Optimization - The Definitive Guide to MongoDB - page 238

■ if you see a detailed query explanation that shows a significantly larger number of scanned records ( nscanned )
than it returns ( n ), then that query is probably a candidate for indexing.

Note

The next step is to determine whether adding an index on the Tags field improves the query's performance:

> db.posts.ensureIndex({Tags:1})

Now run the explain() function again to see the effect of adding the index:

> db.posts.find({Tags:"even"}).explain()
{
"cursor" : "BtreeCursor Tags_1",
"isMultiKey" : true,
"n" : 14997,
"nscannedObjects" : 14997,
"nscanned" : 14997,
"nscannedObjectsAllPlans" : 14997,
"nscannedAllPlans" : 14997,
"scanAndOrder" : false,
"indexOnly" : false,
"nYields" : 0,
"nChunkSkips" : 0,
"millis" : 4,
"indexBounds" : {
"Tags" : [
[
"even",
"even"
]
]
},
"server" : "Pixl.local:27017"
}

The performance of the query has improved significantly. You can see that the query is now using a BtreeCursor
driven by the Tags_1 index. The number of scanned records has been reduced from 29,997 records to the same 14,997
records you expect the query to return, and the execution time has dropped to 4ms.

■ the most common index type, and the only one used by mongoDB, is the btree (binary-tree). a BtreeCursor is
a mongoDB data cursor that uses the binary tree index to navigate from document to document. Btree indexes are very
common in database systems because they provide fast inserts and deletes, yet also provide reasonable performance
when used to walk or sort data.

Note