Download as pdf or txt
Download as pdf or txt
You are on page 1of 27

Data Management in the Cloud

Zhen Shi
Overview
Introduction

3 characteristics of cloud computing

2 types of cloud data management application

2 types of cloud data management architecture

Conclusion
Introduction
What is cloud computing?

Hail as revolutionizing IT

Free corporation

Plug into extremely powerful computing resource


Introduction
Cloud computing platform
3 characteristics of cloud
computing
Three characteristics of a cloud computing

Compute power is elastic


Data is stored at an untrusted host
Data is replicated by crossing large geographic distances
3 characteristics of cloud
computing
Computer power is elastic
Computer resource can be scaled up and down
3 characteristics of cloud
computing
Data is stored at untrusted host
Not really deliver from a celestial location
subject to local rules and regulations

Example:
Amazon S3
3 characteristics of cloud
computing
Data is replicated by crossing large geographic distances
Example:
Amazon EC2

availability
Regions
zones

Persist even
in the face
of failures
of an entire
location
2 types of cloud data management
applications
Based on cloud characteristics

Transactional data management

Analytical data management


2 types of cloud data management
applications
Transactional data management

Not typically use shared-nothing architecture


2 types of cloud data
management applications
What is shared-nothing architecture

Distributed computing architecture

Node is independent and self-sufficient

Single point of connection across the system


2 types of cloud data
management applications
Transactional data management

Hard to maintain ACID when data replication over large


geographic distances
2 types of cloud data
management applications
What is ACID guarantee?

Set of properties that guarantee that database transactions are


processed reliably

Atomicity, Consistency, Isolation, Durability


2 types of cloud data management
applications
Transactional data management

Risks to storing transactional data on an untrusted hosts


mission-critical business processes

Mission-critical
business
processes

Credit card
Customer data
number
2 types of cloud data
management applications
Analytical data management

Perfect match with shared-nothing architecture


2 types of cloud data
management applications
Analytical data management
Unnecessary for ACID guarantees
2 types of cloud data
management applications

Analytical data management


Left out sensitive data
2 types of cloud data
management applications
Comparison with two cloud data management application

Shared- ACID Sensitive data


nothing guarantees
architecture
Transactional Not match Necessary Take care
data
management

Analytical Match Unnecessary Left out


data
management
2 types of cloud data analysis
DBMS market to move into the cloud data management
system

Software solutions to perform the data analysis


MapReduce-like software
Commercially shared-nothing parallel databases
2 types of cloud data analysis
Requirement of cloud DBMS
Efficiency
Fault tolerance
Working under a heterogeneous environment
Operation on encrypted data
Interfacing with business intelligences products
2 types of cloud data analysis
MapReduce-like software

Fault tolerance
High priority

Working under a heterogeneous environment


Able to run in a heterogeneous environment
2 types of cloud data analysis
Operation on encrypted data
No ability to operate on encrypted data

Interfacing with business intelligences products


Not easy interfacing with business intelligences products
Efficiency
Need to discuss
2 types of cloud data analysis
Shared-Nothing parallel database

Fault tolerance
Restart a query upon a failure

Working under a heterogeneous environment


Design to run on homogeneous
2 types of cloud data analysis
Operating on encrypted data
Not able to operate on encrypted data

Interfacing with business intelligences products


Working perfectly

Efficiency
Need to discuss
Conclusion
Parallel database
Advantage:
Efficiency
Performance

MapReduce-like software
Advantage:
fault tolerance
Heterogeneous cluster
Opinion and extension reading

Hybrid solution

Example:
Pig project at Yahoo
SCOPE project at Microsoft
References
https://aws.amazon.com/ec2/

https://aws.amazon.com/s3/

http://cs.yale.edu/homes/dna/papers/abadi-cloud-
ieee09.pdf

You might also like