Introduction to

"MongoDB is a scalable, high-

performance, open source,
schema-free, document-
oriented database."
Goal: be the best database for
(most) web apps

...not the best DB for everything.

What do Web Apps Want?

 Performance
 Scalability
 Availability
What do developers want?
What else do developers want?
 Easier development
 Less maintenance (at 3am)
 Great support
Goals
 Easy-to-use
 Fast
 Always available
 Easy to scale
Goals
 Easy-to-use
 Fast
 Always available
 Easy to scale
5-minute install

Download binaries from

http://www.mongodb.org
and run!
Mongo knows JavaScript
$ ./mongo
MongoDB shell version: 1.3.2
url: test
connecting to: test
type "help" for help
>
Mongo knows JavaScript
$ ./mongo
MongoDB shell version: 1.3.2
url: test
connecting to: test
type "help" for help
> x = 123+1
124
>
mongo.kylebanker.com
Basic Usage
$ ./mongo
MongoDB shell version: 1.3.2
url: test
connecting to: test
type "help" for help
>
Basic Usage
$ ./mongo
MongoDB shell version: 1.3.2
url: test
connecting to: test
type "help" for help
> db
test
>
Database ≈ Database
> use foo
switched to db foo
>
Database ≈ Database
> use foo
switched to db foo
> db
foo
>
Collection ≈ Table
> db.my_collection
foo.my_collection
>
Collection ≈ Table
> db.my_collection
foo.my_collection
>
Collection ≈ Table
> db.my_collection
foo.my_collection
> db.jzxgtx.find()
>
Collection ≈ Table
> db.my_collection
foo.my_collection
> db.jzxgtx.find()
> db.pltzbt.count()
0
>
Document ≈ Row
{
"title" : "My first blog post",
"author" : "Fred",
"content" : "Hello, world!",
"comments" : []
}
Documents
Language Document
JavaScript {"foo" : "bar"}
Perl {"foo" => "bar"}
PHP array("foo" => "bar")
Python {"foo" : "bar"}
Ruby {"foo" => "bar"}
Java DBObject obj = new BasicDBObject("foo", "bar");
Documents
{
"title" : "My first blog post",
"author" : "Fred",
"content" : "Hello, world!",
"comments" : []
}
Documents
> db.posts.insert({
... "title" : "My first blog post",
... "author" : "Fred",
... "content" : "Hello, world!",
... "comments" : [] })
>
Documents
> post = db.posts.findOne()
{
"_id" : ObjectId("2fe3e4d892aa73234c910bed"),
"title" : "My first blog post",
"author" : "Fred",
"content" : "Hello, world!",
"comments" : []
}
>
Documents
> post = db.posts.findOne()
{
"_id" : ObjectId("2fe3e4d892aa73234c910bed"),
"title" : "My first blog post",
"author" : "Fred",
"content" : "Hello, world!",
"comments" : []
}
>
ObjectId
Q: How do we make a unique ID quickly that is guaranteed to
be unique across multiple servers?

A:
> print(post._id)
2fe3e4d892aa73234c910bed
|------||----||--||----|
ts mac pid inc

Built-in document creation timestamp

A few examples...
Chess
{
"name" : "black king",
"symbol" : "♚",
"pos" : {
"x" : "e",
"y" : 8
}
}
Moving
> db.chess.update(

{
"name" : "black king",
"symbol" : "♚",
"pos" : {"x":"e", "y":8}
}
Moving
> db.chess.update(
... {"name" : "black king"},

{
"name" : "black king",
"symbol" : "♚",
"pos" : {"x":"e", "y":8}
}
Moving
> db.chess.update(
... {"name" : "black king"},
... {"$inc" : {"pos.y" : -1}})
>

{
"name" : "black king",
"symbol" : "♚",
"pos" : {"x":"e", "y":8}
}
Moving
> db.chess.update(
... {"name" : "black king"},
... {"$inc" : {"pos.y" : -1}})
>
Moving
> db.chess.update(
... {"name" : "black king"},
... {"$inc" : {"pos.y" : -1}})
> db.chess.find({"name" : "black king"})
{
"name" : "black king",
"symbol" : "♚",
"pos" : {"x" : "e", "y" : 7}
}
Adding Information
> db.chess.update(
... {"name" : /pawn/},
... {"$set" : {"importance" : 1}})
>
Adding Information
> db.chess.update(
... {"name" : /pawn/},
... {"$set" : {"importance" : 1}})
> db.chess.findOne({"symbol" : "♙"})
{
"name" : "white pawn a",
"symbol" : " ♙ ",
"pos" : {"x" : "a", "y" : 2},
"importance" : 1
}
Types
 null  date
 boolean  binary data
 integer  object id
 long  regular expression
 double  code
 string  max value
 array  min value
 object
Querying
> db.chess.find()
Querying

1
a h
Querying
> db.chess.find().sort(
... {"pos.y" : -1, "pos.x" : 1})
Querying
> db.chess.find(
... {"name" : /white/}).sort(
... {"pos.y" : -1, "pos.x" : 1})
Querying
> db.chess.find().sort(
... {"importance" : -1}).limit(1)
Querying
> db.chess.find().sort(
... {"importance" : -1}).limit(1).skip(1)
Query Conditions
> db.chess.find({"pos.y" : 1})
Query Conditions
> db.chess.find({"pos.y" :
... {"$gt" : 2, "$lt" : 7}})
Query Conditions
> db.chess.find({"pos.y" :
... {"$gt" : 2, "$lt" : 7},
... "pos.x" :
... {"$ne" : "b"}})
"$gte"
"<null>"
""
'$gte' or "\$gte"
Or define your own!
":gte"
"=gte"
"?gte"
Or define your own!
" gte"
Or define your own!
" gte"
" ♕ gte"
Or define your own!
" gte"
" ♕ gte"
"xgte"
Other Neat Stuff
Analytics
{
"uri" : "/blog",
"pageviews" : 0
}
Analytics - Increment Pageviews
page = db.analytics.findOne({"uri" : "/blog"});

if (page != null) {
page.pageviews++;
db.analytics.save(page);
}
else {
db.analytics.insert({
"uri" : "/blog",
"pageviews" : 1});
}

...that's 1 round trip + 1 update or insert.

Analytics - Upsert

db.analytics.update(
{ "uri" : "/blog" },
{ "$inc" : { "pageviews" => 1 } },
{ "upsert" : true });

...all in one update!

Capped Collections
4 GB, 50 MB, 97 documents, etc.
Capped Collections
4 GB, 50 MB, 97 documents, etc.
Capped Collections
4 GB, 50 MB, 97 documents, etc.
Storing Files - GridFS
Max: 4 MB
Storing Files - GridFS

(More than 4 MB)

Storing Files - GridFS

J J J

chunks
J J J

J J J
_id : J files
Storing Files - GridFS
$grid = $db->get_gridfs();

$grid->insert($fh,
{"permissions" => 644,
"comment" => "vacation pics"});
Storing Files - GridFS
$file = $grid->find_one($query);

$file->print($another_fh);
JavaScript Functions
> db.eval("function() { return 'hello'; }")
hello
>
JavaScript Functions
> db.eval("function() { return 'hello'; }")
hello
>
> func = "function(x) {" +
... "return 'hello '+ x + '!';" +
... "}"
>
JavaScript Functions
> db.eval("function() { return 'hello'; }")
hello
>
> func = "function(x) {" +
... "return 'hello '+ x + '!';" +
... "}"
> db.eval(func, ["joe"])
hello, joe!
"Stored Procedures"
> db.system.js.insert({
... "_id" : "x",
... "value" : 3});
>
> db.system.js.insert({
... "_id" : "y",
... "value" : 4});
>
"Stored Procedures"
> db.system.js.insert({
... "_id" : "x",
... "value" : 3});
>
> db.system.js.insert({
... "_id" : "y",
... "value" : 4});
>
> db.eval("return x+y");
7
Goals
 Easy-to-use
 Fast
 Always available
 Easy to scale
Query Optimizer

B C

A
Query Optimizer

B C

Done!
Performance
"Any operation that takes longer than 0 milliseconds
is suspect."
- Mongo user on IRC
Performance
100,000 inserts (with indexes for MongoDB and MySQL)

Database Average Insert

MongoDB 0.00011
MySQL 0.00083
CouchDB 0.00640
Memcached 0.00279
Performance

Database Average Query (Indexed)

MongoDB 0.00035
MySQL 0.00772
CouchDB 0.01640
Memcached 0.00015
The Space-Time Continuum

Time

Space
The Space-Time Continuum

Time

Space
Care and Feeding of a Mongod
Server

Laptop
Keeps your database
in memory

Data size
Keeps your database
in memory

Keep indexes
in memory

Data size
Keep the portion of the indexes
Keeps your database you're using in memory
in memory

Keep indexes
in memory

Data size
Keep the portion of the indexes
Keeps your database you're using in memory
in memory
Hits the disk and
takes forever.
Keep indexes
in memory

Data size
Goals
 Easy-to-use
 Fast
 Always available
 Easy to scale
master

slave slave slave

Replica Pairs
master

slave
Replica Pairs

slave
Replica Pairs

master
Replica Pairs
slave

master
Coming soon... replica sets
slave

slave

slave

master
Eventual Consistency
Hey Twitter, I'm eating
a donut.

Fred

Okay
Hey Twitter, I'm eating
a donut.

x 1,000,000

Fred

Fail Whale
Okay

Hey Twitter, I'm eating

a donut.

Fred
Okay

Hey Twitter, I'm eating Nothing new

a donut. from Fred

Fred

Bob
One sec

Hey Twitter, I'm eating

a donut.

Fred
Wait a sec, Bob

Bob
I'd rather have Twitter up and know
what Fred is doing later.

Bob

...eventual consistency works for your customers

I MUST KNOW WHAT FRED IS
DOING RIGHT THIS SECOND!

Bob

...you need "real" consistency

Transactions
A Transaction

Insert this.

Okay, got it.

Phew, my data's safe.

A Transaction
A Transaction
Get Paranoid

Insert this.

Okay, got it.

Phew, my data's safe.

Get Paranoid
Get Paranoid

? I have no idea what

you're talking about.
Get Real Paranoid
Write this to disk

I know better than

he does, I'll just
let this sit in a
buffer for a while.
All over it!
Trust No One!

...trust a bunch of ones. Mostly.

Goals
 Easy-to-use
 Fast
 Always available
 Easy to scale
WARNING!

α
shard
shard
shard

shard
I want Bob's comments
Those are stored here.
Comments from Bob?
Here you go.
Server #3, you're
getting overloaded.

Configuration
Server #3, lockdown.
Split the data,
migrate half to the
new shard.

Router, server #3 has

A-M, server #5 has N-Z.
Shards

Configuration

Routers
Scaling
NoSQL Live
March 11
in Boston

See www.10gen.com/events for details

Thank You!
www.mongodb.org

irc.freenode.net#mongodb
http://groups.google.com/group/mongodb-user

kristina@mongodb.org
@kchodorow