Presentation given to a room full of web designers and developers at DevCon5 in NYC. The presentation introduces concepts critical to understanding how databases work and why it's important to understand even when one typically works much higher in the stack. Topics covered: * History of web development * Picking the right data model * Modern database interaction * How modern databases scale An emphasis is placed on MongoDB in this presentation, but other technologies such as relational (MySQL, Oracle, MSSQL) and Key Value (Cassandra, Riak, DynamoDB) are also covered.

1. Building Modn
Web Acaons

2. @spf13Steve Francia
AKA
Chief Evangelist @
responsible for Developer Experience:
drivers, integrations, web & technical writing

3. ๏ History of web development
๏ Picking the right data model
๏ Modern DB interaction
๏ Modern Web Scale
Agda

4. What is a modn Web A?

5. What is a modn Web A?

6. If you don't ow whe
you've come from,
you don't ow whe
you are. - James Burke

7. te 90’s
๏Web is born
๏Web development mostly
done in perl or C
๏Everyone is a webmaster
๏Relational databases

8. r ’s
๏ Web growth
redefines scale
๏ Javascript avoided
๏ Dynamic languages come of age
๏ LAMP
๏ Everyone is a PHP programmer
๏ Relational databases

9. Mid ’s
๏ Social re-
redefines scale
๏ Multimedia rules
๏ Heavy caching (memcache) required LAM(m)P
๏ Frameworks (Ruby on Rails) with heavy
database abstractions en vogue
๏ Everyone is a OO programmer
๏ Relational databases*

10. is is whe  all
falls apa

11. Condons
๏ Web users exponentially increasing
๏ Excessive layering causes applications
to be slower
๏ Social (dynamic data) limits use of
caching crutch
๏ Cost per byte decreasing rapidly
๏ Data growing in size & complexity

12. Symptoms
๏ Over abstraction
๏ Agile development unsustainable
๏ Needlessly complex
architectures
๏ Memcache

13. 2010 trds
(Areßing r Ißues)
๏Horizontal scale
๏Variety of Choices
(LAMP no more)
๏Specializing

14. Raonal designed for one ing,
used for hing

15. What is a modn Web A?

16. What is a modn Web A?

17. What do we lk for in
a database?
๏ Right structure to match my
data
๏ Performance & Scale
๏ Features that enable me as a
developer

18. K Queson:
WHAT IS A RECORD?

19. K Value
๏ One-dimensional storage
๏ Single value is a blob
๏ Query on key only
๏ Some support secondary indexes
๏ No schema
๏ Value cannot be updated, only replaced
Key Blob
Cassandra, Redis, MemcacheD, Riak, DynamoDB

20. Raonal๏ Query on any field
๏ In-place updates
๏ Two-dimensional storage
๏ Each field contains a single value
๏ Very structured schema (table)
๏ Normalization process requires many tables,
joins, indexes, and poor data locality
Primary
Key
Oracle, MSSQL, MySQL, PostgreSQL, DB2

21. Documt๏ N-dimensional storage
๏ Each field can contain 0, 1,
many, or embedded values
๏ Query on any field & level
๏ Flexible schema
๏ Inline updates
๏ Embedding related data has optimal data locality,
requires fewer indexes, has better performance
_id
MongoDB, CouchDB, RethinkDB

22. Raonal design

23. Documt design

24. Example Blog Post doc
{ _id : ObjectId("4c4ba5c0672c685e5e8aabf3"),
author : "steve",
date : "Sat Apr 24 2013 19:47:11",
text : "About MongoDB...",
tags : [ "tech", "databases" ],
comments : [
{
author : "Fred",
date : "Sat Apr 25 2013 20:51:03 GMT-0700",
text : "Best Post Ever!"
}
]
}

25. What is a modn Web A?

26. What is a modn Web A?

27. MongoDB spks your ngauage
๏ Drivers in 14+ languages
๏ Interface is natural and
idiomatic for each language
๏ Document natively maps to
map/hash/object
array/dict/struct

28. place1 = {
name : "10gen HQ",
address : "229 W 43rd St. 5th Floor",
city : "New York",
zip : "10036",
tags : [ "business", "awesome" ]
}
Start with an object
(or array, hash, dict, etc)

29. Inserting the record
Initial Data Load
> db.places.insert(place1)
> db.places.insert(place1)

30. Querying
> db.places.findOne({ zip: "10036",
tags: "awesome" })
> db.places.find({tags: [ "rad", "awesome" ]})
{
name : "10gen HQ",
address : "229 W 43rd St. 5th Floor",
city : "New York",
zip : "10036",
tags : [ "business", "awesome" ]
}

31. Updating
> db.places.update(
{name : "10gen HQ"},
{ $push :
{ comments :
{ author : "steve",
date : 6/26/2013,
text : "Office hours are great!"
}
}
}
)

32. Nested documents
// Index nested documents
> db.places.ensureIndex({ "comments.author":1 }) // optional
> db.places.find({'comments.author':'Fred'})
{ _id : ObjectId("4c4ba5c0672c685e5e8aabf3"),
name : "10gen HQ",
address : "229 W 43rd St. 5th Floor",
city : "New York",
zip : "10036",
comments : [ {
author : "Fred",
date : "Sat Apr 25 2013 20:51:03",
text : "Best Place Ever!"
} ]
}

33. Multiple values
// Index on tags (multi-key index)
> db.places.ensureIndex({ tags: 1}) // optional
> db.places.find( { tags: 'tech' } )
{
_id : ObjectId("4c4ba5c0672c685e5e8aabf3"),
name : "10gen HQ",
address : "229 W 43rd St. 5th Floor",
city : "New York",
zip : "10036",
tags : [ "business", "awesome", "tech" ],
}

34. Paginating Places in JS
per_page = 10;
page_num = 3;
places = db.places
.find({ "city" : "new york" })
.sort({ "ts" : -1 })
.skip((page_num - 1) * per_page)
.limit(per_page);

35. Paginating Places in Ruby
@per_page = 10
@page_num = 3
@places = @db.places
.find({ :city => "new york" })
.sort({ :ts => -1 })
.skip(( @page_num - 1 ) * @per_page)
.limit(@per_page)

36. ch ftures๏ Rich query language
๏ GeoSpatial
๏ Text search
๏ Flexible schema
๏ Aggregation & MapReduce
๏ GridFS
(distributed & replicated file storage)
๏ Integration with Hadoop, Storm, Solr & more

37. Database ndscape
Scalability&Performance
Depth of Functionality
MongoDB
Key Value
RDBMS

38. NoSQL popu

39. NoSQL popu

40. NoSQL popu

41. What is a modn Web A?

42. What is a modn Web A?

43. Scabi Needs
๏ Data is highly available
๏ Data is consistent
๏ Performant
(caching unnecessary)

44. Difft Aroaches
๏
MultiMaster
๏
Peer to peer
๏
Has Conflicts
๏
Ring based
approach
combines high
availability and
distribution
๏
Complex
application logic
๏
Single Master
๏
Consistent
๏
Slaves have
delayed writes
๏
High availability
๏
No scalable
solution
๏
Single Master
๏
Consistent
๏
Secondaries have
delayed writes
๏
High availability
๏
Range based
distribution

45. MongoDB : bui to scale
๏ Intelligent replication
๏ Automatic partitioning of data
(user configurable)
๏ Horizontal Scale
๏ Targeted Queries
๏ Parallel Processing

46. Igt Repcaon
Node 1
Secondary
Node 2
Secondary
Node 3
Primary
Replication
Heartbeat
Replication

47. Scable Archecture
Node 1
Secondary
Config
Server
Node 1
Secondary
Config
Server
Node 1
Secondary
Config
Server
Shard Shard Shard
Mongos
App Server
Mongos
App Server
Mongos
App Server

48. High Avaibi in Shards
Shard
Primary
Secondary
Secondary
Shard
orMongod
x

49. Targed Requests
Shard Shard Shard
Mongos
1
2
3
4

50. Pall proceßing
Shard Shard Shard
Mongos
1
2 2 2
4 44
3 3 3
6
5

51. What is a modn Web A?

52. e g database
to t your data

53. e g database
for YOUR dopmt

54. e g Database
for scale & Pfoance

55. Gng staed

56. sy deploymt
๏ Heroku
๏ Rackspace
๏ Amazon
๏ Engine Yard
๏ App Fog
๏ ServerGrove
๏ Azure
๏ Nodejitsu

57. Indust acaon
๏ Media &
Entertainment
๏ Retail
๏ Social
๏ Finance
๏ Gaming
๏ Insurance
๏ Healthcare
๏ Government
๏ Archiving
๏ Telecom
๏ Education

58. E IF YOU KED !
Questions?
http://spf13.com
http://github.com/spf13
@spf13
#DevCon5