1403 app dev series - session 5 - analytics

Preview:

DESCRIPTION

 

Citation preview

Application Development SeriesBack to BasicsReporting & Analytics

Daniel Roberts@dmroberts

#MongoDBBasics

2

• Recap from last session

• Reporting / Analytics options

• Map Reduce

• Aggregation Framework introduction– Aggregation explain

• mycms application reports

• Geospatial with Aggregation Framework

• Text Search with Aggregation Framework

Agenda

3

• Virtual Genius Bar

– Use the chat to post questions

– EMEA Solution Architecture / Support team are on hand

– Make use of them during the sessions!!!

Q & A

Recap from last time….

5

Indexing

• Indexes• Multikey, compound,

‘dot.notation’

• Covered, sorting

• Text, GeoSpatial

• Btrees

>db.articles.ensureIndex( { author : 1, tags : 1 } )

>db.user.find({user:"danr"}, {_id:0, password:1})

>db.articles.ensureIndex( { location: “2dsphere” } )

>>db.articles.ensureIndex( { "$**" : “text”,

name : “TextIndex”} )

options db.col.ensureIndex({ key : type})

6

Index performance / efficiency

• Examine index plans

• Identity slow queries

• n / nscanned ratio

• Which index used.

operators .explain() , db profiler> db.articles.find(

{author:'Dan Roberts’})

.sort({date:-1}).explain()

> db.setProfilingLevel(1, 100){ "was" : 0, "slowms" : 100, "ok" : 1 }

> db.system.profile.find().pretty()

Reporting / Analytics options

8

• Query Language– Leverage pre aggregated documents

• Aggregation Framework– Calculate new values from the data that we have– For instance : Average views, comments count

• MapReduce– Internal Javascript based implementation– External Hadoop, using the MongoDB connector

• A combination of the above

Access data for reporting, options

9

• Immediate results– Simple from a query

perspective.

– Interactions collection

Pre Aggregated Reports

{‘_id’ : ObjectId(..),

‘article_id’ : ObjectId(..), ‘section’ : ‘schema’,

‘date’ : ISODate(..),‘daily’: { ‘views’ : 45,

‘comments’ : 150 } ‘hours’ : { 0 : { ‘views’ : 10 }, 1 : { ‘views’ : 2 }, … 23 : { ‘views’ : 14,

‘comments’ : 10 } }}

> db.interactions.find(

{"article_id" : ObjectId(”…..")},{_id:0, hourly:1}

)

10

• Use query result to display directly in application– Create new REST API

– D3.js library or similar in UI

Pre Aggregated Reports

{"hourly" : {

"0" : {

"view" : 1},"1" : {

"view" : 1},……"22" : {

"view" : 5},"23" : {

"view" : 3}

}}

Map Reduce

12

• Map Reduce– MongoDB – JavaScript

• Incremental Map Reduce

Map Reduce

//Map Reduce Example> db.articles.mapReduce(

function() { emit(this.author, this.comment_count); },function(key, values) { return Array.sum (values) },{

query : {},out: { merge: "comment_count" }

})

Output

{ "_id" : "Dan Roberts", "value" : 6 }{ "_id" : "Jim Duffy", "value" : 1 }{ "_id" : "Kunal Taneja", "value" : 2 }{ "_id" : "Paul Done", "value" : 2 }

13

MongoDB – Hadoop Connector

Hadoop Integration

Primary

Secondary

Secondary

HDFS

Primary

Secondary

Secondary

Primary

Secondary

Secondary

Primary

Secondary

Secondary

HDFS HDFS HDFS

MapReduce MapReduce MapReduce MapReduce

MongoS MongoSMongoS

Application ApplicationApplication

Application Dash Boards / Reporting

1) Data Flow, Input / Output via Application Tier

Aggregation Framework

15

• Multi-stage pipeline– Like a unix pipe –

• “ps -ef | grep mongod”

– Aggregate data, Transform documents

– Implemented in the core server

Aggregation Framework

//Find out which are the most popular tags…db.articles.aggregate([

{ $unwind : "$tags" },{ $group : { _id : "$tags" , number : { $sum : 1 } } },{ $sort : { number : -1 } }

])

Output

{ "_id" : "mongodb", "number" : 6 }{ "_id" : "nosql", "number" : 3 }{ "_id" : "database", "number" : 1 }{ "_id" : "aggregation", "number" : 1 }{ "_id" : "node", "number" : 1 }

16

In our mycms application..

//Our new python example@app.route('/cms/api/v1.0/tag_counts', methods=['GET'])def tag_counts():

pipeline = [ { "$unwind" : "$tags" },{ "$group" : { "_id" : "$tags" ,

"number" : { "$sum" : 1 } } },{ "$sort" : { "number" : -1 } }]

cur = db['articles'].aggregate(pipeline, cursor={})# Check everything okif not cur:

abort(400) # iterate the cursor and add docs to a dict tags = [tag for tag in cur] return jsonify({'tags' : json.dumps(tags, default=json_util.default)})

17

• Pipeline and Expression operators

Aggregation operators

Pipeline

$match $sort$limit$skip$project$unwind$group$geoNear$text$search

Tip: Other operators for date, time, boolean and string manipulation

Expression

$addToSet

$first$last$max$min$avg$push$sum

Arithmetic

$add$divide$mod$multiply$subtract

Conditional

$cond$ifNull

Variables

$let$map

18

• What reports and analytics do we need in our application?– Popular Tags– Popular Articles– Popular Locations – integration with Geo Spatial– Average views per hour or day

Application Reports

19

• Unwind each ‘tags’ array

• Group and count each one, then Sort

• Output to new collection– Query from new collection so don’t need to compute for

every request.

Popular Tags

db.articles.aggregate([{ $unwind : "$tags" },{ $group : { _id : "$tags" , number : { $sum : 1 } } },{ $sort : { number : -1 } },{ $out : "tags"}

])

20

• Top 5 articles by average daily views– Use the $avg operator – Use use $match to constrain data range

• Utilise with $gt and $lt operators

Popular Articles

db.interactions.aggregate([ {

{$match : { date : { $gt : ISODate("2014-02-

20T00:00:00.000Z")}}},{$group : {_id: "$article_id", a : { $avg : "$daily.view"}}},{$sort : { a : -1}},{$limit : 5}

]);

21

• Use Explain plan to ensure the efficient use of the index when querying.

Aggregation Framework Explain

db.interactions.aggregate([{$group : {_id: "$article_id", a : { $avg : "$daily.view"}}},{$sort : { a : -1}},{$limit : 5}

],{explain : true}

);

22

Explain output…

{"stages" : [

{"$cursor" : { "query" : … }, "fields" : { … },

"plan" : {"cursor" : "BasicCursor","isMultiKey" : false,"scanAndOrder" : false,"allPlans" : [

{"cursor" :

"BasicCursor",

"isMultiKey" : false,

"scanAndOrder" : false}

]}

}},…

"ok" : 1}

Geo Spatial & Text Search Aggregation

24

• $text operator with aggregation framework– All articles with MongoDB– Group by author, sort by comments count

Text Search

db.articles.aggregate([ { $match: { $text: { $search: "mongodb" } } }, { $group: { _id: "$author", comments:

{ $sum: "$comment_count" } } }{$sort : {comments: -1}},

])

25

• $geoNear operator with aggregation framework– Again use geo operator in the $match statement.– Group by author, and article count.

Utilise with Geo spatial

db.articles.aggregate([ { $match: { location: { $geoNear :

{ $geometry :{ type: "Point" ,coordinates : [-0.128,

51.507] } }, $maxDistance :5000} }

}, { $group: { _id: "$author", articleCount: { $sum: 1 } } } ])

Summary

27

• Aggregating Data…– Map Reduce– Hadoop– Pre-Aggregated Reports– Aggregation Framework

• Tune with Explain plan

• Compute on the fly or Compute and store

• Geospatial

• Text Search

Summary

28

– Operations for you application– Scale out

– Availability

– How do we prepare of production

– Sizing

Next Session – 3th April

Recommended