"The primary online data store for an application is the worst place to take a risk with new technology."  

Adam D' Angelo

(Quora CEO)

CAP Theorem

"Of three properties of shared-data systems-data Consistency, system Availability and tolerance to network Partitions - only two can be achieved at any given moment in time."

2000 Prof. Eric Brewer, PoDC Conference Keynote

2002 Seth Gilbert and Nancy Lynch, ACM SIGACT News 33(2)

It's all about Trade-offs 

Source: http://blog.nahurst.com/visual-guide-to-nosql-systems

But is it really just about the trade-offs?

Trade-offs and theories are confusing!!!

Allow me to share some

Hands-on experience...

Kuai List Headquarters: The most improvised white-board ever!

Requirement #1:

What do we know already?

Requirement #2:

What is trending?

Source: http://db-engines.com/en/ranking_trend

That was easy!!!

Requirement #3:

What about Complex Queries?

AHA!!! Relationships do matter right?

Requirement #4:

What about Transactions?

• It ensures that you atomically make changes to your database.

Requirement #5:

What about Consistency?

• It ensures that clients have the same view of the data

• Achieved through ACID transactions

Requirement #6:

How about Scaling?

• Option 1: I assume we can scale as we go

• Option 2: How about implementing Layers (Memecaches) on top of Relational DBs. (Facebook, Twitter and Airbnb)

.

• Option 3: If scaling is urgent, how about partitioning at the application level. (FriendFeed)

Use Case: Kuai List

• We use PostgreSQL for writing and updating listings.

• For retrieving in desirable speeds we use memecache on top of our core relational database.

• How about data that never changes?

Challenge #1: Host in the Cloud

• Our host (Heroku) offered limited records (10,000 rows).

• We migrated to Amazon RDS: with zero downtime
  • Amazon RDS makes it easy to use replication to enhance availability and reliability for production workloads

Synchronous replication

Data-center B

Data-center C

Data-center A

Data-center D