Quick trip around the Cosmos - Things every astronaut supposed to know

Quick trip around the Cosmos
Things every astronaut supposed to
know

Rafał Hryniewski
@r_hryniewskifb.me/hryniewskinet

Agenda
Cosmos DB origin and (very brief) story
Why Cosmos DB
Available data models
Consistency levels
Scaling, Pricing and Georeplication
Service Level Agreements
Getting started (for free)

Glossary - ACID
A – Atomicity
C – Consistency
I – Isolation
D - Durability

Glossary – CAP Theorem
C – Consistency
A – Availability
P – Partition Tolerant

In the beginning there was only chaos...

2010 – Project Florence goals
Scale elastically
Low read and write latency
At least 99,99% availability
Intuitive and predictable concurrency
Comprehensive SLA
No schema/index management
Multiple models
Low costs
https://azure.microsoft.com/en-us/blog/a-technical-overview-of-azure-cosmos-db/

What already lives under the clouds?

Do we need another database?
https://db-engines.com/en/ranking

Why don’t we stick with MS SQL?

PaaS databases on Azure
Azure SQL
Redis
PostgreSQL
MySQL
Cosmos DB
http://hryniewski.net/2017/05/28/paas-databases-available-on-azure/

Any other database that can be installed on VM

When I don’t love SQL anymore?

So... Should we all abandon earth?

Industry’s first globally distributed, multi-model
database service

Database tools we all love
Stored Procedures
Triggers
User Defined Functions (can be used in SQL syntax)
All in JavaScript

Four models, Four APIs
Document Database with Document DB SQL(like) API
Document Database with Mongo DB API
Key-Value Database with Azure Table Storage API
Graph database with (Tinkerpop) Gremlin API

Document DB – data format sample
[
{
"id": "59be9f01b7cb3b8c0fffd52b",
"firstName": "Camacho",
"lastName": "Castro",
"birthday": "2016-10-14T03:53:19 -02:00",
"project": "in",
"_rid": "DJdRAIh6aAAGAAAAAAAAAA==",
"_self":
"dbs/DJdRAA==/colls/DJdRAIh6aAA=/docs/DJdRAIh6aAAGAAAAAAAAAA==/",
"_etag": ""4200c424-0000-0000-0000-59bea13d0000"",
"_attachments": "attachments/",
"_ts": 1505665341
},
...
]

Document DB – query sample
SELECT * FROM Collection c
SELECT * FROM Collection c WHERE c.gender = 'female’
SELECT c["firstName"] AS FirstName, c["lastName"] AS
LastName FROM Collection c
SELECT {"FirstName" : c.firstName, "LastName": c.lastName}
AS PersonalData FROM Collection c

Table Storage – data format sample

Table Storage - query sample (portal)

Table Storage - query sample (SDK)

Gremlin – data format sample (GraphSON)
{"id": "33843a67-0da8-4dde-81ff-eff607882f23",
"label": "59be9f016f12f58cd5eb0cac",
"type": "vertex",
"properties": {
"firstName": [
{
"id": "742b60bd-cd75-4ec1-b288-c694d88c14ed",
"value": "Santana"
}
],
"lastName": [
{
"id": "70ed7300-cc4d-41c4-8af9-ba65445898ee",
"value": "Peters"
}
],
"gender": [
{
"id": "78b124d0-73af-4ff7-8f93-881377086568",
"value": "male"
}
],
"birthday": [
{
"id": "9140a304-a073-4c55-a86f-f3206d4ecbd9",
"value": "2016-08-15T10:10:17 -02:00"
}
],
"project": [
{
"id": "6cc7ffd1-6427-49a0-85c2-e1f5b1bbcd42",
"value": "elit"
}
]}

Gremlin – query sample
g.V()
g.V().has('gender', 'female')
g.V().valueMap('firstName', 'lastName')

Gremlin – everything can relate with everything

Gremlin – adding edges 101
g.V().has('project', 'magna’) //WHO
.addE('InLoveIn’) //RELATES HOW
.to(V().has('gender', 'female’)) //TO WHOM

Gremlin – traversing graphs 101
g.V().has('firstName', 'Ivy').in('InLoveIn').values('birthday’)
g.V().has('firstName', 'Davenport').out('InLoveIn').values('project')

Mongo DB – data format sample
[
{
"_id" : ObjectId("59beaaa47d85940950b876fd"),
"id" : "59be9f016f12f58cd5eb0cac",
"firstName" : "Santana",
"lastName" : "Peters",
"gender" : "male",
"birthday" : "2016-08-15T10:10:17 -02:00",
"project" : "elit"
},
...
]

Mongo DB – query sample
db.mycollection.find()
db.mycollection.find({gender : "female"})
db.mycollection.find({}, {firstName: 1, lastName: 1, _id:0})

Available SDKs
https://docs.microsoft.com/en-us/azure/cosmos-db/

Non-Azure Equivalents
Document DB - Mongo DB
Mongo DB – Mongo DB (Captain obvious to the rescue!)
Azure Table Storage – Cassandra, HBase
Graph API - Neo4J, Titan

Strong consistency & eventual consistency
Strong Consistency – always reading current data
Eventual Consistency – data will be consistent... eventually

There are other consistency levels? Like what?
Somewhat consistent?

5 of them in Cosmos DB
https://docs.microsoft.com/en-us/azure/cosmos-db/consistency-levels

Strong consistency
+ B
A + C
A
ABA ABC
ABC

Strong consistency
You can use only one Azure region with strong consistency

Bounded-staleness
+ B
A + C
A
AA ABC
ABAB

Bounded-staleness
Introduces acceptable lag for time or number of item versions
Data will manage consistent ordering except while in
staleness window
Georeplication to other Azure region is available if staleness
window is more than 100 000 operations or 5 minutes

Session
+ B
A + C
AB
AA AC
AB
ACB

Session
Scoped to a client session
Reading own writes in consistent order in own session
Better throughput and latency than strong and bounded
staleness consistency levels while managing great session-
scoped consistency

Consistent prefix
+ B
A + C
AB
ABCA ABC
ABCD
AB
+ D

Consistent prefix
Data may not be consistent, but will always be readed in
updates order

Eventual consistency
+ B
A + C
AD
ACA AC
ABD
AB
+ D
ABCD

Eventual consistency
Cheapest writes
Best in terms of latency and throughput
You can read data older than you’ve seen just a second ago
But data will be consinstent...eventually

Consistency can be changed per request
Just add x-ms-consistency-level header to your request
Or look for this option in your SDK

Request Units
Every request costs certain amount of Request Units
It depends on consistency and request itself
You’re paying for provisioning n RU per second
Request Unit is throughput metric

How much is request unit
~1kb read – 1RU
~1kb query by id – 2,5 RU
~1kb create – 15RU

~1kb worth of JSON data
{"id": "59b413f6112b5af91a117944",
"guid": "cf9aa26f-f863-4adf-b6bd-
f4419427219c",
"isActive": false,
"balance": "$1,053.74",
"picture": "http://placehold.it/32x32.jpg",
"age": 32,
"eyeColor": "blue",
"name": "Lynette Campos",
"gender": "female",
"company": "XURBAN",
"email": "lynettecampos@xurban.com",
"secondaryEmail":"lynettecampos@seconda
ry.com"
"phone": "+1 (826) 428-2753",
"address": "511 Rodney Street, Libertytown,
Guam, 5768",
"about": "Sit sunt est Lorem dolore id
magna. Irure non proident culpa dolor enim.
Ex veniam laborum consectetur pariatur
mollit elit non commodo incididunt Lorem
labore. Qui labore ut excepteur id laboris
adipisicing ullamco et nulla irure nostrud
exercitation adipisicing excepteur.
Commodo duis incididunt ea in anim veniam
eiusmod deserunt. In incididunt duis
laborum in.rn",
"registered": "2015-12-15T08:48:18 -01:00",
"tags":
["lorem","ipsum","dolor","nisi","magna","cul
pa","pariatur","tempor"],
"favoriteFruit": "banana",
"favouriteProgrammingLanguage":"C#"}

How much does it cost?
100 RU/s – ~5,02 EUR monthly
1 GB of storage (SSD) - ~0.21 EUR monthly
Minimum 400RU/s (~20.08 EUR monthly)
Each replica multiples the amount!

What if I use all of my Request Units?

What if I use all of my Request Units
HTTP Status 429
Status Line: RequestRateTooLarge
x-ms-retry-after-ms :100

What’s the limit
Minimum request units that can be provisioned is 400RU/s
After 2500 RU/s – partition key is required
After 10000 RU/s – you need to contact support

What partition key should we choose?
{
"id": 9,
"firstName": "Rafał",
"lastName": "Hryniewski",
"birthday": "16-05-1988",
"project": "Hello world!"
}

Let’s go global
~725 documents
West US~0,75-1,10s
West Europe 0,20-0,30s

SLA
https://azure.microsoft.com/en-us/support/legal/sla/cosmos-db/v1_0/

SLA
SLA – Service Level Agreement
Service Credit - Service Fee * Service Credit Percent

Availability SLA
SLA for failed requests percent
99,99% Availability or 10% Service Credit
99% Availability or 25% Service Credit

Throughput SLA
You will be available to use RU/s that you’ve provisioned
99,99% Throughput or 10% Service Credit
99% Throughput or 25% Service Credit

Consistency SLA
Your reads and writes will be executed within chosen
consistency level
99,99% Consistency Attainment or 10% Service Credit
99% Consistency Attainment or 25% Service Credit

Latency SLA
Guarantees 10ms latency for reading up to 1kb document in
same region
Guarantees 15ms latency for writing up to 1kb document in
same region
99,99% Latency Attainment or 10% Service Credit
99% Latency Attainment or 25% Service Credit

Try for free
https://azure.microsoft.com/en-us/try/cosmosdb/
Free Cosmos DB for up to 48h

Dev essentials
https://www.visualstudio.com/dev-essentials/
25 EUR monthly for Azure

Cosmos DB Emulator
https://docs.microsoft.com/en-us/azure/cosmos-db/local-
emulator
Free tool for offline cosmos development

It’s so awesome! Isn’t it?

Links
CosmosDB Technical Overview
• https://azure.microsoft.com/en-us/blog/a-technical-overview-of-azure-
cosmos-db/
Database popularity ranking
• https://db-engines.com/en/ranking
PaaS Database on Azure
• http://hryniewski.net/2017/05/28/paas-databases-available-on-azure/
CosmosDB documentation
• https://docs.microsoft.com/en-us/azure/cosmos-db/
CosmosDB SLAs
• https://azure.microsoft.com/en-us/support/legal/sla/cosmos-db/v1_0/

Thank you!
@r_hryniewskifb.me/hryniewskinet

Quick trip around the Cosmos - Things every astronaut supposed to know

More Related Content

Quick trip around the Cosmos - Things every astronaut supposed to know