1. Azure Cosmos DB: From Internet to Enterprise
9/8/2018 7:26 AM
THR2048
Azure Cosmos DB: From Internet to Enterprise
Aleksey Savateyev
Azure Cosmos DB
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2. Azure Cosmos DB Table API MongoDB API
A globally distributed, massively scalable, multi-model database service
Table API
MongoDB API
Key-value
Column-family
Document
Graph
Guaranteed low latency at the 99th percentile
Elastic scale out
of storage & throughput
Five well-defined consistency models
Turnkey global distribution
Comprehensive SLAs

3. Cosmos DB By Industries (Our Customers)
Retail
Manufacturing
Government / Edu
Automotive/IoT
Financial Services
Professional Services

4. Powering global solutions Field-tested by Microsoft’s planet-scale services and industry-leading enterprises apps
Globally-distributed mission-critical apps
IoT
Personalization
Retail and e-commerce
Gaming
Guarantee uptime to users worldwide with high-availability and low-latency
Scale instantly for uncertain IoT workloads without sacrificing performance
Generate personalized service through low-latency and tunable consistency settings
Support queries over product catalogs, traffic spikes, and rapidly changing inventory
Accommodate bursts of traffic and deliver low-latency multiplayer experiences

5. Turnkey Global Distribution
Worldwide presence as a Foundational Azure service
Automatic multi-region replication
Multi-homing APIs
Manual and automatic failovers
Designed for High Availability

6. Elastic and unlimited scalability
Microsoft Build 2017
9/8/2018 7:26 AM
Designing the database for the cloud
Cost efficiencies with fine-grained multi-tenancy 3
Commodity Hardware
Fine grained multi-tenancy
Adaptive resource
governance
Performance
isolation
IaaS hosted managed database offerings cannot match this!
Elastic and unlimited scalability
Millions of transactions/sec
Petabytes of data
Hundreds transactions/sec
Gigabytes of data 2
Global Distribution 1
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

7. System topology (physical)…
Azure Regions
Datacenters
Stamps
Fault Domains
Container
Machine
Replica
Cluster
Database Engine
Instance
Each cluster spans FDs
Perfect consistency for replica placement
replicas each belonging to a different tenant on a single machine
1000s of tenants in a cluster
100s of tenants on a machine

8. Guaranteeing high availability
Microsoft Build 2017
9/8/2018 7:26 AM
Guaranteeing high availability
Tracking availability at a per tenant and per partition basis at a 5 minute granularity
Self-healing stack where the upper layers automatically recover from the faults raised from the lower layers
Elastic pools of pre-allocated compute with background grow/shrink of clusters
Replica regeneration times are carefully monitored
16+ hours of automated stress tests with failure injection every day
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

9. Reads (1KB) Indexed writes P50 P99 <2ms <10ms <6ms <15ms
Guaranteed low latency at P99 (99th percentile)
Requests are served from local region
Single-digit millisecond latency worldwide
Write optimized, latch-free database engine designed for SSD
Synchronous automatic indexing at sustained ingestion rates
Reads
(1KB)
Indexed writes
P50
P99
<2ms
<10ms
<6ms
<15ms

10. Guaranteeing low latency
Microsoft Build 2017
9/8/2018 7:26 AM
Guaranteeing low latency
Strong
Bounded Staleness
Session
Prefix
Eventual
Indexed Writes
(1 KB)
<10ms + 2RTT
<10ms
<10ms
<10ms
<10ms
Reads
(1 KB)
<10ms
<10ms
<10ms
<10ms
<10ms
Replication protocol is designed to serve reads and writes from the local region
Writes are synchronously indexed upon ingest with durable quorum commits before ACK
Resource governed, log structured and latch free database engine
All data is automatically indexed upon ingestion without requiring explicit schema or secondary index management
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

11. Multiple, well-defined consistency choices
Global distribution forces us to navigate the CAP theorem
Five well-defined consistency levels
Programmatically changeable at any time
Can be overridden on a per-request basis

12. Offering consistency for a price
9/8/2018 7:26 AM
Offering consistency for a price
Consistency distribution among customers
Consistency vs. Throughput
Research in distributed consistency is not rooted in real world scenarios
There is a market for programmable consistency
Tradeoffs between consistency and throughput can be monetized
© Microsoft Corporation. All rights reserved.

13. Elastically scalable storage and throughput
Provisioned request / sec
Time
Nov 2016
Dec 2016
Black Friday
Hourly throughput (request/sec)
Single machine is never a bottle neck
Transparent server-side partition management
Elastically scale storage (GB to PB) and throughput (100 to 100M req/sec) across many machines and multiple regions
Automatic expiration via policy based TTL
Pay by the hour, change throughput at any time for only what you need
Support for both request per second and requests per minute to handle spikes cost-effectively

14. Multi-model, multi-API
Database engine operates on Atom-Record-Sequence type system
All data models can be efficiently translated to ARS
Multi-model: Key-value, Document, and Graph
Multi-API: SQL (DocumentDB), MongoDB, Table, and Gremlin
More data-models and APIs to be added

15. Schema-agnostic, automatic indexing
At global scale, schema/index management is painful
Automatic and synchronous indexing
Hash, range, and geospatial
Works across every data model
Highly write-optimized database engine
Schema
Physical index

16. Schema-agnostic database engine{
"locations": [
{ "country": "Germany", "city": "Berlin" },
{ "country": "France", "city": "Paris" } ],
"headquarter": "Belgium",
"exports":[{"city":"Moscow"}, {"city":"Athens"}] }
At global scale ALTER TABLE, CREATE INDEX, DROP INDEX are a non-starter
Logical index layouts (inverted, tree, columnar, …)
Automatic and synchronous indexing of all ingested content
No schemas or secondary indices ever needed
Resource governed, write optimized database engine
locations
headquarter
exports 1
country
Germany
city
Berlin
France
Paris
Moscow
Athens
Belgium

17. Industry-leading, enterprise-grade SLAs
99.99% availability – even with a single region
Made possible with highly-redundant storage architecture
Guaranteed durability – writes are majority quorum committed
First and only service to offer SLAs on:
Low-latency
Consistency
Throughput

18. Fully resource-governed stack
Capacity management, COGS, SLA, margins, pricing, all depend on stringent resource governance across the entire stack
Request Unit (RU)
Rate based currency (/sec, /min, /hr)
Normalized across various database operations
ML pipeline to calculate the query charges across different datasets and query patterns
Need to remain consistent across hardware generations
Automated perf and RG runs every four hours to detect regressions
All engine micro-operations are finely calibrated to live within the fixed budgets of system resources
% Memory
READ ==
INSERT ==
% CPU
UPSERT ==
% IOPS
DELETE ==
Request Unit (RU)
QUERY == …
Database operations consume RUs A B
Scans
Index Lookups C
Query Complexity D
Instruction
Cost
Query
Cost (Query) == ω x A+ ϕ x B+ λ (C+D)
ω, ϕ, and λ are calculated from ML models trained on
production & test data

19. Security & Compliance Always encrypted at rest and in motion
Fine grained “row level” authorization
Network security with IP firewall rules
Comprehensive Azure compliance certification:
ISO 27001
ISO 27018
EUMC
HIPAA
PCI
SOC1 and SOC2

20. Lowest Total Cost of Ownership (TCO) $ 10x
Deeply exploit cloud properties and economies of scale
Significantly cheaper than DynamoDB, Cassandra, Spanner and MongoDB
Designed from the ground up as a multi-tenant service with end-to-end resource governance to provide performance isolation.
Fully managed as a service - no dev/ops expenses needed
Cost $ 3x
Cosmos DB
DynamoDB
On-premises
MongoDB /
Cassandra

21. Call to Action! Getting Started Technical Overview ->
9/8/2018 7:26 AM
Getting Started
Service Page:
Azure Portal:
Downloadable emulator:
Free Trial:
Technical Overview ->
on Twitter and Tweet #CosmosDB
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

22. We are Hiring #CosmosDB 9/8/2018 7:26 AM
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

23. Please evaluate this session
Tech Ready 15
9/8/2018
Please evaluate this session
From your
Please expand notes window at bottom of slide and read. Then Delete this text box.
PC or tablet: visit MyIgnite
Phone: download and use the Microsoft Ignite mobile app
Your input is important!
© 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

24. 9/8/2018 7:26 AM
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.