedObjects" : 2, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ 7, 1.7976931348623157e+308 ] ] } } All keys in valid range are scanned, but the matcher rejects duplicate documents making n == 1.

Multikeys 1 2 3 4 5 6 7 8 9

Range Types
Explicit inequality
db.c.find( {x:{$gt:4,$lt:7}} )
db.c.find( {x:{$gt:4}} )
db.c.find( {x:{$ne:4}} )
Regular expression prefix
db.c.find( {x:/^a/} )
Data type
db.c.find( {x:/a/} )

Range Types db.c.find( {x:/â/} ) "indexBounds" : { "x" : [ [ "a", "b" ], [ /â/, /â/ ] ] } 2 ranges scanned of 2 different types: string and regex

Range Types db.c.find( {x:/a/} ) "indexBounds" : { "x" : [ [ "", { } ], [ /a/, /a/ ] ] } Here the index only helps to restrict type, not efficient in practice

Set Match
db.c.find( {x:{$in:[3,6]}} )
Index {x:1}

Set Match 8 1 2 3 4 5 6 7 9 3 , 6

Set Match > db.c.find( {x:{$in:[3,6]}} ).explain() { "cursor" : "BtreeCursor x_1 multi", "nscanned" : 3, "nscannedObjects" : 2, "n" : 2, "millis" : 8, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 3, 3 ], [ 6, 6 ] ] }} Why is nscanned 3? This is an algorithmic detail, when there are disjoint ranges for a key nscanned may be higher than the number of matching keys.

Set Match 1 2 3 4 5 6 7 8 9

All Match
db.c.find( {x:{$all:[3,6]}} )
Index {x:1}

All Match 8 1 2 3 4 5 6 7 9 3 ? {_id:4,x:[3,6]}

All Match > db.c.find( {x:{$all:[3,6]}} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 1, "nscannedObjects" : 1, "n" : 1, "millis" : 0, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ 3, 3 ] ] } } The first entry in the $all match array is always used for index bounds. Note this may not be the least numerous indexed value in the $all array.

All Match 1 2 3 4 5 6 7 8 9

Limit
db.c.find( {x:{$lt:6},y:3} ).limit( 3 )
Index {x:1}

Limit 8 1 2 3 4 5 6 7 9 6 ? < y:3 y:1 y:3 y:3 y:3

Limit > db.c.find( {x:{$lt:6},y:3} ).limit( 3 ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 4, "nscannedObjects" : 4, "n" : 3, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ -1.7976931348623157e+308, 6 ] ] } } Scan until three matches are found, then stop.

Skip
db.c.find( {x:{$lt:6},y:3} ).skip( 3 )
Index {x:1}

Skip 8 1 2 3 4 5 6 7 9 6 ? < y:3 y:1 y:3 y:3 y:3

Skip > db.c.find( {x:{$lt:6},y:3} ).skip( 3 ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 5, "nscannedObjects" : 5, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ -1.7976931348623157e+308, 6 ] ] } } All skipped documents are scanned.

Sort
db.c.find( {x:{$lt:6}} ).sort( {x:1} )
Index {x:1}
Sorting along index key uses index btree ordering

Sort 8 1 2 3 4 5 6 7 9 6 ? < y:3 y:1 y:3 y:3 y:3

Sort > db.c.find( {x:{$lt:6},y:3} ).sort( {x:1} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 5, "nscannedObjects" : 5, "n" : 4, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ -1.7976931348623157e+308, 6 ] ] } } Find uses the btree cursor to easily sort data

Sort
db.c.find( {x:{$lt:6}} ).sort( {y:1} )
Index {x:1}
Using non-indexed key to sort data will need to scan & order

Sort 8 1 2 3 4 5 6 7 9 6 ? < y:3 y:1 y:3 y:3 y:3

Sort Results are sorted on the fly to match requested order. The scanAndOrder field is only printed when its value is true. > db.c.find( {x:{$lt:6},y:3} ).sort( {y:1} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 5, "nscannedObjects" : 5, "n" : 4, "scanAndOrder" : true, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : true, "indexOnly" : false, "indexBounds" : { "x" : [ [ -1.7976931348623157e+308, 6 ] ] } }

Sort and scanAndOrder
With “scanAndOrder” sort, all documents must be touched even if there is a limit spec.
With scanAndOrder, sorting is performed in memory and the memory footprint is constrained by the limit spec if present.

Count
Count uses the same indexed but only scans in the index, not the object data in storage
With some operators the full document must be checked. Some of these cases:
$all
$size
array match
Negation - $ne, $nin, $not, etc.
With current semantics, all multikey elements must match negation constraints
Multikey de duplication works without loading full document

Covered Indexes
db.c.find( {x:6}, {x:1,_id:0} )
Index {x:1}
Id would be returned by default, but isn’t in the index so we need to exclude to return only indexed fields.

Covered Indexes > db.c.find( {x:6}, {x:1,_id:0} ).explain() { "cursor" : "BtreeCursor x_1", "nscanned" : 1, "nscannedObjects" : 1, "n" : 1, "millis" : 0, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : true, "indexBounds" : { "x" : [ [ 6, 6 ] ] } } IndexOnly is true, and isMultiKey must be false. Currently we set isMultiKey to true the first time we save a doc where the field is a multikey array.

Two Equality Bounds
db.c.find( {x:5,y:’c’} )
Index {x:1,y:1}

Two Equality Bounds ? 5 c 1 b 3 d 4 g 5 d 5 f 6 c 7 a 9 b 5 c

Two Equality Bounds > db.c.find( {x:5,y:'c'} ).explain() { "cursor" : "BtreeCursor x_1_y_1", "nscanned" : 1, "nscannedObjects" : 1, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 5, 5 ] ], "y" : [ [ "c", "c" ] ]}} 2 Ranges applied to narrow down the data to scan.

Two Equality Bounds ? 1 b 3 d 4 g 5 c 5 d 5 f 5 c 6 c 7 a 9 b

Two Set Bounds
db.c.find( {x:{$in:[5,9]},y:{$in:[’c’,’f’]}} )
Index {x:1,y:1}

Two Set Bounds , , , 5 c 1 b 3 d 4 g 5 d 5 f 6 c 7 a 9 f 5 c 5 f 9 c 9 f

Two Set Bounds > db.c.find( {x:{$in:[5,9]},y:{$in:['c','f']}} ).explain() { "cursor" : "BtreeCursor x_1_y_1 multi", "nscanned" : 5, "nscannedObjects" : 3, "n" : 3, "millis" : 0, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, ... "indexBounds" : { "x" : [ [ 5, 5 ], [ 9, 9 ] ], "y" : [ [ "c", "c" ], [ "f", "f" ] ] } }

Disjoint $or Criteria
db.c.find( {$or:[{x:5},{y:’d’}]} )
Indexes {x:1}, {y:1}
Does 2 sequential find for each clause
Must not return same document twice, so it checks whether it satisfies previous clause

Disjoint $or Criteria ? ? 1 b 3 d 4 g 5 d 6 a 7 e 9 f 5 c d 7 g 5 1 b 3 d 4 g 5 d 6 a 7 e 9 f 5 c 7 g

Disjoint $or Criteria > db.c.find( {$or:[{x:5},{y:'d'}]} ).explain() { "clauses" : [ { "cursor" : "BtreeCursor x_1", "nscanned" : 2, "nscannedObjects" : 2, "n" : 2, "millis" : 0, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "x" : [ [ 5, 5 ] ] } }, { "cursor" : "BtreeCursor y_1", "nscanned" : 2, "nscannedObjects" : 2, "n" : 1, "millis" : 1, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { "y" : [ [ "d", "d" ] ] } }], "nscanned" : 4, "nscannedObjects" : 4, "n" : 3, "millis" : 1}

Unindexed $or Clause
db.c.find( {$or:[{x:5},{y:’d’}]} )
Index {x:1} (no index on y)

Unindexed $or Clause > db.c.find( {$or:[{x:5},{y:'d'}]} ).explain() { "cursor" : "BasicCursor", "nscanned" : 9, "nscannedObjects" : 9, "n" : 3, "millis" : 0, "nYields" : 0, "nChunkSkips" : 0, "isMultiKey" : false, "indexOnly" : false, "indexBounds" : { } } Since y is not indexed, we must do a full collection scan to match y:’d’. Since a full scan is required, we don’t use the index on x to match x:5.

Automatic Index Selection (Query Optimizer)

Optimal Index
find( {x:5} )
Index {x:1}
Index {x:1,y:1}
find( {x:5} ).sort( {y:1 } )
Index {x:1,y:1}
find( {} ).sort( {x:1} )
Index {x:1}
find( {x:{$gt:1,$lt:7}} ).sort( {x:1} )
Index {x:1}

Optimal Index
Rule of Thumb
No scanAndOrder
All fields with index useful constraints are indexed
If there is a range or sort it is the last field of the index used to resolve the query
If multiple optimal indexes exist, one chosen arbitrarily.

Multiple Candidate Indexes
find( {x:4,y:’a’} )
Index {x:1} or {y:1}?
find( {x:4} ).sort( {y:1} )
Index {x:1} or {y:1}?
Note: {x:1,y:1} is optimal
find( {x:{$gt:2,$lt:7},y:{$gt:’a’,$lt:’f’}} )
Index {x:1,y:1} or {y:1,x:1}?

The only index selection criterion is nscanned
find( {x:4,y:’a’} )
Index {x:1} or {y:1} ?
If fewer documents match {y:’a’} than {x:4} then nscanned for {y:1} will be less so we pick {y:1}
find( {x:{$gt:2,$lt:7},y:{$gt:’b’,$lt:’f’}} )
Index {x:1,y:1} or {y:1,x:1} ?
If fewer distinct values of 2 < x < 7 than distinct values of ‘b’ < y < ‘f’ then {x:1,y:1} chosen (rule of thumb)

The only index selection criterion is nscanned
Pretty good, but doesn’t cover every case, eg
Overhead of using an index versus doing a collection scan
Cost of scanAndOrder vs ordered index
Cost of loading full document vs just index key
Cost of scanning adjacent btree keys vs non adjacent keys/documents

Competing Indexes
At most one query plan per index
Run in interleaved fashion
Plans kept in a priority queue ordered by nscanned. We always continue progress on plan with lowest nscanned.

Competing Indexes
Run until one plan returns all results or enough results to satisfy the initial query request (based on soft limit spec / data size requirement for initial query).
We only allow plans to compete in initial query. In getMore, we continue reading from the index cursor established by the initial query.

“ Learning” a Query Plan
When an index is chosen for a query the query’s “pattern” and nscanned are recorded
find( {x:3,y:’c’} )
{Pattern: {x:’equality’, y:’equality’}, Index: {x:1}, nscanned: 50}
find( {x:{$gt:5},y:{$lt:’z’}} )
{Pattern: {x:’gt bound’, y:’lt bound’}, Index: {y:1}, nscanned: 500}

“ Learning” a Query Plan
When a new query matches the same pattern, the same query plan is used
find( {x:5,y:’z’} )
Use index {x:1}
find( {x:{$gt:20},y:{$lt:’b’}} )
Use index {y:1}

“ Un-Learning” a Query Plan
100 writes to the collection
Indexes added / removed

Bad Plan Insurance
If nscanned for a new query using a recorded plan is much worse than the recorded nscanned for an earlier query with the same pattern, we start interleaving other plans with the current plan.
Currently “much worse” means 10x

Thanks! Feature Requests jira.mongodb.org Support groups.google.com/group/mongodb-user

2011 Mongo FR - Indexing in MongoDB

by antoinegirbal on Mar 25, 2011

Accessibility

Tags

Upload Details

Usage Rights

Statistics

2011 Mongo FR - Indexing in MongoDB — Presentation Transcript