2. WARNING – This session is rated as a ‘Grandma session’ (=Level 200)
3. 3 goals of this presentations
When you leave this presentation you should have learned
How easy it is to get started using MongoDB
How using MongoDB changes the way you design and build your applications
How MongoDB’s flexibility supports evolutionary design
That giving speakers beer before a session is never a good idea
14. MongoDB
HuMongous
General purpose database
Document oriented database using JSON document syntax
Features:
- Flexibility
- Power
- Scaling
- Ease of Use
- Built-in Javascript
Users: Craigslist, eBay, Foursquare, SourceForge, and The New York Times.
15. Written in C++
Extensive use of memory-mapped files
i.e. read-through write-through memory caching.
Runs nearly everywhere
Data serialized as BSON (fast parsing)
Full support for primary & secondary indexes
Document model = less work
High Performance
17. MongoDB Database Architecture: Collection
Logical group of documents
May or may not share same keys
Schema is dynamic/application maintained
18. Why should I use it?(or how do I convince my boss?)
Developer productivity
Avoid ORM pain, no mapping needed
Performance(again)
Scaling out is easy(or at least easier)
Optimized for reads
Flexibility
Dynamic schema
19. How to run it?
Exe
Windows service
Azure
3rd party commercial hosting
20. How to talk to it?
Mongo shell Official and non official drivers
>12 languages supported
23. 23
First step in any application
is determine your
domain/entities
24. In a relational based app
We would start by doing
schema design
25. In a MongoDB based app
We start building our app
and let the schema evolve
26. Comparison
Album
- id
- artistid
- title
Track
- no
- name
- unitPrice
- popularity
Artist
- id
- name
Album
- _id
- title
- artist
- tracks[]
- _id
- name
Relational Document db
28. Modeling
Start from application-specific queries
“What questions do I have?” vs “What answers”
“Data like the application wants it”
Base parent documents on
The most common usage
What do I want returned?
30. Product
Single collection inheritance
Product
- _id
- price
Book
- author
- title
Album
- artist
- title
Jeans
- size
- color
- _id
- price
- author
- title
Relational Document db
- _id
- price
- size
- color
31. Product
Single collection inheritance
Product
- _id
- price
Book
- author
- title
Album
- artist
- title
Jeans
- size
- color
_type: Book
- _id
- price
- author
- title
Relational Document db
_type: Jeans
- _id
- price
- size
- color
32. One-to-many
Embedded array / array keys
Some queries get harder
You can index arrays!
Normalized approach
More flexibility
A lot less performance
BlogPost
- _id
- content
- tags: {“foo”, “bar”}
- comments: {“id1”, “id2”}
36. ACID Transactions
No support for multi-document
transactions commit/rollback
Atomic operations on document level
Multiple actions inside the same
document
Incl. embedded documents
By keeping transaction support
extremely simple, MongoDB can
provide greater performance
especially for partitioned or replicated systems
39. Storing binary documents
Although MongoDB is a document database, it’s not good for documents :-S
Document != .PNG & .PDF files
Document size is limited
Max document size is 16MB
Recommended document size <250KB
Solution is GridFS
Mechanism for storing large binary files in MongoDB
Stores metadata in a single document inside the fs.files collection
Splits files into chunks and stores them inside the fs.chunks collection
GridFS implementation is handled completely by the client driver
40. Demo 4 – Evolving your domain model
------------& GRIDFS
41. Evolving your domain model
Great for small changes!
Hot swapping
Minimal impact on your application and database
Avoid Migrations
Handle changes in your application instead of your database
43. Avoid table collections scans by using indexes
> db.albums.ensureIndex({title: 1})
Compound indexes
Index on multiple fields
> db.albums.ensureIndex({title: 1, year: 1})
Indexes have their price
Every write takes longer
Max 64 indexes on a collection
Try to limit them
Indexes are useful as the number of records you want to return are limited
If you return >30% of a collection, check if a table scan is faster
Creating indexes
44. Aggregations with the Aggregation Framework
$project Select()
$unwind SelectMany()
$match Where()
$group GroupBy()
$sort OrderBy()
$skip Skip()
$limit Take()
Largely replaces the original Map/Reduce
Much faster!
Implemented in a multi-threaded C ++
No support in LINQ-provider yet (but in development)
47. Benefits
Scalable: good for a lot of data & traffic
Horizontal scaling: to more nodes
Good for web-apps
Performance
No joins and constraints
Dev/user friendly
Data is modeled to how the app is going to use it
No conversion between object oriented > relational
No static schema = agile
Evolvable
48. Drawbacks
Forget what you have learned
New way of building and designing your application
Can collect garbage
No data integrity checks
Add a clean-up job
Database model is determined by usage
Requires insight in the usage
51. Things we didn’t talk about…
Security
- HTTPS/SSL
Compile the code yourself
Eventual Consistency
Geospatial features
Realtime Aggregation
52. Things we didn’t talk about…
Many to Many
- Multiple approaches
References on 1 site
References on both sites
53. Things we didn’t talk about…
Write Concerns
- Acknowledged vs Unacknowledged writes
- Stick with acknowledged writes(=default)
54. Things we didn’t talk about…
GridFS disadvantages
- Slower performance: accessing files from
MongoDB will not be as fast as going directly
through the filesystem.
- You can only modify documents by deleting
them and resaving the whole thing.
- Drivers are required
55. Things we didn’t talk about…
Schema Migrations
- Avoid it
- Make your app backwards compatible
- Add version field to your documents
56. Things we didn’t talk about…
Why you should not use regexes
- Slow!
Advanced Indexing
- Indexing objects and Arrays
- Unique vs Sparse Indexes
- Geospatial Indexes
- Full Text Indexes
MapReduce
- Avoid it
- Very slow in MongoDB
- Use Aggregation FW instead
57. Things we didn’t talk about…
Sharding
Based on a shard key (= field)
Commands are sent to the shard that includes
the relevant range of the data
Data is evenly distributed across the shards
Automatic reallocation of data when adding or
removing servers