Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

22
Data Warehouse Data Warehouse Accelerator Accelerator Michael Wallace Principal Systems Consultant

Transcript of Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

Page 1: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

Data Warehouse Data Warehouse AcceleratorAccelerator

Michael Wallace

Principal Systems Consultant

Page 2: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

2

Why a Data Warehouse Accelerator?Why a Data Warehouse Accelerator?

“Forces related to the new business climate may be fueling an acceleration in the growth of the data

warehouse.... Doing more with less is both the mandate and the mantra across the board.

“More than ever, businesses need dramatic increases in the price/performance of their

systems.

-Richard Winter, VLDB Expert, Winter Corporation

Page 3: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

3

Sound All Too Familiar?Sound All Too Familiar?

How do I protect my revenue that is at risk?How can I better leverage detailed (non-aggregate)data?How can I keep more data online without incurring prohibitive storage expense?How do I perform real-time analysis?How do I run ad hoc queries anytime?How do I get queries to return within seconds or minutes, not hours or days?How do I free up my DBAs from constant warehouse tuning?How do I add users and data without causing performance or architectural disruptions?How can I help my business users be more independent and productive?

• How do I get this project up and running this quarter?• How do I demonstrate a quick and solid return on my investment?• How do I minimize my risk?

Page 4: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

4

Need to get your data back under control?Need to get your data back under control?

Page 5: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

5

Hundreds of companies have benefited from the from the

power, flexibility, scalability and low, low TCO of the IQ

Warehouse Accelerator

Use all your warehouse data….

a subset of your warehouse data….

Or a new set of data…..

Here’s how.

Page 6: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

6

Typical Data Warehouse Architecture

Page 7: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

7

The IQ Accelerator enhances the performance and ROI of any existing data warehouse.

Gauranteed!

Page 8: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

8

Sybase IQ – What is it?Sybase IQ – What is it?

The Database for Analytic Applications

The Only RDBMS 100% designed for decision support

It’s Faster It’s More Scalable

It’s More Economic It has huge market momentum

Page 9: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

9

Sybase IQ Architecture Sybase IQ Architecture

An IQ Database

Is a database with tables

The tables have columns

Uses “indexes” to speed retrieval

Schema Independent

Star

Relational

Flat

Applications Connect to IQ via:

ODBC

JDBC

Open Client

An IQ Database Provides:

Stored Procedures, Functions, and Batches

Views

On-Line Backup

Concurrent Readers/Writers

Transactions

User Defined Functions and User Defined data types

Crash Recovery and logging

Common Language ProcessorANSI 92 SQL

Transact SQL

JAVA

Page 10: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

10

Sybase IQ – What’s Different?Sybase IQ – What’s Different?

Data is Stored Vertically

Each column is stored separately

Bit-Mapped Index

Index on every column

Optimized Storage

Input data is typically compressed

Usually = 30-40%

Database smaller than input data

Even with all the indexes

NOT an MPP Solution

Much simpler implementation

Much simpler management

Query Engine Retrieves Only Columns Used in the Query

Reduces system I/O dramatically

Average 90% Less than competition

Permits better data manipulation

Schema Design Not Restricted

Design based on application use

Flat, Star, Relational, Snowflake

Any Schema

Page 11: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

11

IQ Multiplex ArchitectureIQ Multiplex Architecture Scalability Scalability

A single copy of data shared across multiple computer nodes

All data and indexes stored in the shared database

No partitioning of data required

No distributed lock management

System does not lock on queries or refresh

Individual nodes are Independent of Other Nodes.

Each IQ Node has its own local Temp Space and catalog.

Individual nodes can be different configurations (CPUs, memory, disk).

Data Store (SAN)

Fiber Channel Backbone

ASIQ ReaderASIQ

Writer/Reader

Page 12: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

12

IQ Multiplex ArchitectureIQ Multiplex ArchitectureVertical ScalabilityVertical Scalability

Data Store (SAN)

Fiber Channel Backbone

ASIQ ReaderASIQ

Writer/Reader

ASIQ Reader

Individual nodes can be different configurations (CPUs, memory, disk).

Each IQ engine runs independently, using all available CPUs on its own node.

Additional CPUs scale linearly when added to existing nodes

IQ is CPU, not I/O, bound

Page 13: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

13

IQ Multiplex ArchitectureIQ Multiplex ArchitectureHorizontal ScalabilityHorizontal Scalability

Data Store (SAN)

Fiber Channel Backbone

ASIQ Reader

ASIQ Writer/Reader

ASIQ Reader

No data redistribution

No change in schema

Start small and grow HUGE.

Load balancing can be used to spread out users.

Up to 120 nodes

Page 14: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

14

Data Store (SAN)Data Store (SAN)

IQ Multiplex ArchitectureIQ Multiplex ArchitectureFlexible ScalabilityFlexible Scalability

Data Store (SAN)

Fiber Channel Backbone

ASIQ Reader

ASIQ Writer/Reader

ASIQ Writer/Reader

ASIQ Reader ASIQ Reader

Re-use old hardware

Grow writer node as needed

Increase disk storage without adding nodes

Up to 30+ PetaBytes of disk storage

SMP-like management & tuning

High Availability provided through multiplexing

Page 15: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

15

IQ Multiplex ArchitectureIQ Multiplex ArchitectureUser ScalabilityUser Scalability

EnterpriseEnterprise

WorkgroupWorkgroup

VLDBVLDB

WebWeb

Flexible Scalability = User Scalability

Page 16: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

16

Why Sybase IQ?Why Sybase IQ?

Speed: IQ speeds up queries of your existing warehouse by 10-1000X

Scalability: Scales to thousands to users with virtually no degradation of performance

Flexibility: Allows ad hoc queries anytime with no additional tuning required

Low risk: Already tested, tuned, configured to insure success (world’s largest data warehouse built and certified on IQ/Sun boasts 48.2TB input data, 22TB final size)

Simplicity and elegance: Leading-edge, patented architecture guarantees quick installation, low management and maintenance costs.

Economy: Saves approx. $1 million per terabyte of input data

Page 17: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

17

Need Proof?Need Proof?

Best price-performance in both 300GB and 1000 GB scales

Lowest disk to data ratio: 3 to 10 X better than any other systemAt 1000GB, IQ used 54 disks compared to 1263 and 1408 for competing systems

Best storage efficiency by factor of 10 -- 1TB raw data = 2.4TB storage in IQ, 22TB in competitor

At 300 and 1000 GB scales, IQ is 9 to 25 times less expensive than competitive systems

IQ scores big in TPCH benchmarks

Page 18: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

18

Need proof?Need proof?GIGA study illustrates hefty ROI*GIGA study illustrates hefty ROI*

“The organizations interviewed by Giga information group showed actual or expected returns on their investment ….that ranged from 72% to 175%.

“Simply put, for every dollar invested in the Sun/Sybase RA (IQ running on Sun hardware), $1.63 would be returned to the organization in direct cost savings or increased bottom-line profit as a result of increased business.

“Giga Information Group projects that a composite organization facing some of the same business and IT pressures will likewise achieve a return on investment greater than most standard IT hurdles, and such an investment will pay back its investment in a period of between 13 and 15 months of use.”

*The Total Economic Impact of Deploying the Sun-Sybase Enterprise Data Warehouse Reference Architecture, c. 2003

Page 19: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

19

Need Proof?Need Proof?

“If you are used to queries that take 24 or more hours to run, and then you are told that you can run them in a matter of minutes with Sybase IQ, this may be hard to swallow. The truth is that using a column-based approach really can produce such performance improvements.… IT managers should be ready to fall off their chairs.”

--Bloor Research, 2002

We are able to deliver one data warehouse for all of our applications at one third the storage of conventional technologies, while seeing performance gains as advertised with IQ.

-Kim Ross, CIO Nielsen Media Research

“Conservatively speaking, the CDW’s (Compliance Data Warehouse) Return on Investment is expected to be 200 to 1.”

--Jeffrey Kmonk IRS

Sybase IQ reduced loading and indexing from 30 minutes to 2.5 to 3 minutes. Query speeds were 20 – 50 times faster than Oracle. Time to add a column was reduced from 4 hours with Oracle to 15 minutes with IQ.

Jeff ButlerDepartment of TransportationBureau of Transportation Statistics

Page 20: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

20

Need Proof?Need Proof?Fortis Bank Fortis Bank

Before After

ONE Data Mart (Marketing) Many Data Marts

4 days to load and create all mainframe data into the data warehouse

1.5 TB data loaded every month

8 days to run (in batch) all Business objects reports

115,000 ad-hoc queries/month

Average response time is 4 hours 75% of queries executed within 3 seconds.

Raw Data is only 6 Gigabyte 450 Gigabyte IQ storage

Data warehouse can only be refreshed once a quarter

Data warehouse refreshed every day

20 users 1,000 users

7 external (PWC) consultants write each report

0 external consultants, users write reports themselves.

Full time DBA work DBA work = 10 min. a day

“This is the strongest product I have come across in my career, something I wouldn’t admit that often. There is no doubt whatsoever that another technology would not have offered our users the same service as Sybase IQ.”

--Jean-Louis Catin, IT

Page 21: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

21

Hard to believe? Challenge Us!

Nothing to loose –Everything to Gain!

Proof of Concept.

Page 22: Data Warehouse Accelerator Michael Wallace Principal Systems Consultant.

Sybase IQSybase IQ::

The Ultimate Data Warehouse Accelerator