Introduction to Azure DocumentDB

Alexander Zyl
.NET Developer
azyl@scnsoft.com

NoSQL solutions in Azure
HBase
TS
• Column Family Store • Key/Value Store
Redis
TS

Features:
• Fully managed
• Schema agnostic
• Scalable
• Tunable consistency levels
• Tunable indexing policies
• Familiar SQL syntax for querying
• JavaScript execution
DocumentDB

DocumentDB resource model
RESTAPI
DocumentDB Infrastructure

Databases
/dbs/{id}
DocumentDB
Account

Collections
/colls/{id}
Databases
/dbs/{id}
Users
/users/{id}
Permissions
/permissions/{id}

Triggers
/triggers/{id}
Functions
/functions/{id}
Stored Procedures
/sprocs/{id}
Attachments
/attachments/{id}
Documents
/docs/{id}
Collections
/colls/{id}

DocumentDB
Account
Users
/users/{id}
Databases
/dbs/{id}
Permissions
/permissions/{id}
Collections
/colls/{id}
Attachments
/attachments/{id}
Documents
/docs/{id}
Triggers
/triggers/{id}
Functions
/functions/{id}
Stored Procedures
/sprocs/{id}

Server ZServer CServer BServer A
Logical containers
Physical containers
DocumentDB
Account
CollectionsCollections Collections Collections

= 𝑓 𝑀𝑒𝑚𝑜𝑟𝑦, 𝐶𝑃𝑈, 𝐼𝑂
Request Unit

Performance levels
RUs per second RUs per second20 20 20RUs per second RUs per second250 RUs per second 1k RUs per second 2.5k RUs per second

{
"Id": 44,
"ReleaseYear": "2014",
"Make": "Aston Martin",
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
"Address": "Some st. 9"
},
"GPSLocation": {
"Latitude": 44.6516185,
"Longitude": -63.5820275
},
"StatusData": […]
}
Vehicle Dealer
StatusData
VehicleIdPK PK
PK
ReleaseYear
Make
DealerId
Name
StatusId
VehicleId
EngineOn
TimeStamp
Address
Vin
Model
Latitude
Longitude

Approaches to document modeling
 Reference data
 Embed data

Vehicle document:
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
"GPSLocation": {
"Latitude": 44.6516185,
"Longitude": -63.5820275
},
"StatusData": [
{ "Id": 1, "TimeStamp": "2014-07-04", "EngineOn": true, "FuelLevel": 40 },
{ "Id": 2, "TimeStamp": "2014-07-04", "EngineOn": false, "FuelLevel": 33 },
{ "Id": 999, "TimeStamp": "2014-08-12", "EngineOn": true, "FuelLevel": 23 }
]
}
Bad design

When to embed:
 One-to-few relations
 Infrequent changes
 Embedded data has
bounds
Modeling relations

Vehicle document:
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
"GPSLocation": {
"Latitude": 44.6516185,
"Longitude": -63.5820275
},
"StatusData": [
]
}

When to embed:
 One-to-few relations
 Infrequent changes
 Embedded data has
bounds
When to reference:
 One-to-many relations
 Many-to-many relations
 Data changes frequently
 Unbounded reference
Modeling relations

Vehicle document:
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
"GPSLocation": {
"Latitude": 44.6516185,
"Longitude": -63.5820275
},
"StatusData": [
]
}
Vehicle document:
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
"StatusData": [
]
}
VehicleLocation document:
{
"VehicleId": 44,
"GPSLocation": {
"Latitude": 44.651617,
"Longitude": -63.582027
}
}

Vehicle
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
"StatusData"95,: [
]
}
Vehicle document:
{
"Id": 44,
"Vin": "2G1WT58KX79250102",
"Model": "DBS",
"Dealer": {
"Name": "Atlant-M",
},
}
VehicleStatus documents:
{ "Id": 1, "TimeStamp": "2014-07-04", "EngineOn": true, "FuelLevel": 40, "VehicleId": 44 },
{ "Id": 2, "TimeStamp": "2014-07-04", "EngineOn": false, "FuelLevel": 33, "VehicleId": 44 },
{ "Id": 5, "TimeStamp": "2014-07-06", "EngineOn": false, "FuelLevel": 55, "VehicleId": 44 },
{ "Id": 999, "TimeStamp": "2014-08-12", "EngineOn": true, "FuelLevel": 23, "VehicleId": 44 }

Documents in a collection
{
"id": 16,
"text": "Bonjour",
"user": {
"name": "Francois",
"nickname": "@franky"
},
"entities": {
"hashtags": [
{
"text": "#heof"
}
]
}
}
{
"id": 4,
"text": "Hello",
"user": {
"name": "Jerome",
"nickname": "@juim"
},
"entities": {
"hashtags": [
{
"text": "#rutib",
"indices": [ 10, 26 ]
}
]
}
}

Index tree
textid userentities
4 16 Hello Bonjourname nickname
Jerome Francois @juim @franky
hashtags
text indices
0
10 26
0
#rutib #heof

What we can controlAutomatic indexing
 Manual include
 Manual exclude

What we can control
Automatic indexing
 Manual include
 Manual exclude

What we can control
Indexing modes
 Consistent
 Lazy
 None

Building paths
{
"id": 16,
"text": "Bonjour",
"user": {
"name": "Francois",
"nickname": "@franky"
},
"entities": {
"hashtags": [
{ "text": "#heof" }
]
}
}
_.text
_.user.name
_.entities.hashtags
_.entities.hashtags[0].text

Building paths
Applicable wildcards:
? – single selection
/text/?
/user/nickname/?
/entities/hashtags/[]/text/?

{
"id": 4,
"text": "Hello",
"user": {
"name": "Jerome",
"nickname": "@juim"
},
"entities": {
"hashtags": [
{
"text": "#rutib",
"indices": [ 10, 26 ]
}
]
}
}
/text/?
Building paths: examples

{
"id": 4,
"text": "Hello",
"user": {
"name": "Jerome",
"nickname": "@juim"
},
"entities": {
"hashtags": [
{
"text": "#rutib",
"indices": [ 10, 26 ]
}
]
}
}
/user/nickname/?

Building paths
Applicable wildcards:
? – single selection
* – recursive selection/text/?
/user/nickname/?
/entities/hashtags/[]/text/?
/user/*
/entities/*
/entities/hashtags/[]/*

{
"id": 4,
"text": "Hello",
"user": {
"name": "Jerome",
"nickname": "@juim"
},
"entities": {
"hashtags": [
{
"text": "#rutib",
"indices": [ 10, 26 ]
}
]
}
}
/user/*

{
"id": 4,
"text": "Hello",
"user": {
"name": "Jerome",
"nickname": "@juim"
},
"entities": {
"hashtags": [
{
"text": "#rutib",
"indices": [ 10, 26 ]
}
]
}
}
/entities/hashtags/*

Indexing options:
Include to index
Exclude from index
Applying rules

Index kinds:
Hash – equality queries
Applying rules
SELECT *
FROM collection c
WHERE c.prop = 'value'

Index kinds:
Range – range + OrderBy queries
Applying rules
SELECT *
FROM collection c
WHERE c.prop = 'value'
SELECT *
FROM collection c
WHERE c.prop >= 15
ORDER BY c.prop

Index kinds:
Range – range + OrderBy queries
Spatial – ST_DISTANCE, ST_WITHIN
Applying rules
SELECT *
FROM collection c
WHERE c.prop >= 15
ORDER BY c.prop
SELECT *
FROM collection c
WHERE ST_DISTANCE(c.Location, {
"type": "Point",
"coordinates": [-122.19, 47.36]
}) < 100 * 1000

Index precision:
For numbers: 1-8 bytes
For strings: 1-100 bytes
Applying rules

•Strong consistency
•Eventual consistency
DocumentDB
•Strong consistency
•Eventual consistency
•Session
•Bounded staleness
Offered consistency models

Strong consistency: Write operation
Version2
Version1
Version2
User
Replica A
Replica B
Replica CGateway
Async
Infra
Version2

Strong consistency: Read operation
Version2
Version1
Version2
User
Replica A
Replica B
Replica CGateway
Infra
Version2

Eventual consistency: Write operation
Version2
Version1
Version1
User
Replica A
Replica B
Replica CGateway
Async
Infra
Version2

Eventual consistency: Read operation
Version2
Version1
Version1
User
Replica A
Replica B
Replica CGateway
Infra
Version1

Session: Write operation
Version2
Version1
Version1
User
Replica A
Replica B
Replica CGateway
Async
InfraSession Id
Version2

Session: Read operation
Version2
Version1
Version1
Replica A
Replica B
Replica CGateway
Infra
Version2
User A
User B
Session Id
Version1

Bounded staleness: Write operation
Version2
Version1
Version1
User
Replica A
Replica B
Replica CGateway
Async
Infra
Version2

Bounded staleness: Read operation
Version3
Version1
Version2
User
Replica A
Replica B
Replica CGateway
Sync
Infra
Version2

Scalability issues
Out of space
>Data Collection

Scalability issues
Too many requests
Out of space
>Data Collection

Horizontal scaling
Vertical scaling
Solutions?

Collection
RequestPartitioning our data

Partition 1
Request
Request
Partition 2
Logical grouping
Partitioning our data

Partitioning strategies
Hash partitioning
Partition 1
Partition 3
Partition 2
UserId: 14 Hash(14)=>
P3
Infra
Logical grouping

Range partitioning
Partition 1
Partition 3
Partition 2
Name: David
Infra
K > David >= A
A-I
K-Q
R-Z
Logical grouping

Lookup partitioning
Region name Partition Id
Asia Partition1
Version3Europe Partition2
United States Partition3
Partition 1
Partition 3
Partition 2
Region: Europe
Infra
Logical grouping

Use Cases: storing and analyzing logs
Logs

Use Cases: storing materialized views

Introduction to Azure DocumentDB

Recommended

More Related Content

What's hot (19)

Viewers also liked (12)

Similar to Introduction to Azure DocumentDB (20)

Recently uploaded (20)

Introduction to Azure DocumentDB

Editor's Notes