8/7/2019 Data in the Cloud Presentation

1/13

Data in the cloud

OReilly MySQL ConferenceMrten MickosCEO, Eucalyptus Systemswww.twitter.com/martenmickos

8/7/2019 Data in the Cloud Presentation

2/13

2011 Eucalyptus Systems, Inc.

Infrastructure software

MF Mini Client/Server Cloud

Internet era

Web,Web2.0, Mobile internet

LAMP

1m

10m

100m

1bn

10bn

users now

8/7/2019 Data in the Cloud Presentation

3/13

8/7/2019 Data in the Cloud Presentation

4/13

2011 Eucalyptus Systems, Inc.

8/7/2019 Data in the Cloud Presentation

5/13

2011 Eucalyptus Systems, Inc.

FOSS model

Ten years ago

disruptive movement

lead by innovative

growth companies distribution-centric

software model

enabled by the

internet

Today

an established

practice lead by large

corporations service-centric

software model

being redefined by

cloud computing

8/7/2019 Data in the Cloud Presentation

6/13

2011 Eucalyptus Systems, Inc.

Dominant db designs

scale-upRDBMS scale-out RDBMS

big data

key/value

doc

db graphdb

80s & 90s 00s & 10s 10s & 20s

?

NoSQL /

NewSQL land

0% 1% 10% 30% of world population online

8/7/2019 Data in the Cloud Presentation

7/13

2011 Eucalyptus Systems, Inc.

MySQL - NoSQL - NewSQL

Akiban, Cassandra, Citrusleaf, Clustrix,

Couchbase, Dynomite, FlockDB, GenieDB,

Hadoop, Hive, HyperGraphDB, HyperTable,

MarkLogic Server, Memcached, MemSQL,MongoDB, MySQL, MySQL Cluster, MySQL

with HandlerSocket, Neo4J, NimbusDB,

Objectivity/DB, Ravel, Redis, RethinkDB,Riak, SimpleDB, Terracotta, Tokyo Cabinet,

Voldemort, VoltDB, Xeround

8/7/2019 Data in the Cloud Presentation

8/13

2011 Eucalyptus Systems, Inc.

Reversal of DBMS design principles

Client/Server

Strength

complex feature set

scale-up

ACID

Weaknesssimplicity

scale-out

non-ACID

Cloud

Strength

Weakness

8/7/2019 Data in the Cloud Presentation

9/13

2011 Eucalyptus Systems, Inc.

Data in the cloud

1. Design the database for scale-out on multiple nodes

separate read activity from write activity

scale reads through replication

scale writes through sharding & partitioning

consider different dbms:s for different uses within theapplication

plan for failures

plan for 10x growth steps

2. Deploy on a cloud service (public) or cloud platform

(private, on-prem) with common abstraction layer (API)

8/7/2019 Data in the Cloud Presentation

10/13

8/7/2019 Data in the Cloud Presentation

11/13

2011 Eucalyptus Systems, Inc.

Future

Cloud, Mobile & Social are the drivers, and

they are bigger than anything weve seen

Existing models ofFOSS, community anddistribution being disrupted

Major rewards for creating the winningsolution(s) for data in the cloud

8/7/2019 Data in the Cloud Presentation

12/13

2011 Eucalyptus Systems, Inc.

Thank you!

Mrten Mickosmarten@eucalyptus.com

twitter.com/martenmickos

8/7/2019 Data in the Cloud Presentation

13/13

2011 Eucalyptus Systems, Inc.

Eucalyptus cloud platform

Over 25,000 new Eucalyptus clouds were started in 2010