Slow performance in mongodb compound geospatial index

[sabber sabber]

I have a collections called "location" one of which documents look like :

    {

    "_id" : ObjectId("5726ad18bdb861058819ea00"),

    "loc" : {

    "type" : "Point",

    "coordinates" : [

    -90.426407,

    34.850685

    ]

    },

    "name" : "restaurant name",

    "tags" : [

    "Chinese",

    "thai",

    "American",

    "indian"

    ]

    }

and i have the following index :

    db.location.getIndexes()

    [

    {

    "v" : 1,

    "key" : {

    "_id" : 1

    },

    "name" : "_id_",

    "ns" : "chatx.location"

    },

    {

    "v" : 1,

    "key" : {

    "loc" : "2dsphere",

    "tags" : 1

    },

    "name" : "loc_2dsphere_tags_1",

    "ns" : "chatx.location",

    "2dsphereIndexVersion" : 3

    }

    ]

i used tags to identify restaurant food type types. 

I have been trying to find out nearest restaurants based on food(s) type. My query is :

    db.runCommand(

       {

         geoNear: "location",

         near: { type: "Point", coordinates: [-90.426407, 34.850685] },

         limit:10,

         maxDistance:5000,

         spherical: true,

         query: { tags:{$all:["chinese", "thai"]}}

       }

    )

In order to do that i used $geoNear (which is very good performative in 3.2 version ). However when i combine with tags (using $all) , query then becomes slower, which makes sense, because $all searches all the documents based on first element then refines based on other elements. The main reason for a query becomes slower is increasing limit (e.g: 10 to 100 or more) and array elements["thai", "indian", .....]. For example if i search restaurants with 4 food type tags, query becomes super slow rather than if i search restaurants with 1 food type tags.

I was wondering if anybody can help me either figuring out best schema for this collection or providing better query options. 

[Kevin Adistambha]

Hi,

In order to do that i used $geoNear (which is very good performative in 3.2 version ). However when i combine with tags (using $all) , query then becomes slower

Could you clarify a few things:

• What are the numbers on “fast”, “slow”, and “super slow” (e.g. in milliseconds)?
• How many documents are in the “location” collection?
• What is your deployment topology (e.g. standalone, replica set, sharded cluster)?
• What is the exact version of your MongoDB (i.e. 3.2.6 is the latest in the 3.2 series).
• Could you also please post the output of the db.runCommand({geoNear ...}) query?

which makes sense, because $all searches all the documents based on first element then refines based on other elements

This is correct, as per the Performance section in the $all manual.

Best regards,
Kevin

​