These materials are 2017 John Wiley & Sons, Inc.

. Any dissemination, distribution, or unauthorized use is strictly prohibited.

 Cloud Data
Warehousing
Snowflake Special Edition

by Joe Kraynak

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Cloud Data Warehousing For Dummies, Snowflake Special Edition

Published by
John Wiley & Sons, Inc.
111 River St.
Hoboken, NJ 07030-5774
www.wiley.com
Copyright 2017 by John Wiley & Sons, Inc., Hoboken, New Jersey

No part of this publication may be reproduced, stored in a retrieval system or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise,
except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without
the prior written permission of the Publisher. Requests to the Publisher for permission should be
addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken,
NJ07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/
permissions.
Trademarks: Wiley, For Dummies, the Dummies Man logo, The Dummies Way, Dummies.com,
Making Everything Easier, and related trade dress are trademarks or registered trademarks of
John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries, and may
notbe used without written permission. All other trademarks are the property of their respective
owners. John Wiley & Sons, Inc., is not associated with any product or vendor mentioned in
thisbook.

LIMIT OF LIABILITY/DISCLAIMER OF WARRANTY: THE PUBLISHER AND THE AUTHOR MAKE NO

REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE ACCURACY OR COMPLETENESS OF
THE CONTENTS OF THIS WORK AND SPECIFICALLY DISCLAIM ALL WARRANTIES, INCLUDING
WITHOUT LIMITATION WARRANTIES OF FITNESS FOR A PARTICULAR PURPOSE.NO WARRANTY
MAY BE CREATED OR EXTENDED BY SALES OR PROMOTIONAL MATERIALS. THE ADVICE
AND STRATEGIES CONTAINED HEREIN MAY NOT BE SUITABLE FOR EVERY SITUATION.THIS
WORK IS SOLD WITH THE UNDERSTANDING THAT THE PUBLISHER IS NOT ENGAGED IN
RENDERING LEGAL, ACCOUNTING, OR OTHER PROFESSIONAL SERVICES. IF PROFESSIONAL
ASSISTANCE IS REQUIRED, THE SERVICES OF A COMPETENT PROFESSIONAL PERSON SHOULD BE
SOUGHT.NEITHER THE PUBLISHER NOR THE AUTHOR SHALL BE LIABLE FOR DAMAGES ARISING
HEREFROM.THE FACT THAT AN ORGANIZATION OR WEBSITE IS REFERRED TO IN THIS WORK
AS A CITATION AND/OR A POTENTIAL SOURCE OF FURTHER INFORMATION DOES NOT MEAN
THAT THE AUTHOR OR THE PUBLISHER ENDORSES THE INFORMATION THE ORGANIZATION
OR WEBSITE MAY PROVIDE OR RECOMMENDATIONS IT MAY MAKE. FURTHER, READERS
SHOULD BE AWARE THAT INTERNET WEBSITES LISTED IN THIS WORK MAY HAVE CHANGED OR
DISAPPEARED BETWEEN WHEN THIS WORK WAS WRITTEN AND WHEN IT IS READ.

For general information on our other products and services, or how to create a custom For
Dummies book for your business or organization, please contact our Business Development
Department in the U.S. at 877-409-4177, contact [email protected], or visit www.wiley.com/go/
custompub. Forinformation about licensing the For Dummies brand for products or services,
contact BrandedRights&[email protected].
ISBN 978-1-119-35192-4(pbk); ISBN 978-1-119-35190-0(ebk)

Manufactured in the United States of America

10 9 8 7 6 5 4 3 2 1

Publishers Acknowledgments

Project Editor: Christina Guthrie Production Editor: Antony Sami

Acquisitions Editor: Steve Hayes Snowflake Review Team:
Editorial Manager: Rev Mengle VincentMorello, Jon Bock,
KentGraziano
Business Development
Representative: Karen Hattan

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Introduction
A
s an executive, manager, or analyst, youre well aware that
knowledge is power and that data properly analyzed on a
timely basis provides the insight necessary to make well-
informed decisions and achieve a competitive advantage.

Today, companies have a much greater collection of more relevant

data than ever before. This includes a diverse range of sources,
internal and external, including data marts, cloud-based applica-
tions, and machine-generated data.

Unfortunately, the data warehouse architecture of the past strains

under the burden of extremely large, diverse data sets. Analysts
often wait 24 hours or more for data to flow into the data ware-
house before its available for analysis. They can wait even longer
for complex queries to run on that data. In many cases, the stor-
age and compute resources required to process and analyze that
data are insufficient. This leads to systems hanging or crashing.
To avoid this, users and workloads must be queued, which results
in even longer delays.

To remain efficient and competitive, organizations must be able

to harness the power of the vast amounts of data constantly being
generated and conduct complex analysis on that data. Fortunately,
advances in computer hardware, architecture, and software can help
your organization meet this challenge and exceed your expectations.

About This Book

Welcome to Cloud Data Warehousing For Dummies, where you
discover how your organization can tap the power of massive
amounts of data conveniently and affordably to enhance effi-
ciency and transform raw data into valuable business intel.

More data opens the door to more and bigger opportunities, which
are almost always accompanied by equally big challenges. To take
advantage of these big opportunities, you need to find and imple-
ment a data warehouse solution that can store and organize data
in diverse formats, provide convenient access to it, and improve
the speed at which you can analyze that data. And it must be done
as cost-effectively as possible. This book shows you how.

Introduction 1

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Foolish Assumptions
We surmise that you grasp the concept of data warehousing and
the challenges and opportunities it presents.

We also assume youre an analyst, database administrator, or

other stakeholder or influencer in your organization, who wants a
fundamental understanding of cloud data warehousing and how it
can support the efforts and expertise of the people in your orga-
nization who need to access and analyze data. Or you may be a
decision maker who needs the information and insight to choose
the best data warehouse solution for your company.

Icons Used in This Book

Throughout this book youll find the following icons that high-
light tips, important points to remember, and more:

This icon guides you to faster, easier ways to perform a task

or better ways to put cloud data warehousing to use in your
organization.

This icon highlights concepts worth remembering as you immerse

yourself in the understanding and application of cloud data
warehousing.

Throughout this book are case studies that reveal how various
companies applied cloud data warehousing in real-world situa-
tions. They significantly improved the speed and performance of
their data storage and analytics systems and saved money in the
process.

Beyond the Book

If you like what you read in this book and want to know more, we
invite you to visit www.snowflake.net, where you can find out more
about the company and what they offer, obtain details about differ-
ent plans and pricing, view webinars, access news releases, get the
scoop on upcoming events, access documentation and other sup-
port, and get in touch with them theyd love to hear fromyou!

2 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Data warehousing: past to present

Understanding the benefits of a cloud

data warehouse

Recognizing where cloud data

warehousing fits in todays economy

Chapter 1
Getting Up to Speed on
Cloud Data Warehousing

I
n one form or another, cloud computing and software-as-a-service
(SaaS) have been around for decades. But cloud data warehouse-
as-a-service (DWaaS) has only recently emerged as an alternative
to conventional, on-premises data warehousing and similar types of
solutions that have appeared in recent years. Why? Why now? Whats
changed? In this chapter, we answer these questions, and more.

We begin by defining what a data warehouse is and explore the

evolution of data warehousing to understand how this technology
has made its way to the cloud. Then we look at how organizations
can benefit from cloud DWaaS and explain why more companies
rely on cloud data warehousing to compete in todays data-driven
economy.

What Is a Data Warehouse?

A data warehouse is a computer system dedicated to storing and
analyzing data to reveal trends, patterns, and correlations that
provide information and insight. Traditionally, organizations have
used data warehouses to store and integrate data collected from
their internal sources (usually transactional databases), includ-
ing marketing, sales, production, and finance. The data warehouse

CHAPTER 1 Getting Up to Speed on Cloud Data Warehousing 3

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 emerged when companies realized that analyzing data directly
from those internal systems competed with the day-to-day activi-
ties of business users such as data entry and operational reporting.

Over the years, data sources have expanded beyond internal busi-
ness operations and external transactions. They now include
exponentially greater volumes of data and more complex data
from websites, mobile phones, online games, online banking apps,
and even machines. Most recently, companies are capturing huge
amounts of data from IoT (Internet of things)-enabled devices.

The Evolution ofData Warehousing

Historically, businesses collected data in well-defined, highly
structured forms at a reasonably predictable rate and volume.
Even as the speed of older technologies advanced, data access and
usage were carefully controlled and limited to ensure acceptable
performance for every user. This required businesses to be more
tolerant of longer analytics cycles.

Times have changed (see Figure 1-1). Advances in technology

mean companies can now make significant business decisions
backed by large amounts of data. And its not just the market
leaders or mature companies. Smaller, nimble market entrants
continue to transform well-established industries within months
or just a couple of years. Theyre doing so with data to reveal
opportunities and develop products and services that change how
retail and business vendors engage their customers.

Illustration supplied by Snowflake.

FIGURE1-1: Data warehousing has evolved over four decades.

4 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Recognizing the limitations of
conventional data warehousing
Conventional data warehouse solutions were not designed to
handle the volume, variety, and complexity of todays data. And
newer systems designed to address these shortcomings struggle
to accommodate the data access and analysis that organizations
now require. Todays challenges reveal:

Data sources are more numerous and varied, resulting in

more diverse data structures that must co-exist in a single
location to enable exhaustive and affordable analysis.
Traditional architectures inherently cause competition
between users and data integration activities, making it
difficult to simultaneously pipe new data into the data
warehouse and provide users with adequate performance.
Scaling up a conventional data warehouse to meet todays
increasing storage and workload demands, when possible, is
expensive, painful, and slow.
The more recent, alternative data platforms are often complex,
requiring specialized skills and lots of tuning and configura-
tion. This struggle worsens when trying to handle the
growing number and diversity of data sources, users, and
queries.

But all is not lost! Like all great things, technology evolves. New
ideas and new methods emerge to address the significant busi-
ness problems of today and the aspirations of tomorrow.

Technology and design tothe rescue!

The good news is that technology, and data warehousing
architecture (the design and building blocks of the modern data
warehouse), have evolved to address the demands of the data-
driven economy with the following innovations:

The cloud: A key factor driving the evolution of the modern

data warehouse is the cloud. This creates access to near-
infinite, low-cost storage; improved scalability; the outsourc-
ing of data warehousing management and security to the
cloud vendor; and the potential to pay for only the storage
and computing resources actually used.

CHAPTER 1 Getting Up to Speed on Cloud Data Warehousing 5

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Massively parallel processing (MPP): MPP emerged in the
previous decade, which involves dividing a single computing
operation to execute simultaneously across a large number
of separate computer processors. This division of labor
facilitates faster storage and analysis of data when software
is built to capitalize on this approach.
Columnar storage: Traditionally, databases stored records
in rows, similar to how a spreadsheet appears. For example,
this could include all information about a customer or a retail
transaction. Retrieving data the traditional way required the
system to read the entire row to get one element. This is
laborious and time-consuming. With columnar storage, each
data element of a record is stored in a column. With this
approach, a user can query just one data element, such as
gym members who have paid their dues, without having to
read everything else in that entire record, which may include
each members ID number, name, age, address, city, state,
payment info, and so on. The approach can provide a much
faster response to these kinds of analytic queries.
Vectorized processing: This form of data processing for
data analytics (the science of examining data to draw
conclusions) takes advantage of the recent and revolutionary
computer chip designs. This approach delivers much faster
performance versus older data warehouse solutions built
decades ago for older, slower hardware technology.
Solid state drives (SSDs): Unlike hard disk drives (HDDs),
SSDs store data on flash memory chips, which accelerates
data storage, retrieval, and analysis. A solution that
takesadvantage of SSDs can deliver significantly better
performance.

For more about advances in technology and other trends that

drive the evolution of data warehousing, see Chapter2.

Introducing the cloud data warehouse

Cloud data warehousing is a cost-effective way for companies to
take advantage of the latest technology and architecture with-
out the huge upfront cost to purchase, install, and configure the
required hardware, software, and infrastructure. The various
cloud data warehousing options are generally grouped into the
following three categories:

6 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Traditional data warehouse software deployed on cloud
infrastructure: This option is very similar to a conventional
data warehouse, as it reuses the original code base. So you
still need IT expertise to build and manage the data ware-
house. While you do not have to purchase and install the
hardware and software, you may still have to do significant
configuration and tuning, and perform operations such as
regular backups.
Traditional data warehouse hosted and managed in the
cloud by a third party as a managed service: With this
option, the third party provider supplies the IT expertise, but
youre still likely to experience many of the same limitations
of a conventional data warehouse. The data warehouse is
hosted on hardware installed in a data center managed by
the vendor. This is similar to what the industry referred to as
an ASP or application service provider. The customer still has
to specify in advance how much disk space and compute
resources (CPUs and memory) they expect to use.
A true SaaS data warehouse: With this option, often
referred to as data-warehousing-as-a-service, (DWaaS), the
vendor delivers a complete cloud data warehouse solution
that includes all hardware and software and the IT and
database administration (DBA) expertise required. Clients
typically pay only for the storage and computing resources
they use, when they use them. This option should scale up
and down on demand.

For a more detailed comparison of cloud data warehousing solu-

tions, turn to Chapter5.

Why You Need a Cloud Data Warehouse

Any organization that depends on data to better serve their
customers, streamline their operations, and lead their industry
will benefit from a cloud data warehouse. Unlike massive, tradi-
tional data warehouses, the cloud means businesses big and small
can size their data warehouse to meet their needs and their bud-
get, and dynamically grow and contract their system as things
change from day-to-day and year-to-year.

CHAPTER 1 Getting Up to Speed on Cloud Data Warehousing 7

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Here are a few areas where cutting-edge cloud data warehouse
technology can significantly improve a companys operations:

Customer experience: Monitoring end-user behavior can

help companies tailor products, services, and special offers
to the needs and demands of individual consumers. With
customer sentiment analysis, companies understand better
what customers think by analyzing their social media
postings, tweets, and other online messaging.
Quality assurance: Companies can also use the analysis of
customer sentiment to monitor for early warning signs of
customer service issues or product shortcomings and take
action sooner than was previously possible when the only
data source was call center complaint logs.
Operational efficiency: Operational intelligence (OI) consists
of monitoring business processes and analyzing events to
identify where a company can reduce costs, boost margins,
streamline processes, respond to market forces more
rapidly, and so on.
Innovation: Instead of only checking the rearview mirror to
understand an industrys recent past, companies can use
new sources of data to spot and capitalize on trends, thereby
disrupting their industry before an unknown or unforeseen
competitor does so first.

Nearly all of a companys data is stored in a multitude of disparate

databases. The key questions to ask are: How accessible is that
data? How much will it cost to extract, store, and analyze all of
your data? And, what will happen if you dont? This is where data
warehousing comes into play.

8 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Adapting to increasing demands for data
access and analytics

Adjusting to how data is created and

used today

Tackling the challenges with new and

improved technologies

Chapter 2
Why the Modern Data
Warehouse Emerged

C
loud data warehousing emerged from the convergence of
three major trends changes in data sources, volume and
complexity; increased demand for data access and analyt-
ics; and technology improvements that significantly increased the
efficiency of data storage, access, and analytics.

In this chapter, we describe these trends in greater detail for a

clearer understanding and appreciation of cloud data warehousing.
We also reveal how a data warehouse can take advantage of the
benefits of cloud to address these trends.

Trends in Data: Volume, Complexity,

and Usage
When we talk data in this book, were talking petabytes. One
petabyte is equal to about 500 billion pages of standard, printed
text and can appear neatly typed in rows and columns in a data-
base. Data also comes from people using websites and software
applications on their phones and other mobile devices, or data
produced by digital or mechanical devices.

CHAPTER 2 Why the Modern Data Warehouse Emerged 9

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 In this section, we focus on changes in data, and data use, that
have led to demand for cloud data warehousing.

The data tsunami

In the not-so-distant past, businesses generally managed data
that was entered manually into the system by human beings. They
may have also had data from external sources, such as customers,
clients, and partners. The amount of data was relatively small and
predictable, and stored, managed, and secured in a companys
data center known as on-premises.

Today, the business world is experiencing a data tsunami with data

available from a variety of sources already mentioned in this book,
and other sources too numerous and varied to list. The volume
and complexity of this data can quickly overwhelm a conventional,
on-premises data warehouse, and often cause data processing and
analysis to hang, or even crash the system.

Adapting to the exponential increase of data requires a fresh per-

spective (see Figure2-1). The conversation must shift from how
big a companys data warehouse must be to whether it can scale
cost-effectively, without friction, and on the order of magnitude
necessary to handle massive volumes of data.

The use cases that cloud data warehousing has sparked continue
to emerge. For example, SaaS-born companies and big enterprises
that use the cloud to store their data are monetizing (selling) that
data. They package it as a service and sell it to other organizations
keen to make better business decisions.

Illustration supplied by Snowflake.

FIGURE2-1: The modern data warehouse enables all data for all users.

10 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Data born in the cloud
The business world has experienced a rapid adoption of SaaS,
including customer relationship management (CRM) software,
business management (ERP) software suites, advertising buying
platforms, and online marketing tools, just to name a few. Thanks
to the cloud, new SaaS companies can set up shop with just the
price of a laptop or two. These products can create huge amounts
of valuable data stored in the cloud.

Demand for SaaS/cloud applications has also grown. Ease of

deployment pales to what on-premises applications require to get
up and running. In the past, a company may have operated only
five to ten significant enterprise applications generating data.
Now, its normal for even midsize organizations to have hundreds
or even thousands of applications, each with the potential of cre-
ating its own data silo marketing data in one system, finance in
another, product information in yet another and none of them
integrated for complete and optimum analysis.

With the majority of an organizations data now in the cloud,

the natural place to integrate this data is also in the cloud. With
cloud data warehousing, you can integrate cloud data in the cloud.
Youre no longer forced to pull it inside your data center, which is
very expensive and time-consuming, and makes less sense as the
amount of cloud-native data grows.

Machine-generated data
Machine-generated data is a key topic related to the Internet of
Things (IoT). Its an endless collection of devices that communi-
cate data via the Internet, including smart phones, thermostats,
refrigerators, oil rigs, home security systems, smart meters, and
much more. Data collected and analyzed from IoT devices can
enhance products and processes, monitor equipment, and predict
needed maintenance to avoid failure.

But a lot of machine-generated data has a poor signal-to-noise

ratio. It contains valuable data but also a lot of noise. Therefore,
you often must store all of it to find the valuable bits. In addition,
a growing share of this data originates outside your data center.
This makes cloud, and its near-infinite scalability, the natural
location to store and integrate this data.

CHAPTER 2 Why the Modern Data Warehouse Emerged 11

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Data exploration
Analyzing data starts with data exploration identifying inter-
esting and valuable connections and serving them up to data users
in the form of reports and analytics. Although data exploration
isnt a new concept, the growth in data volume makes it a more
resource-intensive exercise.

Data exploration often involves large data sets. Its also often
experimental in nature, which complicates the ROI assessment
needed to support the significant upfront cost of deploying a tra-
ditional, on-premises data warehouse. In response, the cloud can
enable a data warehouse to scale up and down as needed, and
offers a pay-for-use model to avoid the challenge of whether or
not to make an expensive, upfront commitment.

Data lakes
The growing need to have massive amounts of raw data in different
formats, all in a single location, spawned the data lake. But com-
panies quickly realized that transforming that data and extracting
valuable insight was a cost-prohibitive, labor-intensive process
to even attempt these efforts.

But the original interest in data lakes reveals that companies want
to store all of their data in one location at a reasonable cost. With
a modern data warehouse, the cloud supports the cost-effective
methods to store and transform data with on-demand resources
to minimize this resource-intensive process.

Trends in Reporting and Analytics

Data-driven decision-making is no longer relegated to the execu-
tive team or data scientists. Its now used to improve nearly every
operational aspect of a company. But this increasing demand
for data access and analytics across an organization can slow or
crash a system as workloads compete for storage and comput-
ing resources from traditional data warehouses. Efficiency drops,
requiring more time and money invested in additional infrastruc-
ture to maintain the system.

In this section, we identify some of the trends that are changing

how people access and use data and how those trends drive the
need for modern, built-for-the-cloud data warehouse solutions.

12 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Elasticity to enable analytics
Here are a few scenarios where true, cloud-based elastic data
warehousing can make it possible to do more with data:

As discussed earlier, data exploration has many benefits. But

no one really knows in advance the computing resources
needed to analyze huge data sets. This makes on-demand,
elastic scalability ideal for this analysis.
Ad hoc data analysis answers a single, specific, business
question, which emerges all the time. Dynamic elasticity
provides the flexibility and adaptability to perform these
queries without slowing down other workloads.
Event-driven analytics demand constant data. They incorporate
new data to update reports and dashboards on a continual basis
so senior managers can monitor the business in real time or close
to it. Ingesting and processing streaming data requires an elastic
data warehouse to handle variations and spikes in data flow.

Rapid iteration replacing exhaustive

pre-planning
Entrepreneurs typically have two paths to follow when ensur-
ing the marketability of a new idea: exhaustive preplanning or
rapid iteration. The first option is a traditional, time-consuming
process that involves thinking through an opportunity or a new
product idea, kicking ideas back and forth, and hoping it creates
consumer demand. Rapid iteration involves quickly testing the
idea in the market to iterate over and over until a viable version of
the product shows success. From there, the process begins again.

Rapid iteration has emerged as the more effective process to dis-

mantling established competitors and altering how an entire indus-
try does business. But it requires high-speed collection and analysis
of a large amount of data to be successful. Advances in cloud data
warehousing and analytics have made rapid iteration more practical.

Embedded analytics
For many companies, analytics operate as a separate and distinct
business process. But a growing trend is to build analytics into
business applications, which are increasingly built in the cloud.
These applications handle significant variability in the number

CHAPTER 2 Why the Modern Data Warehouse Emerged 13

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 of users that query the applications and the number of queries
(workloads) users run to analyze that data. The cloud facilitates
data transfers from cloud-based applications to the organiza-
tions cloud data warehouse, where its scalability and elasticity
can better support fluctuations in users and workloads.

MEETING INCREASED DEMAND FOR

DATA ANALYTICS AT JANA
Jana provides free, unrestricted Internet access to more than 30 mil-
lion smartphone users in more than 15 emerging markets. With its
mCent Android app, Jana shifts the cost of mobile Internet from cus-
tomers to more than 4,000 brands via sponsored content.

When new, branded content or mCent features are introduced, Jana

analyzes and measures key metrics, including user attention, lifetime
user value, and key performance indicators (KPIs).

As Jana and its data grew, the companys initial analytics architecture
could no longer efficiently serve its business. Queries slowed and
table scans became unfeasible. Adding capacity and backup systems
and administering its open-sourced data repository required more
and more administration time.

As illustrated in the following figure, Jana upgraded most of its data

platform components to streamline its system with a cloud-based
data warehouse to overcome these barriers and gain the following
benefits:

Keep pace with the business demands of processing and analyzing

a rapidly growing stream of disparate data.
Encourage an increased use of analytics throughout the company;
80 percent of Janas employees access the data warehouse.
Significantly reduce administration overhead.

Janas transformation to a faster, cheaper and more effective data

warehouse.

14 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Technology Musts for Any Modern Data
Warehouse
Technology innovations can improve data warehousing and ana-
lytics with regard to availability, simplicity, cost, and perfor-
mance. In this section, we focus on the key technologies that
should be part of any modern data warehouse.

Cloud
The properties of cloud make it particularly well-suited for data
warehousing. Weve mentioned these in other contexts but its
also important to know that they came from cloud:

Unlimited resources: Cloud infrastructure delivers near-

unlimited resources, on demand, and within minutes or
seconds. Organizations pay only for what they use, making it
possible to dynamically support any scale of users and
workloads without compromising performance.
Save money, focus on data: Companies that choose a
cloud-built solution avoid the costly, up-front investment of
hardware, software, and other infrastructure, and the costs
of maintaining, updating, and securing an on-premises
system. They instead focus on analyzing data.
Natural integration point: By some estimates, as much as
80 percent of data you want to analyze comes from applica-
tions outside your companys data center. Bringing that data
together in the cloud is dramatically easier and cheaper than
building an internal data center.

Columnar storage, processing

As mentioned earlier, columnar storage significantly improves
the efficiency and performance of data storage, retrieval, and
analysis, enabling quicker access to results for system users.

Solid state drives (SSDs)

Unlike hard disk drives (HDDs), SSDs store data on flash memory
chips, which accelerate data storage, retrieval, and analysis. These
improvements augment the computing power of data warehouses
architected to use SSDs effectively.

CHAPTER 2 Why the Modern Data Warehouse Emerged 15

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 noSQL
noSQL, short for not only structured query language (SQL), describes
a technology that enables storing and analyzing newer forms
of data, such as data generated from machines and from social
media, to enrich and expand an organizations data analytics.
Traditional data warehouses dont accommodate these data types
very well. Therefore, newer systems have emerged in recent years
to handle these semi-structured data forms such as JSON,
Avro, and XML.

Some of these noSQL systems were designed with the intent to

replace traditional data warehouses but only ended up comple-
menting them. To get value from this semi-structured data,
organizations often have to extract and transform data from a
noSQL system and load it into a traditional data warehouse for
easy access by business users. As a result, this adds another layer
of complexity, and cost, for companies (such as Jana; see the
earlier sidebar) that attempt to capitalize on the benefits of both
types of systems.

Therefore, the modern cloud-based data warehouse must

incorporate, and optimize for, the ingestion and query of both
semi-structured and structured (traditional) data formats so
organizations avoid paying for and managing two systems.

16 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Choosing the right data warehouse
solution

Getting a high performance-to-price ratio

Making data security and protection a

priority

Chapter 3
The Criteria for Selecting a
Modern Data Warehouse

T
he trends discussed in Chapter 2 have led to a need and
anopportunity for a new kind of data warehouse. One built
for the volume, diversity, and velocity of todays data,
andfor the new ways organizations use their data. Such a solu-
tion must take advantage of key technology innovations, includ-
ing the cloud.

When youre in the market for a data warehouse, a checklist of

criteria will help determine which alternative best meets your
needs. Consider this chapter your checklist for finding the best
data warehouse solution for your organization.

Meets Current and Future Needs

True elasticity has its business benefits but theres more to that
story. You should be able to scale both compute and storage inde-
pendently so you are not forced to add more storage when you
really just need more compute, and vice versa. These are key
capabilities of an elastic data warehouse.

CHAPTER 3 The Criteria for Selecting a Modern Data Warehouse 17

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Accommodates and Integrates All Data
in One Place
Non-traditional, or semi-structured data, as discussed in pre-
vious chapters, can enrich the insight of data analytics beyond
the limits of traditional data. But this requires a new approach to
loading and transforming these new data types before an orga-
nization can analyze that data. Most traditional data warehouses
sacrifice performance or flexibility to handle these data types. A
modern data warehouse should eliminate the need to design and
model rigid, traditional structures upfront that would require
transforming semi-structured data before loading. It should also
optimize query performance against semi-structured data while
still in its native form. Overall, the data warehouse should support
diverse data with flexibility and avoid performance issues.

Efficiently loading all of your data into one location is crucial. But
integrating all of those diverse data types for more precise analyt-
ics is something else. A modern data warehouse should automati-
cally integrate your semi-structured data, once confined to noSQL
systems, with structured data inherent to a traditional, corporate
relational database. There should be nothing to install and con-
figure, with tuning and performance built in. Most importantly,
you shouldnt have to maintain and pay for two separate systems
to manage all of your data.

Supports Existing Skills, Tools, and Expertise

Traditional data warehouses are only outdated because the tech-
nology spans four decades and is not easily re-engineered for the
cloud. That also means the language they rely on, SQL, remains an
industry mainstay. Because of this, there is a broad array of mature
and emerging data management, data transformation, integration,
visualization, business intelligence, and analytics tools that com-
municate with a SQL data warehouse. The well-established role of
standard SQL also means a huge number of people have SQL skills.

Traditional data warehouses support SQL but dont support

the capabilities needed to effectively store and process semi-
structured data. Many organizations have therefore turned to
alternative approaches, such as noSQL solutions. The limitations
of these systems pose another problem. They require specialized

18 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 ANALYZING DISPARATE
DATA AT CHIME
Chime (chimecard.com) is smarter banking for the mobile generation.
Chime gathers and analyzes data across mobile, web, and back-end
server platforms to enhance its members experiences while deliver-
ing value to its business.

Analyzing key business metrics at Chime were laborious and involved

gathering and analyzing data from a large number of services, includ-
ing Facebook and Google ad services. Chime also pulled events
fromother, third-party analytics tools, most of which provided
semi-structured data such as JSON.

Chime satisfied the following requirements with its new, cloud data
warehouse:

Efficiently deliver structured and semi-structured data, and make

it available for query in near real time using standard SQL data-
base tables.
Simplify its data pipeline without the need to design a new model
for every new data type loaded into its data warehouse.
Scale up and down to meet workload demands and control costs.
Integrate quickly and painlessly with third-party, data analytics tools.
Enable SQL instead of other options that require complicated pro-
gramming languages to extract and analyze data.

Chimes analysts now model more scenarios to enhance member

services, spend less time waiting on query results, and spend more
time analyzing data.

knowledge and skills that arent broadly available and may not
support SQL. A modern data warehouse should be architected
with leading technology but built on inclusive and established
standards (such as SQL) compatible with skills and tools com-
monly available in the industry.

When youre in the market for a new data warehouse solution,

consider the skills and expertise required to manage and use the
solution. Hiring people with hard-to-find skills and retraining
others throughout your organization to deploy and maintain a
data warehouse can cost time and money.

CHAPTER 3 The Criteria for Selecting a Modern Data Warehouse 19

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Saves Your Organization Money
A conventional data warehouse can cost millions of dollars in:
licensing fees, hardware, and services; the time and expertise
required to set up, manage, deploy, and tune the warehouse; and
the costs to secure and back up data. In addition, building a data
warehouse that meets the business requirements and takes full
advantage of the volume and variety of todays data is often cost
prohibitive for any organization.

A modern data warehouse should meet these challenges at a much

lower price point. For example, does it scale storage and compute
separately so you only pay for the resources you need? Does it also
scale workloads and concurrency? Will it support diverse data struc-
tures and integrate diverse data in one place? And finally, can it do
all of this automatically without the complexity, expense, and head-
ache of manually tweaking and tuning the system to get the best
performance? (See Chapter5 for comparing cloud data warehouses.)

With cloud data warehousing, your service fee should cover

everything for a small fraction of the cost of a conventional, on-
premises solution. But not all cloud-based solutions are the same.
Their differences also determine how much a customer must pay,
in one way or another, to gain valuable data insight.

Provides Data Resiliency and Recovery

Many types of data warehouse failures can cause data loss or
inconsistencies. Therefore, your data warehouse must keep your
data safe, up to date, and available. Traditional data warehouses
typically protect data by performing periodic backups, which
consume valuable compute resources and interfere with ongo-
ing workloads. Periodic backups also require additional storage
and often fail to include the most recent data, resulting in data
inconsistencies.

A modern data warehouse should manage itself when it comes to

ensuring the durability, resiliency, and availability of the system.
It shouldnt interfere with any ongoing workloads, degrade
performance, or result in service unavailability due to backup
processes running in the background. And it should be cheap,
with clever ways to preserve your data without having to copy and
move it somewhere else.

20 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Secures Data at Rest and in Transit
Data security covers the following two main areas:

Confidentiality: Preventing unauthorized access to data

Integrity: Ensuring the data isnt modified or corrupted
A modern data warehouse must also support multilevel, role-
based access control (RBAC). This insures users only have access
to data they are permitted to see. For better security, require
multi-factor authentication (MFA). With MFA, as a user logs in, the
system sends a secondary verification request, often to a mobile
phone. The passcode sent to the phone must then be entered. This
insures that an unauthorized person with a stolen username and
password cannot access the system

Encrypting the data, which means applying an encryption algo-

rithm to translate the clear text into cipher text, is another required
security feature. A bigger part of the solution is key management.
Once you encrypt your data, youll use an encryption key to decrypt.
In addition to data, you have to protect the key that decodes your
data (see Figure3-1). How long do you use the same key? What hap-
pens if the key is compromised? All of this must be managed. The
data warehouse should use a hierarchical key wrapping approach,
which encrypts the encryption keys, as well as a robust key-rotation
process, which limits the time any single key is used.

Illustration provided by Snowflake.

FIGURE3-1: How to protect the keys that decode your data.

CHAPTER 3 The Criteria for Selecting a Modern Data Warehouse 21

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 In addition, the solution provider of a modern cloud data ware-
house must perform periodic security testing, known as penetra-
tion testing, to proactively check for vulnerabilities. The vendor
must administer these measures consistently and automatically
without impacting performance.

Choose a data warehouse with industry standard, end-to-end

security. Find a solution that has passed security audits like SOC 2.

Streamlining the data pipeline

The data pipeline refers primarily to the extract, transform, and load
(ETL) processes that import data into the warehouse and in a for-
mat that supports queries. A slow data pipeline forces users, such as
analysts, to spend too much time waiting to access data. The rapid
growth in diversity, number, and volume of non-relational data
sources streaming in from multiple sources compounds the problem.

A modern data warehouse should reduce the overall complexity of

the process to move data through the data pipeline faster. Modern
solutions should be able to efficiently load semi-structured data
in its native format and make it immediately available for query
without needing additional and intricate systems, such as noSQL,
to transform data. This allows users to immediately access data in
the same way they query a SQL database. Such solutions can pro-
vide access to new data exponentially faster, reducing the inges-
tion and transformation process from a day to less than an hour.

Optimizes Your Time to Value

Deploying a solution should not be a major undertaking, and cru-
cial aspects that were once manual should be automated. Most
of all, the solution you choose should be available all the time to
all users, and encompass all data types at a fraction of the cost of
traditional systems. Such a system should deliver immediate data
insight to help streamline an organization and increase its ability
to serve customers and lead its industry.

Every data warehouse solution requires time to implement. How-

ever, some solutions take considerably more time than others.
Think about how long your company will need to wait before it
can start to capitalize on the new system.

22 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Compressing the time-to-value gap

Slashing storage and compute costs

Taking advantage of elastic scaling

Outsourcing administration and security

Chapter 4
On-Premises versus
Cloud Data Warehousing

W
hen youre in the market for a new data warehouse, the
first choice to consider is where you want your data
warehouse located: your organizations data center or in
the cloud and provided as software-as-a-service. Traditional on-
premises data warehousing is a mature, well-established technol-
ogy designed well before cloud became a viable platform. With the
rapid adoption of cloud, theres a need for data warehouse solu-
tions that can take full advantage of what the cloud offers.

In this chapter, we present the key considerations for cloud data

warehousing as we compare it to traditional, on-premises systems.

Evaluating Time to Value

Deploying a conventional data warehouse (see Chapter 3) can
take at least a year and extend to a multi-year project before
you extract insight from your data. The agility of business today
means key stakeholders who support the project, and key busi-
ness and technical enablers responsible for the projects success,
may leave the team or the company before going live. Such a long
cycle also exposes the project to economic downturns, company

CHAPTER 4 On-Premises versus Cloud Data Warehousing 23

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 revenue shortfalls, and the risk of never implementing the project
due to scope creep.

In addition, on-premises solutions arent geared to handle

todays semi-structured data. That requires adding an open-
source, noSQL platform, which adds another layer of complexity.
This added burden lengthens the implementation phase of a new
data warehouse.

Done right, a cloud data warehouse can be up and running in less

than six months, or even in three months. Therefore, most of the
time required to get up and running should be extracting data
from your other data sources and configuring a front-end, ana-
lytics tool to extract insight from the data warehouse.

Accounting for Storage and

Computing Costs
On-premises data warehouses are expensive in terms of hard-
ware, software, and administration:

Hardware: When estimating hardware costs, consider the

costs of servers, additional storage devices, data center
space to house the hardware, a high-speed network to
access the data, and the power and redundant power
supplies needed to keep the system up and running. If your
warehouse is mission critical, add the costs to configure a
disaster recovery site.
Software (licensing): Organizations frequently pay hundreds
of thousands of dollars in software licensing fees for data
warehouse software and add-on packages. Additional end
users, including customers and suppliers, who are given
access to the data warehouse, can significantly increase those
costs. Then add the ongoing cost for annual support contracts,
which often comprise 20 percent of the original license cost.
Administration: An on-premises data warehouse needs
specialized, information technology (IT) personnel to deploy
and maintain the system. This creates a potential bottleneck
when issues arise and keeps responsibility for the system
with the customer, not the vendor.

24 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 A cloud data warehouse replaces the upfront and ongoing cost of
an on-premises system, with simple usage-based pricing. You
pay a monthly fee based on how much storage and computing
resources you use. Conservatively speaking, the annualized cost
for a cloud data warehouse solution can be one-tenth of a similar,
on-premises system.

Sizing, Balancing, and Tuning

For optimum performance, an on-premises data warehouse must
be sized, balanced, and tuned, which requires a significant up-
front investment along with ongoing monitoring and administra-
tion costs. Such a configuration often includes:

Number and speed of central processing units (CPUs)

Amount of memory
Number and size of disks for required storage capacity
Input/output (I/O) bandwidth (a measure of how much data
can be transferred at a given time)

With an on-premises data warehouse, companies often size their

system for peak usage, which may represent only a small period
of the year. For example, a company may need the full power of
the data warehouse only at the end of each financial quarter or
year. But they must pay for that peak capacity 24 hours a day
because the system cant easily scale up or down.

Elastic cloud data warehousing delivers two key advantages:

The complexities and cost of capacity planning and

dministration sizing, balancing, and tuning the system
a
should be built into the system, automated, and covered by
the cost of your subscription.
The same goes for dynamically provisioning storage and
compute resources on the fly to meet the demands of your
changing workloads in peak and steady usage periods.
Capacity is whatever you need whenever you need it.

CHAPTER 4 On-Premises versus Cloud Data Warehousing 25

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Considering Data Preparation
and ETL Costs
An on-premises data warehouse must extract data from all of
your data sources. Then it must transform that data to adhere to
the often rigid data structure inside the system before loading it
into the warehouse. A key challenge includes adhering to a finite
and expensive amount of processing capacity and storage. As a
result, data transformation must happen outside normal business
hours to avoid competing with other data processing jobs. This
is expensive. In addition, semi-structured data doesnt arrive in
neatly organized and consistent rows and columns inherent to
traditional data structures. The data is also high-volume, high-
velocity data.

The best cloud-based solutions can load semi-structured data

directly without transforming it. These solutions can provide
access to fresh data up to fifty times faster than a traditional data
warehouse. In addition, the lower cost of unlimited, cloud storage
provides data analysts access to all of the data instead of being
limited to periodic aggregates of that data.

Adding the Cost of Specialized Business

Analytic Tools
As mentioned in Chapter 3, traditional, on-premises data
warehouses arent geared to handle the volume, variety, and
velocity of todays data. As a result, organizations operate two data
platforms: an on-premises, enterprise SQL data warehouse for
storage of traditional relational data, and a noSQL big data plat-
form, which can run on-premises or in the cloud, for storing non-
relational data.

Unfortunately, these newer systems require specialized tools and

expertise, which dont compare to the prevalence of SQL tools and
expertise. After all, SQL has been around for decades, while noSQL
systems are relative newcomers.

The ideal cloud data warehousing solution delivers the best of

both worlds the flexibility to integrate relational and non-
relational data along with support for the readily available SQL
tools and skills for querying that data.

26 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 OPTIMIZING DOUBLEDOWNS
DATAPIPELINE
DoubleDown, an online gaming studio, added a noSQL system to
their data pipeline to prepare data for loading into their data ware-
house. But this approach meant DoubleDowns daily event log (user
clicks and other data generated by gamers activities) required long
processing times. The company couldnt access the previous days
data until 3pm the next day. Even worse, if one of its data computing
clusters went down, the company actually lost data.

DoubleDown chose a system that could directly load its semi-structured

data without transforming it first, making that data immediately avail-
able for queries. This improved the quality and performance of their
data pipeline by:

Getting data to analysts nearly 100 times faster 15 minutes

versus 24 hours.
Eliminating nearly all of the frequent failures in their previous
pipeline.
Providing analysts full data granularity instead of periodic
aggregates.
Reducing DoubleDowns cost of their data pipeline by 80 percent.
DoubleDown analysts now have immediate access to data from
newproduct releases to enable the business to make far faster,
data-driven decisions.

When youre in the market for a new data warehouse, consider the
cost and availability of the skills and expertise required to manage
the data warehouse, and the many analytics and other tools used
in conjunction with a data warehouse.

Making Allowances for Scaling and Elasticity

Conventional data warehouses are prone to system slowdowns
and crashes as users and processes compete for limited resources.
These systems tightly connect storage and compute onto a sin-
gle computer cluster (a group of computers), making it costly to
increase one without increasing the other.

CHAPTER 4 On-Premises versus Cloud Data Warehousing 27

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Newer, cloud-built data warehouse solutions provide virtually
unlimited storage and compute; however, consider a data ware-
house that scales storage separate from compute (see Figure4-1).
Ideally, the cloud data warehouse should scale in three ways:

Storage: Cloud storage is inherently scalable, easily adjust-

ing the amount of storage to meet changing needs.
Compute: The resources used for processing data loads and
queries should easily scale up or down, at any time, as the
number and intensity of the workloads change.
Users and workloads (concurrency): Solutions with fixed
computing resources slow as users and workloads increase.
Organizations are often forced to replicate data into separate
data marts, shift some workloads outside of normal business
hours, and queue users to preserve performance. Only the
cloud can enable a data warehouse to scale out by distribut-
ing virtual data warehouses, or clones, of the data across
separate compute clusters.

Look for a cloud data warehousing solution that decouples storage

from compute so both can scale easily and independent of each
other to keep costs low. The solution should also scale out, or
horizontally, to support more users and workloads without nega-
tively impacting performance.

Illustration provided by Snowflake.

FIGURE4-1: The ideal data warehouse should scale in three ways.

28 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Deliberating Over Delays andDowntime
Many companies with on-premises solutions have two main
complaints. They must wait hours or more than a day before data
collected the previous day is in the warehouse and available. They
must wait the same time for a complex query to run on a large
data set. In some cases, multiple, concurrent processes can freeze
or crash the system, extending delays and downtime.

With virtually unlimited storage and compute resources, cloud

data warehouse solutions, architected as dynamically elastic, are
better equipped to scale up, and out, to meet increased demands.
However, decreasing delays and eliminating unplanned down-
time requires more than simply ramping up system resources.
Better solutions streamline the data pipeline and store data to
make queries run more efficiently without manual tuning.

As you evaluate your options, look for solutions that address all
these types of performance issues. How quickly you can access
your data and analytics can significantly impact your operations
and your ability to maintain a competitive edge.

Considering the Costs of Security and

Potential Security Breaches
A single breach can quickly turn into a public relations nightmare
and result in lost business, and steep fines from regulatory agen-
cies. Although the cloud attracts the fear of security risks, it can
be more secure than your data center.

If you opt for an on-premises data warehouse, youre solely

responsible for securing sensitive data, which involves careful
and constant attention to the following details:

Firewall protection
Security protocols
Data encryption, at rest and in transit
User roles and privileges
Monitoring and adapting to emerging security threats

CHAPTER 4 On-Premises versus Cloud Data Warehousing 29

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Effective data security is complex and costly to implement, espe-
cially in terms of human resources. Poorly implemented security
measures expose you to even more costs if breached.

Because cloud data warehousing providers serve a number of

customers, they can afford the expertise and resources to pro-
vide industrial-strength, end-to-end data warehouse security.
Look for a provider that ensures industry-standard, end-to-end
encryption to secure data both at rest and in transit.

Paying a Premium for Data Protection

and Recovery
On-premises data warehouses are vulnerable to data loss from
equipment failure, power outages or surges, theft or vandalism,
and disasters (fire, flood, earthquake, and so on). To protect your
data, you must back it up regularly and store backups at a remote
location. A backup power supply is also necessary to prevent data
loss and ensure that your data warehouse is always available to
process incoming data and queries. If disaster does strike, youll
need skilled personnel in place to recover data using the most
recent backups. If your data warehouse is mission critical, you
may also need a geographically separated disaster recovery site
(an additional data center) along with the software, licenses, and
processes to insure automatic failover so there is no gap in service.

The cloud provides an ideal solution for data protection and recov-
ery. By its nature, it stores data off premises. Some, but certainly not
all, cloud-based solutions automatically back up data to two or more
separate physical locations. If the data centers are geographically
isolated, then they also provide built-in disaster recovery. Cloud data
centers are equipped with redundant power supplies so they remain
up and running even during lengthy power outages. Cloud data
warehousing providers can deliver these protections at a much lower
cost than you by distributing the cost over thousands of clients.

If you do not want to administer your own data backups be sure

to ask your potential cloud data warehouse provider how they
configure their service. Likewise, if you need disaster recovery
protection, be sure to confirm the providers architecture uses
geographically separated centers.

30 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Evaluating differences between cloud
data warehouse options

Considering factors that affect

performance

Choosing a solution that ensures data

protection and security

Gauging your savings in administrative

costs

Chapter 5
Comparing Cloud Data
Warehouse Solutions

T
he growing adoption of cloud has caused legacy on-premises
vendors and recent market entrants to offer cloud versions
of their data warehouse products. Like any product or
service, no two cloud data warehouse solutions are the same. In
this chapter, we explain some of the differences and what to look
for among cloud data warehouse solutions.

Understanding Approaches to Data

Warehousing in the Cloud
Technology vendors all provide access to a data warehouse
deployment via the Internet and typically operate on a subscrip-
tion or usage-based model. But the following cloud approaches
offer significantly different product capabilities:

Infrastructure-as-a-service (IaaS): This approach requires

the customer to install traditional data warehouse software
on computers provided by the cloud platform provider. The
customer manages all aspects of the cloud hardware and
data warehouse software. In addition, the capabilities of the

CHAPTER 5 Comparing Cloud Data Warehouse Solutions 31

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 data warehouse are identical to the same software deployed
using on-premises hardware.
Platform-as-a-service (PaaS): With this hybrid approach,
the data warehouse vendor provides the hardware and
software as a cloud service. The vendor manages the
hardware deployment, software installation, and software
configuration. However, the customer manages, tunes, and
optimizes the data warehouse software.
Software-as-a-service (SaaS): With the SaaS approach, the
data warehouse vendor provides all hardware and software
as part of its service, including all aspects of managing the
hardware and software. Typically included in the service are
software and hardware upgrades, security, availability, data
protection, and optimization.

With all of these scenarios, the task of purchasing, deploying, and

configuring the data center space, and the hardware to support
the data warehouse, transfers from the customer to vendor.
Beyond that advantage, the benefits and drawbacks of the differ-
ent offerings can vary from ease of use to security and availability.
This chapter discusses some of the key differences among various
cloud data warehouse offerings.

If a data warehouse provider merely supplies access to its tra-

ditional data warehouse via the cloud, the solution is likely to
resemble its original, on-premises architecture and funtionality.

Comparing Architectures
Many vendors offer a cloud data warehouse originally designed
and deployed for on-premises environments. These traditional
architectures were created long before the cloud and its benefits
emerged as a viable option. Alternatively, any data warehouse
solution built for the cloud should capitalize on the benefits of the
cloud (See Figure5-1).

To identify a solution built on a cloud-optimized architecture,

look for the following characteristics:

Centralized data storage

Independent scaling of compute and storage resources

32 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Illustration provided by Snowflake.
FIGURE5-1: How a cloud-optimized architecture streamlines performance.

Increased concurrency without competing for resources

Load and query data simultaneously without degrading
performance
A robust metadata service that applies across the entire
system. (Metadata is data about other data such as file size,
the author of the data, and when it was created.)

A cloud-optimized architecture also takes advantage of storage-

as-a-service, where data storage expands and contracts
automatically and transparently to the user. Data storage
designed for older architectures remains expensive and has
limited scalability.

Assessing a Solutions Ability toManage

Data Diversity
A key factor driving the adoption of cloud data warehousing stems
from the growing volume of data that originates in the cloud
outside a companys data center. In most cases, this non-relational
data must be transformed before loaded into a traditional data
warehouse on-premises or in the cloud. This approach adds
significant complexity and delays to accessing new data.

With this greater volume and variety of data, the cloud has
become a natural integration point. An ideal way to address this
issue is with a cloud data warehouse that can handle both rela-
tional and non-relational data, and without having to transform

CHAPTER 5 Comparing Cloud Data Warehouse Solutions 33

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 the non-relational data or compromise performance during the
data loading or subsequent query process.

Data must be transformed before its loaded into a traditional,

cloud-based warehouse. Or the organization must buy and main-
tain an additional system to handle non-relational data.

Gauging the Scaling and Elasticity of

Different Options
Not all cloud data warehouses feature the same type of elasticity.
Advanced cloud data warehouse solutions can scale up and down,
on the fly, and without taking the system offline or putting it into
a read-only mode.

Consider the drawbacks of solutions that dont scale well:

A cloud data warehouse that requires manual reconfigura-

tion involves careful planning and coordination with the
vendor to scale resources.
Scaling may require downtime or a switch to read-only mode
to redistribute data and reconfigure the system.
Most cloud data warehouse offerings bundle compute and
storage on the same node, requiring customers to scale both
when they need to increase just one or the other.

Comparing Concurrency Capabilities

Concurrency is the ability to perform two or more tasks simulta-
neously or allow two or more users access to a computing solu-
tion. In a traditional data warehouse, fixed compute and storage
resources limit concurrency. With cloud, however, compute and
storage are not fixed. Cloud-optimized architectures support
concurrency in the following two ways:

Dynamic distribution of data and additional compute

resources makes it possible for multiple users to query the
same data simultaneously without degrading performance.

34 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Loading and querying can happen concurrently. The cloud
data warehouse solution should enable separate compute
resources (nodes or even clusters) for different workloads.
This approach enables simultaneous loading and querying of
data without contention by assigning the conflicting pro-
cesses to independent compute clusters.

Ensuring SQL Support

Almost all business intelligence (BI); extract, transform, and load
(ETL); and data analytics tools can communicate with a data ware-
house that supports standard SQL.However, not all cloud data ware-
housing solutions fully support standard SQL.For example, big data
solutions positioned as cloud data warehouses are often noSQL
solutions and have only incomplete or non-standard SQL support.

Checking Backup/Recovery Support

With on-premises and many cloud data warehousing solutions,
customers must protect their own data with backup and data
replication tools. However, some cloud data warehouse solutions
include data protection as a part of the service.

For optimum protection, look for a solution that automatically

saves past versions of data or automatically duplicates data for
use as an online backup. The solution should also allow for self-
service recovery of lost or corrupted data.

Confirming Resiliency andAvailability

Resiliency is the ability of the data warehouse to continue to func-
tion automatically in the midst of component, network, or even
data center failure. Availability is the ability users have to access
the system at all times (often referred to as uptime). Cloud data
warehouse services vary as to how much the customer is respon-
sible for availability and resiliency. At the most basic level, a cloud
data warehouse service may require the customer to handle sys-
tem monitoring to detect and possibly prevent a failure. The cus-
tomer may also have to administer data replication so a duplicate

CHAPTER 5 Comparing Cloud Data Warehouse Solutions 35

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 copy of the data warehouse is available in case of a failure. At the
other end of the spectrum, the vendor provides monitoring, rep-
lication and automatic failover as part of the service.

Availability is also a factor for software upgrades. Different ven-

dors take different approaches during the upgrade:

Basic: Customers manage upgrades and related downtime.

Better: The vendor manages upgrades and informs users of
upcoming upgrades so they can plan for the downtime.
Best: The vendor provides transparent upgrades without
involving users or subjecting them to any downtime.

Look for how many 9s of availability the cloud data warehouse

solution supports (99.9XX% uptime).

Gauging Performance
One of the great promises of the cloud is the ability to have huge
amounts of resources available that you can pay for only when
you need them. Imagine renting a Ferrari for that one impor-
tant occasion but using a fuel-efficient hybrid for your everyday
drive to work. Look for a cloud data warehouse solution that can
optimize performance on demand and eliminates administrative
effort to incorporate new resources.

Steer clear of data warehouses that disrupt or delay activity to add

or subtract resources. Some solutions also require administrative
work, including redistributing data and recalculating metadata
such as statistics.

Evaluating Cloud Data Security

The cloud is often perceived as less secure than on-premises data
storage. However, cloud solutions have gained increasing accep-
tance due to the continuing barrage of break-ins into secure,
on-premises data centers. These incidents reveal that companies
are limited in their ability to secure their own data.

All cloud data warehousing offerings shift the responsibility for

physical data center security to the cloud solution vendor. But be
aware that security features vary among vendors:

36 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 The most basic cloud data warehouse offerings provide only
security capabilities, leaving things like encryption, access
control, and security monitoring to the customer.
Other solutions offer features such as encryption and access
controls, which customers can choose to turn on, but they
leave the system vulnerable if not enabled.
Cloud data warehouse offerings that are more service-
orientated incorporate features for security and also assume
the burden of security management by providing encryption,
encryption key management, key rotation, intrusion detec-
tion, and more, as part of the service.

Accounting for Administration

Traditional data warehouses require a significant amount of
the customers time, effort, and expertise. One or more data-
base administrators (DBAs) must perform software patches and
upgrades, data partitioning and repartitioning, index manage-
ment, workload management, statistics updates, security man-
agement and monitoring, backups and replication, query tuning
and rewriting, and more.

At a base level, a cloud data warehouse solution thats built on

older, on-premises technology still requires the customer to
manage all of these aspects. Newer data warehousing offerings
reduce or eliminate much of this management overhead through
new designs and automation.

Ensuring the Ability to IsolateWorkloads

A key factor in the speed and performance of a data warehouse
is its ability to isolate workloads. Without this ability, workloads
compete for the same limited resources.

With traditional on-premises systems, the only way to isolate

workloads is to schedule them to run at different times. This
assumes the load can finish during those off hours and also
assumes no one starts a report that continues to run into the wee
hours of the night. In this way, there is no contention for the CPU
and memory of the system. Otherwise, the only real option is to
buy additional memory or CPUs, or both.

CHAPTER 5 Comparing Cloud Data Warehouse Solutions 37

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 A cloud data warehouse built on an older architecture is likely to
behave the same. To be effective, the cloud data warehouse should
easily configure multiple pools of compute resources (of varying
sizes) to separate the workloads of users and processes that need to
run concurrently. This approach eliminates contention and provides
resources sized to each workload. Ideally, these separate workloads
should access the same data simultaneously, and turn on and off
easily, based on need.

Enabling All Use Cases: Cloning

and Time Travel
In traditional environments, different data systems handle dif-
ferent use cases a data warehouse for operational reporting,
data marts for departmental reporting and analytics, data lakes
for data exploration, and specialized tools for activities such as
predictive analytics. Each of these require hardware, a copy of
data, individual management, and so on.

To bring these diverse use cases together in the cloud, a data

warehouse should support fast and efficient ways to clone data to
avoid the traditional and costly forms of duplicating data.

A cloud data warehouse should also make it easy to recover from

errors or problems created by data transformation jobs with fea-
tures such as time travel, which enable simple access and rollback
to previous versions of data.

38 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 IN THIS CHAPTER
Listing your data warehouse needs and
success criteria

Considering all factors in the total cost of

ownership

Taking your data warehouse for a test

drive before you buy

Chapter 6
Six Steps to Getting
Started with Cloud Data
Warehousing

N
ow that you know data warehousing basics, the differences
between on-premises and cloud data warehousing, and the
differences between various cloud data warehouses, youre
probably wondering how to apply your newly acquired knowledge
to choose the right cloud data warehouse for your organization.

When choosing a cloud data warehouse, you want to be sure to

check off the following six steps:

1. Evaluate your data warehousing needs.

2. Migrate or start fresh.
3. Establish success criteria.
4. Evaluate cloud data warehouse solutions.
5. Calculate your total cost of ownership.
6. Set up a proof of concept (POC).

CHAPTER 6 Six Steps to Getting Started with Cloud Data Warehousing 39

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 In this chapter, we guide you through these six key steps to
choosing a cloud data warehouse. It starts with evaluating your
data warehouse needs and ends with the process of testing your
top choice. By the end of this chapter, youll have a plan to help
you choose your next solution with confidence.

Step 1: Evaluate Your Data

Warehousing Needs
The data warehouse thats right for you should meet your cur-
rent needs and be flexible enough to accommodate your future
needs. Therefore, consider the nature of the data you work with,
the skills and tools already in place, your usage needs, the future
plans for your business, and how a data warehouse can take your
business further than you imagined:

Data: What types of data must the data warehouse contain?

At what rate is new data created? How often will data move
into the warehouse? What crucial data cant you access
today?
Fits with existing skills, tools, and processes: What tools
need to connect to the data warehouse? What skills from
your current team are applicable to the various cloud data
warehousing options? What processes will a cloud data
warehouse impact?
Usage: Which users and applications will access the data
warehouse? What types of queries will you run? How much
data will users need to access, and how quickly? How will
workloads vary over time? What performance requirements
do your users and applications require? Ideally, how many
users should access the data warehouse but dont today due
to resource constraints?
Resources: What human resources are available to manage
the data warehouse? How much investment do you wish to
make to monitor and manage availability, performance, and
security?

40 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Step 2: Migrate or Start Fresh
Every cloud data warehouse project should start with assessing
how much of your existing environment should migrate to the
new system, and what should be built new for a cloud data ware-
house. These decisions may address everything from design of
the extract, transform, and load (ETL) processes to data models
and software development lifecycle methods. Here are a few key
considerations:

Is this a brand new project? If so, it often makes sense to

design the project to take full advantage of the capabilities of
a cloud data warehouse rather than carry forward an
implementation with significant constraints.
How satisfied are users with your current implementa-
tion? High levels of satisfaction may indicate its best to
maintain many of the existing processes and approaches
from those deployments. Low levels of satisfaction/adoption
indicate change is needed.
What aspects of current deployments were designed to
accommodate constraints that are no longer present
with a cloud data warehouse? Tools and processes
designed to work around resource constraints, avoid the
disruptive effort required to add capacity, or to optimize cost
may be unnecessary for a cloud solution.
How do current users and applications access the data
warehouse? Users and applications that primarily rely on
industry standard interfaces, such as SQL, and use standard
ETL and BI tools, will experience significantly less change to
adapt to a new approach.
How are your data and analytics requirements likely to
change in the future? A solution built to evolve is less likely
to be replaced sooner than expected.

If you have a large and complex traditional data warehouse envi-

ronment, you may want to migrate a small part of the system to
get comfortable with using a cloud data warehouse. Then you can
iteratively expand your cloud footprint.

CHAPTER 6 Six Steps to Getting Started with Cloud Data Warehousing 41

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Step 3: Establish Success Criteria
How will you measure the success of moving to a new cloud data
warehouse? Choose important business and technical require-
ments. Criteria should focus on the performance, concurrency,
simplicity, and total cost of ownership.

RESOLVING LATENCY ISSUES

ATWHITE OPS
White Ops is a leading provider of cyber security services. Unlike tradi-
tional approaches that employ statistical analysis, White Ops combats
criminal activity by differentiating between robotic and human inter-
action, working to uncover and characterize new fraud patterns. This
constant process requires storing and processing massive amounts
ofdata.

White Ops had previously relied on noSQL systems to store and pro-
cess that data. However, that approach required a developer to build
a custom query. The latency for results was at least 24 hours, depend-
ing on the workload. The more requests, the longer the delays.

To simplify and standardize the process to increase productivity and

performance, White Ops implemented a cloud data warehouse with
SQL as its core language and delivered as a service. Analysts can now
access data directly and without a developer. The data warehouse
enables White Ops to:

Have all data in one place

Scale elastically
Query diverse data with standard SQL
Accelerate the pace of evolution of its fraud prevention offerings
White Ops can now consolidate and scale massive amounts of data,
enable access to the data without waiting for specialists with deep
programming skills, and help its customers avoid the potentially dev-
astating effects of online fraud.

42 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 If your new cloud data warehouse has capabilities that werent
available in your previous system, and those capabilities are rele-
vant to evaluating the business and technical success of your new
solution, be sure to include them.

As you establish the success criteria of your new solution, deter-

mine how youll measure that success:

Determine which criteria are quantifiable and which are

assessed qualitatively.
Determine how youll measure the quantifiable criteria.
Figure out how youll assess the qualitative criteria.

Step 4: Evaluate Cloud Data Warehouse

Solutions
Once you determine your data warehouse needs and success cri-
teria, youre ready to start evaluating solutions. Throughout this
book weve described the differences between available options in
detail (see Chapters3,4, and5). As you compare solutions, make
sure they meet the following criteria:

Addresses current and future needs

Integrates structured and semi-structured data, stores it all
in one place, and avoids the need to create data silos
Supports existing skills, tools, and expertise
Guards against data loss and enables you to easily recover
lost data
Secures your data with industry-standard password
protection and encryption
Ensures data and analytics are always available
Streamlines the data pipeline so that new data is available
for analysis in the shortest possible amount of time
Optimizes time to value, so you can begin to reap the
benefits of your new data warehouse as soon as possible

CHAPTER 6 Six Steps to Getting Started with Cloud Data Warehousing 43

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Isolates workloads, so workloads do not compete for limited
resources
Scales compute and storage independently and automati-
cally, and scales concurrency without impacting performance

Step 5: Calculate Total Cost of Ownership

If your choice of a cloud data warehouse solution comes down
to price, consider the total cost of ownership. For a conventional
data warehouse, total cost of ownership includes the cost of each
of the following:

Licensing, typically based on the number of users

Hardware servers, storage devices, networking, and so on
Office space to house the data center
Electricity to run the hardware and keep it cool
Administration and maintenance
Data security password protection and encryption
Ongoing management of the data warehouse
Solutions to ensure availability and resiliency
Support for scaling and concurrency
Creation of development and staging environments
For some solutions, you may need to consider additional costs,
such as building and managing multiple data marts, having mul-
tiple copies of data in different data marts, training people, hav-
ing multiple systems (for example, SQL and noSQL) to handle
diverse data, and so on.

Calculating the costs of cloud data warehouse options is usually

easier, but it varies according to the vendors services. Assum-
ing you outsource everything to the vendor by choosing a data-
warehouse-as-a-service (DWaaS), you can calculate the total
cost of ownership based on the monthly subscription fee. If you

44 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 opt for an infrastructure-as-a-service (IaaS) or platform-as-a-
service (PaaS) solution (see Chapter5), you need to add the costs
of whatever software, administration, and services the solution
doesnt include.

Companies typically calculate the cost ownership over the

expected lifetime of the data warehouse, which is commonly one
to three years. A key caveat is that in their calculations people
often assume that a cloud system runs 24/7, which is not the case.
They overlook the savings possible when a cloud solution is scaled
up and down dynamically in response to changing demand.

Step 6: Set Up a Proof of Concept (POC)

After investigating different cloud data warehouse options,
viewing demos, asking questions, and meeting with each ven-
dors team, you should do a proof of concept (POC) before decid-
ing which to choose. A POC is a process of testing a solution to
determine how well it serves your needs and meets your success
criteria. Think of it as a test drive. It typically lasts a day or two,
but it can be conducted over the course of several weeks. You
request a POC from a prospective vendor with the general under-
standing that if the solution performs satisfactorily, youll buy
the product. Or, in the case of cloud data warehousing, subscribe
to the service.

This is also your chance to think outside the box. Consider what
else you could do above and beyond what you do today. If you had
this cloud data warehouse solution in place, what additional busi-
ness value could this system deliver?

When setting up your POC, list all requirements and success cri-
teria, not just the issues youre trying to resolve. For example, if
your primary complaint about your current data warehouse is that
queries take too long to run, dont focus solely on that issue. Your
POC should cover everything, including ease of migrating your
data to the new warehouse, loading new structured and semi-
structured data, running queries and multiple workloads, and
using your existing business intelligence tools.

CHAPTER 6 Six Steps to Getting Started with Cloud Data Warehousing 45

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
 Develop a comprehensive check list. Use your list of data ware-
housing needs and your criteria for success as a starting point.
But dont overlook positive qualities of your existing data ware-
house that are non-issues for you now. In other words, make
sure the new data warehouse can do everything your current data
warehouse does but better, and that it overcomes the drawbacks
of your current warehouse.

If you do a POC with multiple vendors, try to use the same check
list for each.

46 Cloud Data Warehousing For Dummies, Snowflake Special Edition

These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
These materials are 2017 John Wiley & Sons, Inc. Any dissemination, distribution, or unauthorized use is strictly prohibited.
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wileys ebook EULA.