RDBMS & noSQL

RDBMS & noSQL Mixed for best performance

Javier Tomás Zon Senior Technical Operations

April, 2016

www.percona.com

RDBMS & noSQL

• Who am I? • Differences • Best Practices • Q&A

www.percona.com

@jtomaszon

• 14 years as System Administrator

• 9 years working with MySQL

• 5 years using DevOps culture

• ex-TOTVS

• IoT Enthusiastic (Arduino, Raspberry Pi)

www.percona.com

RDBMS & noSQL

• Vertical Scale • Horizontal Scale

www.percona.com

RDBMS & noSQL

• Structured Data • Non-structured Data

www.percona.com

RDBMS & noSQL

• Structured Data • Non-structured Data

www.percona.com

RDBMS & noSQL

• Structured Data • Non-structured Data

ID USER COLOR

1 jav blue

USER_ID COLOR

1 blue

1 red

{ “id”: 1, “user”: “jav”, “color”: “blue" }

{ “id”: 1, “user”: “jav”, “color”: [“blue”, “red"] }

www.percona.com

RDBMS & noSQL

• Atomic Transactions • Eventual Consistency

put node

node

node

node

node

node

node

www.percona.com

Best Practices

• RDBMS

• Transactions

• Data Consistency

• Structured Data

www.percona.com

Best Practices

• noSQL

• Map & Reduce

• Eventual Consistency

• Geo-references

• Denormalized Data

MongoDB NoSQL Quick Talk

Adamo Tonete MongoDB Senior Support Engineer

April, 2016

www.percona.com

MongoDB: noSQL Quick Talk

• Who am I? • What is the best

NoSQL product to scale out? • When consider scale out? • Let's talk about MongoDB • Q&A

www.percona.com

@adamotonete

• 8 years working as DBA

• MongoDB certified

• A few interesting companies before Percona

• Jointed Percona 5 Months ago.

www.percona.com

What is the best NoSQL product to scale out?

• There are many flavours of NoSQL:

• Key/Value • Documents • Column-oriented • Objects

www.percona.com

When consider scale out?

• Workset bigger than RAM • High CPU usage • Disk O/I exhausted • Redundancy • Decrease Latency

www.percona.com

Let's talk about MongoDB

• A few things that you need to know before scale MongoDB:

• MongoDB has a single master for each replicaset.

• Sharding must be very well configured, otherwise you will only have a few machines idle.

• Look at what you want to scale: • reads? • writes?

www.percona.com

Let's talk about MongoDB

• High consistence will cost more time to confirm a operation.

• Downsize is not so easy operation all data will be removed by one single thread, by design.

• Keep your eyes open to MongoS and Config Servers those guys are responsible to keep a shard working.

www.percona.com

Resources

• sitepoint • Differences:

http://www.sitepoint.com/sql-vs-nosql-differences/ • How to Choose:

http://www.sitepoint.com/sql-vs-nosql-choose/

• percona.com/blog • AWS Training • youtube.com

• CloudU channel • Rackspace Channel

www.percona.com

QUESTIONS?