Cosmos DB

Before we start

Survey

https://poznajAzure.pl/survey.html

Who am I?

https://poznajAzure.pl/survey.html

Piotr Stapp

Unique name - just sing a song: "Don't Stapp me know" ;)
CEO @ Dotnetomaniak
VP @ devWarsztaty
Software architect @ Demant Technology Center
M.Sc (distction) Oxford Brooks University in Web Tech
M.Sc. Warsaw University of Technology in Computer Science
And ......

https://www.linkedin.com/in/piotrstapp/

https://poznajAzure.pl/survey.html

Databases in Azure

Options (built-in)

SQL
- Azure SQL
- Azure Database for PostgreSQL
- Azure Database for MySQL
- Azure Database for MariaDB
NoSQL
- Azure Cosmos DB
- Azure Cache for Redis
- Table storage (simple key value)
Self-hosted :)

Where is MongoDB or Cassandra?

Azure Cosmos DB

Survey result

https://admin.sli.do/event/hqjwlbt1/polls

What's inside?

Relational DB

Document DB

Relational vs document

Model

User
Address
Orders

{
    "id": "9f4657a4-43e3-4a47-9392-6c418035d552",
    "key": "7a9f3c4b-3adc-4fb1-a539-47734db30ae0",
    "user": {
        "Id": 215,
        "FirstName": "Wayne",
        "LastName": "Hayes",
        "FullName": "Wayne Hayes",
        "UserName": "Wayne19",
        "UserNameLower": "wayne19",
        "Email": "Wayne.Hayes@hotmail.com",
        "CartId": "a78e749e-ebb2-434c-b658-6795ada0e086",
        "Gender": 0,
        "Orders": [
            {
                "OrderId": 217,
                "Item": "banana",
                "Quantity": 1
            }
        ],
        "Address": {
            "Type": "company",
            "Address": {
                "City": "Florineport",
                "ZipCode": "38056-0892",
                "Street": "Miller Center",
                "CountryCode": "QA"
            }
        }
    },
    "_rid": "vE5vAKy0d4mWzgMAAAAAAA==",
    "_self": "dbs/vE5vAA==/colls/vE5vAKy0d4k=/docs/vE5vAKy0d4mWzgMAAAAAAA==/",
    "_etag": "\"0e0000c6-0000-0d00-0000-6005f1f10000\"",
    "_attachments": "attachments/",
    "_ts": 1611002353
}

Using

Azure CLI

- Creating is easy :)

- 5-15 minutes :(

- I will cheat a bit




az cosmosdb create \
   --name $COSMOS_DB_NAME \
   --resource-group $group \
   --kind MongoDB

Basics

Rules

No JOIN at all (only client side)
Bad design == heavy queries == $$$
Indexes are important 🤪
Relations:
- Embedded (like Orders)
- Reference (like CartId)

DEMO

Let's look on it 😁

DEMO - Simple search

--simple select
SELECT *
FROM c
WHERE c.user.LastName = 'Bernhard'

--simple count
SELECT count(1)
FROM c
WHERE c.user.LastName = 'Bernhard'

DEMO - Nested search

SELECT c.id, c.user.Id, 
	c.user.FirstName, c.user.LastName,
	o.Item, o.Quantity
FROM c
JOIN o IN c.user.Orders
WHERE 1=1
AND c.user.LastName = 'Bernhard'
AND o.Item='kiwi'
and o.Quantity=1

Request Units (RU)

Abstraction over:
- CPU
- Memory
- IOPS
Read 1 KB item costs 1RU

Request Units (RU)

Bullet One
Bullet Two
Bullet Three

https://docs.microsoft.com/en-us/azure/cosmos-db/request-units

DEMO - Calculator

https://cosmos.azure.com/capacitycalculator/

DEMO - Bad one

🙉🙊🙈

SELECT count(1)
FROM c
JOIN o IN c.user.Orders
WHERE 1=1
AND o.Item>'kiwi'
and o.Quantity=1

High Availability

Azure Cosmos DB

Theory

PACELC theorem is an extension to the CAP theorem. It states that in case of network partitioning (P) in a distributed computer system, one has to choose between availability (A) and consistency (C) (as per the CAP theorem), but else (E), even when the system is running normally in the absence of partitions, one has to choose between latency (L) and consistency (C).

(wikipedia)

Azure Cosmos DB

DEMO SLA DOCUMENTATION

(sorry)

https://docs.microsoft.com/en-us/azure/cosmos-db/high-availability

Consistency levels

https://twitter.com/markmccaughrean/status/1101791016737300480

Consistency levels

Strong
Bounded Staleness
Session
Consistent Prefix
Eventual

Strong

Bullet One
Bullet Two
Bullet Three

Bounded Staleness

Bullet One
Bullet Two
Bullet Three

Session

Bullet One
Bullet Two
Bullet Three

Consistent Prefix

Bullet One
Bullet Two
Bullet Three

Eventual

Bullet One
Bullet Two
Bullet Three

Indexes

Types

kind - range (default) or hash
precision - default is -1==maximum
dataType - string or number

Indexes

{
    "indexingMode": "consistent",
    "automatic": true,
    "includedPaths": [
        {
            "path": "/*"
        }
    ],
    "excludedPaths": [
        {
            "path": "/user/Address/*"
        },
        {
            "path": "/\"_etag\"/?"
        }
    ],
    "compositeIndexes": [
        [
            {
                "path": "/user/LastName",
                "order": "ascending"
            },
            {
                "path": "/user/FirstName",
                "order": "descending"
            }
        ]
    ]
}

DEMO - Indexes

SELECT count(1)
FROM c
JOIN o IN c.user.Orders
where 1=1
AND c.user.LastName > 'Z'
AND c.user.FirstName < 'L'
AND CONTAINS(c.user.Email, 'gmail', true)

DEMO - Indexes

-- run on data & db2
SELECT count(1)
FROM c
where c.user.Address.Address.CountryCode="CZ"

Composite indexes

--only on db2
SELECT * FROM c
ORDER BY c.user.LastName desc,
         c.user.FirstName asc

Backup

Managed backup

Default:
- every 4 hours
- only last 2
Not default -> let's check 😁

How is it working?

Bullet One
Bullet Two
Bullet Three

Restore

Bullet One
Bullet Two
Bullet Three

Pricing tips&tricks

Title Text

Bullet One
Bullet Two
Bullet Three

3 options

Manual scale
Autoscale (in theory 1.5x manual)
Serverless (new option)
Free tier

Serverless

100k RU == 0.0258 EUR
Availability Zone 1.25x
Multiregion - not supported yet

Manual / provisioned

100 RU == ~4.92 EUR
Minimum 400 RU per container in container scaling mode
In database mode formula is more complicated: MAX([containers count]x100RU, 400 RU)

Autoscale

100 RU autoscale == 150 RU manual
Autoscale is always a range: from 10% X to 100% X
Database model pricing is "better"
Minimum for 25 containers is still 400 RU, instead of 2500 RU

Free tier

With free tier, you'll get the first 400 RU/s and 5 GB of storage in this account for free. To keep your account free, keep the total RU/s across all resources in the account to 400 RU/s.

ORM - problems

DEMO - Translation

var query = _dbClient.Where<UserDocument>( _collectionUri,
 d =>d.User.UserName.ToLower() == criteria.SearchString);

SELECT * FROM c
WHERE LOWER(c.user.UserName) = "{search}" 

------------------------------

var query = _dbClient.Where<UserDocument>( _collectionUri,
 d => d.User.UserName.Equals(originalSearchString,
 		StringComparison.OrdinalIgnoreCase)); 

SELECT * FROM c
WHERE STRINGEQUALS(c.user.UserName, "{search}", true)

----

var query = _dbClient.Where<UserDocument>( _collectionUri,
 d => d.User.UserNameLower == originalSearchString); 
 
SELECT * FROM c
WHERE c.user.UserNameLower = "{search}"

More stappuff

Things

Different languages and types than SQL & documents:
- Mongo, Cassandra, Gremlin, etcd ...
Geospatial data + search
Storing & querying dates
Change feed
...

The end?

Not yet :)

Q&A

https://www.linkedin.com/in/piotrstapp/

Twitter: @ptrstpp950