open-source, high-performance,

document-oriented database
Antoine Girbal
antoine@10gen.com
Data Store Analytics
RDBMS
Helper
RDBMS
(MemCache, New Gen.
Application layer)
(Oracle, MySQL) OLAP
(vertica, aster,
greenplum)

Non-relational
Operational Stores
(“NoSQL”)
NoSQL Really Means:


non-relational, next-generation
operational datastores and databases

focus on scalability

ease of modeling and changing data

(also no sql syntax, thanks!)
 no joins
+ no complex transactions

Horizontally Scalable
Architectures
 JSON-style Documents
Just like a light and friendly XML
{ hello: “world” }

represented as BSON

\x16\x00\x00\x00\x02hello\x00
\x06\x00\x00\x00world\x00\x00
http://bsonspec.org
 Flexible “Schemas”
In collection db.posts:

{author: “mike”,
links: 3,
date: "Sun Jul 18 2010 14:40:20 GMT-0700 (PDT)"
text: “blah blah”}

{author: “eliot”,
date: "Sun Jul 18 2010 14:40:22 GMT-0700 (PDT)"
text: “Here is MongoDB ...”,
views: 10}

Potentially all documents in the same collection

 Embedded Document
little need for joins
or transactions across documents!
{
_id: ObjectId("4d1009c7262bb4b94af1cea4")
author_id: “1346”,
date: "Sun Jul 18 2010 14:40:20 GMT-0700 (PDT)",
title: “my story”
text: “once upon a time ...”,
tags: [“novel”,”english”],
Comments:[
{user_id: 234, text: “awesome dude”},
{user_id: 1235, text: “that made me cry”}]
}
 Data Model
Normalized: Natural: Just
Many tables Collections BLOBS
Row/Column Documents

PERFORMANCE
As fast as BLOBs if well
Easy to scale (Dynamo)
Can't scale modeled, horizontal
can easily use caching
scalability

FEATURES
Most regular SQL
(too) many No feature
features (satisfies 90% of
users)
 Features
• Complex querying
• Atomic updates with modifiers
• Indexing (unique, compound, Geo)
• Aggregation and Map / Reduce
• Capped Collections
• Powerful Shell (Javascript)
• GridFS: file storage
 Replication
Using Replica Set:
- pool of servers with 1 master
- automatic master election and failover
- distributed reads (slaveOk)

Client Client

master

slave slave slave

 Sharding
For large datasets, or write heavy system
Shards

mongod

mongod

mongod

mongod

mongod

mongod

mongod

mongod

mongod
...
Config ...
Servers

mongod

mongod

mongod mongos mongos ...

client
 Support
• OS: Mac OS X, Windows, Linux, Solaris,
32/64 bits
• Drivers: C, C#, C++, Haskell, Java,
Javascript, Perl, PHP, Python, Ruby, Scala...
+ community drivers
• Open-source project with active
community, Wiki, Google Group, 10gen
consulting / support
 Production Examples

Shutterfly 
Intuit

Fourquare 
College humor

Craigslist 
Evite

bit.ly 
Disqus

IGN 
Justin.tv

Sourceforge 
Heartbeat

Etsy 
Hot Potato

the New York Times 
Eventbrite

Business Insider 
Sugar crm

Gilt Groupe 
Electronic Arts

...
 New Post
> post = {author: "mike",
... date: new Date(),
... text: "my blog post",
... tags: ["mongodb", "intro"]}

> db.posts.save(post)

> db.posts.findOne()
{ "_id" : ObjectId("4d2f944103e8fdbb36f6d205"),
"author" : "mike",
"date" : ISODate("2011-01-14T00:08:49.933Z"),
"text" : "my blog post",
"tags" : ["mongodb","intro"]}
A Quick Aside

_id • special key

• present in all documents
• unique across a
Collection
• any type you want
 Update
> db.posts.update({_id: post._id},
... { $set: {author: "tony"}})

> c = {author: "eliot", date: new Date(), text:

"great post!"}
> db.posts.update({_id: post._id},
... { $push: {comments: c}})

> db.posts.update({_id: post._id},

... { $inc: {views: 1}})
> db.posts.findOne()
Querying
{
"_id" : ObjectId("4d2f944103e8fdbb36f6d205"),
"author" : "tony",
"comments" : [
{
"author" : "eliot",
"date" : ISODate("2011-01-14T00:13:52.463Z"),
"text" : "great post!"
}
],
"date" : ISODate("2011-01-14T00:08:49.933Z"),
"tags" : [
"mongodb",
"intro"
],
"text" : "my blog post",
"views" : 1
}
 More Querying
Find by Author
> db.posts.find({author: "tony"})

10 most recent posts:

> db.posts.find().sort({date: -1}).limit(10)

Posts since April 1st:

> april_1 = new Date(2010, 3, 1)
> db.posts.find({date: {$gt: april_1}})

Adding an index to speed up:

> db.posts.ensureIndex({author: 1})
> db.posts.ensureIndex({date: 1})
 More Querying

Find with regexp:

> db.posts.find({text: /post$/})

Find within array:

> db.posts.find({tags: "intro"})
> db.posts.ensureIndex({tags: 1})

Find within embedded object:

> db.posts.find({"comments.author": "eliot"})
> db.posts.ensureIndex({"comments.author": 1})
 More Querying
Counting:
> db.posts.find().count()
> db.posts.find({author: "tony"}).count()

Paging:
> page = 2
> page_size = 15
> db.post.find().limit(page_size).skip(page * page_size)

Advanced operators:
$gt, $lt, $gte, $lte, $ne, $all, $in, $nin, $where
> db.posts.find({$where: "this.author == 'tony' || this.title == 'foo'"})
 Download MongoDB
Current 1.6, soon 1.8
http://www.mongodb.org

and let us know what you think

@mongodb