1. Hello My Friends, Welcome to CosmosDB
Peter Shore
SQL Saturday Cincinnati

2. About Me Database Administrator – Thirty- One Gifts
Intentionally Accidental DBA
Over 20 years IT experience
Server Engineer
Desktop Engineer
Network Infrastructure
Desk side support
President CBusPASS
Co-Organizer SQL Saturday Columbus
How to find me

4. Warning
MongoDB
GraphDB
Gremlin
Cassandra
Key-Value
JSON
XML
NoSQL

5. Relational “vs.” NoSQL Relational Pros Data Integrity
ACID
Atomic – all or none of a transaction
Consistency – transaction never half finished
Isolation – transactions are separated until finished
Durability – pending changes tracked for recovery from abnormal termination
Standardized access via ANSI SQL
Structured
Performance
Scale up

6. Relational “vs.” NoSQL Relational Cons Complexity Performance
Normalization
Data Types
Maintenance
Performance
Abundance of information sources
Scale up

7. Relational “vs.” NoSQL Pros of NoSQL Cons of NoSQL No formal structure
Flexibility
Performance
Scale out
Large volumes of data
Minimal maintenance
Lack of ACID
Eventually consistent
Lack of formal structure
Lack of standardization

8. What is CosmosDB? Microsoft says*
Globally distributed
Multi-model database
Multi-master
Elastically and independently scale
Throughput
Storage
Across any number of Azure's geographic regions
Offers throughput, latency, availability, and consistency guarantees with comprehensive service level agreements (SLAs). *

9. SLAs

10. What is CosmosDB? Globally Distributed
Across any number of Azure regions
Regions are sets of data centers
Tier 1 Azure application
Add and remove regions from CosmosDB
No application redeployment
Highly available
Data where the users are
CosmosDB’s multi-homing API
Requests to nearest data center
No configuration changes required

12. What is CosmosDB? Multi-model Graph Use graph structures to store data
Graph structures are made up of nodes, edges, and properties
Nodes represent entities
Edges represent relationships
Properties are germane information to nodes
Key concept the graph (or edge or relationship)
Directly relates items in the data-store
Links allow data in the store to be connected directly and often accessed in one operation
Uses Gremlin API in CosmosDB
Fraud Detection & Analytics Solution. ...
Knowledge Graph. ...
Network and Database Infrastructure Monitoring. ...
Recommendation Engine & ...
Master Data Management. ...
Social Media and Social Network Graphs.

14. What is CosmosDB? Multi-model continued
Key-value database (aka key-value store)
Designed for storing, retrieving and managing associative arrays
Associative arrays are known as dictionaries or hash tables
Collections of objects/records
Records are stored and retrieved using keys
Records may contain different fields
Each record treated as an opaque collection
Phone directory
Stock trading (Symbol & value)
Artist Info
IP Forwarding table

15. What is CosmosDB? Multi-model continued MongoDB Documents
Not inherently opaque
Relies on inner structure of documents to extract metadata
Can be XML, JSON, YAML, or BSON
Also binary formats such as PDF or Office topics
Core operations are Create, Retrieve, Update, Delete (CRUD)
Documents are addressed by a unique key which is indexed
Organized
Collections
Tags/non-visible meta data
Directories
Variable length documents
JSON/BSON
Performance
Scale out
Dynamic
Operational Intelligence, log storage
Product catalogs

16. What is CosmosDB? Multi-model continued Table
API for applications written to use Azure Table storage
Column-family (Cassandra Data Model)
Maximize writes and data duplication
Minimize partitions used
SQL API
Document database
Accessible with familiar SQL syntax
Schema not required

17. What is CosmosDB? Atom-Record-Sequence* Core system type of CosmosDB
Atoms are a small set of primitive types
String
Bool
Number
Records are structs
Sequences are arrays of atoms, records, or sequences
Azure CosmosDB engine efficiently translates and projects other data models onto ARS
Allows for native support of APIs *
Enable customers to elastically scale throughput and storage based on demand, globally. The system should deliver the configured throughput within 5 seconds at the 99th percentile, from the time of the request to scale.
Enable customers to build highly responsive, mission-critical applications. The system must deliver predictable and guaranteed end-to-end low read and write latencies at the 99th percentile.
Ensure that the system is “always on”. The system must provide 99.99% availability regardless of the number of regions associated with their database. To enable customers to test the end-to-end availability properties of the applications, (in steady state) the service must also allow customers to simulate regional failures or mark regions associated with their database offline. This helps validate the end-to-end availability properties of applications.
Enable developers to write correct globally distributed applications. The system must offer an intuitive and predictable programming model around data consistency. While strong consistency comes with a price, writing large globally distributed applications against an “eventually consistent” database results in an application code which is hard to reason about, is brittle, and rife with correctness bugs.
Offer stringent financially-backed comprehensive SLAs for 1, 2, 3 and 4 above.
Relieve the developers from the burden of database schema/index management and versioning. Keeping database schema and indexes in-sync with an application’s schema is especially painful for globally distributed applications.
Natively support multiple data models and popular APIs for accessing data. The translation between the externally exposed APIs and internal data representation needed to be efficient.
Operate at a very low cost to pass on the savings to customers.

18. Request Units Request Units/second Measurement used for charges
Can vary per API
Abstracts system resources (CPU, IOPS, and memory)

19. Good, Bad, and Ugly of Backups
Microsoft managed
Taken every 4 hours
Bad
Only last two snapshots retained
Must contact support for restore
If container or database is deleted, last backup retained for 30 days
Ugly
Copy data with ADF

20. Partitioning, Not What You Think
Splitting items into distinct subsets known as logical partitions
Scale containers within a database
Partitions formed based on value of the partition key
Once defined partitions automatically managed
New logical partitions are created as new items are added

21. Choosing a Key Avoid hot spots
Requests to a partition key cannot throughput allocated to partition
Requests limited if throughput exceeded
Key needs to spread workload evenly across partitions and evenly over time
Needs a wide range of values
Portion cannot exceed 10GB of storage

22. Indexing By default all items are indexed
Default policy can be overridden
Include/exclude terms or properties
Can add additional index types, such as spatial
Index Types
Data Types (String, Number, Point, Polygon, or LineString)
Index Kind (Range – equality, range or order by queries) or Spatial
Precision - Range index

23. Index Kinds Range Spatial Index Supports equality queries JOIN queries
Range queries
Data type of string or Number
Spatial Index
Efficient within and distance queries

24. Use Cases Common Uses Cases IoT Retail (Windows Store)
Gaming Services (XBoxLive)
Gaming Statistics

25. IoT Use Case

28. Access Data MongoDB & Table Gremlin Cassandra
Primary Connection String
Secondary Connection String
Read Only Connection String
Gremlin
Gremlin Endpoint
Cassandra
Read-write
Read-only

29. Further Data Access PowerBI
Analysis Services/Azure Analysis Services (in-memory only)
SQL Server
Linked Servers
Polybase (SQL Server 2019)

30. Questions?