CLOUD PLATFORMS

Zulqar Nain
Department of AI and Big Data
Woosong University

Date: 2024-04-02
COURSE CONTENT
 https://
eddiejackson.net/web_docume
nts/The_Definitive_Guide_to_Cl
oud_Computing.pdf
 Aws documentation

Cloud Platforms
Department of AI and Big Data
2
 RECAP
 Cloud computing is relatively young, but in the short
time since its inception, it has managed to create a
host of competing definitions, architectures, service
models, and delivery methods.
 Across all of these varying ways of looking at cloud
computing, we find common characteristics, including
massive scalability, ease of allocating resources, and
a service management platform.
 Building on this foundation, cloud providers can
deliver a range of services, from infrastructure to
platforms to applications and business services.
 No single delivery model meets all needs, but the
combination of public, private, and hybrid clouds offer
a range of options suitable for many business
requirements. Cloud Platforms
Department of AI and Big Data
3
 BUSINESS INNOVATIONS ENABLED BY
CLOUD
 Many discussions of cloud computing focus on its
technological advantages—and there are many—but
there are business advantages as well
 This lecture shifts focus from questions of architecture
and operations to issues of service delivery and return
on investment (ROI).
 We will compares service delivery under traditional IT
service models and under cloud computing. Example
scenarios will illustrate some of the key differences.
 Advantages of doing business with cloud computing—
The advantages of doing business with cloud
computing include the reduced time required to deliver
new services, new means to control costs, the ability to
scale to demand, and the adaptability of cloud
computing.
Cloud Platforms
Department of AI and Big Data
4
 BUSINESS INNOVATIONS ENABLED BY
CLOUD
 ROI in cloud computing comes from both reduced
capital costs and lower operational costs.
 As with other technologies, the ROI in the cloud is
highly dependent on more than just the technology;
how you implement and manage cloud services
contributes to how much of the potential ROI is actually
realized.
 As a first step to understanding the source of ROI in
cloud computing, let’s consider a couple of
hypothetical examples of how service delivery in the
cloud differs from traditional IT service delivery.

Cloud Platforms
Department of AI and Big Data
5
 LAUNCHING A NEW BUSINESS SERVICE
 The computing, storage, and network services
required to support the service
 Software that captures the functional
requirements of the new service while providing a
usable interface.
 A well‐developed plan for deploying elements in

the proper order so that dependencies are in place

as new components are put in place
 Policies and procedures to govern how the service

infrastructure is managed and maintained

 A recovery strategy and corresponding systems to

mitigate the risk of data loss or service delivery

failure
Cloud Platforms
Department of AI and Big Data
6
LAUNCHING A NEW BUSINESS SERVICE

Cloud Platforms
Department of AI and Big Data
7
 LAUNCHING A NEW BUSINESS SERVICE
 To better understand how service models influence
service delivery, let’s assess delivering a couple of
different types of services under different models.
 In the first example, we will consider a home

improvement retailer with a plan to offer tutorial

videos on home improvement projects for the do‐
it‐yourself (DIY) customer.
 In our second example, we will see how business

analysts deal with the problem of “big data” and

the need for advanced business intelligence and
analytics services.

Cloud Platforms
Department of AI and Big Data
8
 LAUNCHING A NEW BUSINESS SERVICE
 They are significantly different types of services—
one is a customer‐facing Web application and the
other is a more batch‐oriented back office service.
 They require a different combination of computing

resources
 They have different usage patterns over time

 Cloud computing can reduce the cost of delivery of

both services regardless of the differences in the

type of application an demand profile

Cloud Platforms
Department of AI and Big Data
9
 UNDER A TRADITIONAL IT SERVICE MODEL :
TUTORIAL VIDEOS FOR THE DIY CUSTOMER
 Not all of us are gifted carpenters or skilled
plumbers, but some of us think we could do a
fairly decent job around the house if we just had
the right tools and a few tips to get us started.
 A home improvement retailer that has traditionally

done well serving the small contractor segment of

the market has decided to target the potential DIY
customer in an effort to improve sales and expand
their share of that market segment.
 The following list highlights key features and non‐

functional requirements:

Cloud Platforms
Department of AI and Big Data
10
 UNDER A TRADITIONAL IT SERVICE MODEL :
TUTORIAL VIDEOS FOR THE DIY CUSTOMER
 The service will provide short tutorial videos on a
range of home improvement topics. Videos will
range from 1 to 10 minutes in duration with an
average of 5 minutes.
 Videos will be streamed over the Web and

delivered through the company’s Web site.

 The service will be launched in beta to customers

in the Northeast United States for 4 weeks

followed by an extended 4‐week beta to the
Northeast, Mid‐Atlantic, and Southeast United
States.
 After that, it will be made available throughout the

company’s North American market.

Cloud Platforms
Department of AI and Big Data
11
 UNDER A TRADITIONAL IT SERVICE MODEL :
TUTORIAL VIDEOS FOR THE DIY CUSTOMER
 The initial launch will support up to 500 videos; at
the end of the beta testing phase, 1000 videos will
be available. Content will grow at an average rate
of 200 videos per month after that.
 Metadata will be assigned to each video to

improve search and browsing. Tags will include

structured data, such as repair type, tools
required, and time to complete the task.
Unstructured data describing the video content is
also included.
 Videos will be accessible through a centralized

“How‐to Video Library” in the Web site as well as

through product pages that link to relevant videos.
Cloud Platforms
Department of AI and Big Data
12
 UNDER A TRADITIONAL IT SERVICE MODEL :
TUTORIAL VIDEOS FOR THE DIY CUSTOMER
 Customers will be encouraged to review and rate
videos. The results will be analyzed to improve the
overall quality of instruction, expand the scope of
topics, and eliminate the least‐useful content

Cloud Platforms
Department of AI and Big Data
13
UNDER A TRADITIONAL IT SERVICE MODEL :
TUTORIAL VIDEOS FOR THE DIY CUSTOMER

Cloud Platforms
Department of AI and Big Data
14
UNDER A TRADITIONAL IT SERVICE MODEL :
ADVANCED ANALYTICS FOR AUTO INSUR-
ANCE PREMIUM CALCULATIONS

Cloud Platforms
Department of AI and Big Data
15
 UNDER A TRADITIONAL IT SERVICE MODEL :
ADVANCED ANALYTICS FOR AUTO INSUR-
ANCE PREMIUM CALCULATIONS
 Existing data sets on age and sex of the driver,
past accidents, number of moving violations,
primary garaging location of the vehicle, and so on
must be available for data mining.
 Additional data on household income, including

income by age, disposable income, household net

worth, disposable income, and so on; consumer
spending data by category, such as financial
services, automotive, medical, recreation, and so
on; business activity data by location; and
publically available data, including census data
and crime statistics

Cloud Platforms
Department of AI and Big Data
16
 UNDER A TRADITIONAL IT SERVICE MODEL :
ADVANCED ANALYTICS FOR AUTO INSUR-
ANCE PREMIUM CALCULATIONS
 On a monthly basis, internal and external data will
be collected and analyzed to build a predictive
model that categorizes each customer by fine‐
grained risk estimate.
 New extraction, transformation, and load (ETL)

procedures will be developed to collect data from

multiple sources and copy it to project storage;
data will not be stored once the model is
constructed.
 To improve the quality of predictions, multiple

prediction models will be constructed and results

will be combined to make final classifications.

Cloud Platforms
Department of AI and Big Data
17
UNDER A TRADITIONAL IT SERVICE MODEL :
ADVANCED ANALYTICS FOR AUTO INSUR-
ANCE PREMIUM CALCULATIONS

Cloud Platforms
Department of AI and Big Data
18
 UNDER A TRADITIONAL IT SERVICE MODEL :
ADVANCED ANALYTICS FOR AUTO INSUR-
ANCE PREMIUM CALCULATIONS
 On a monthly basis, internal and external data will
be collected and analyzed to build a predictive
model that categorizes each customer by fine‐
grained risk estimate.
 New extraction, transformation, and load (ETL)

procedures will be developed to collect data from

multiple sources and copy it to project storage;
data will not be stored once the model is
constructed.
 To improve the quality of predictions, multiple

prediction models will be constructed and results

will be combined to make final classifications.

Cloud Platforms
Department of AI and Big Data
19
 TUTORIAL VIDEOS IN CLOUD
 Initially, the systems architects decide to allocate
two virtual servers for the beta‐test period;
however, if demand warrants additional or fewer
servers, systems administrators will adjust as
needed.
 Planning for long‐term storage is not a significant

issue because additional storage will be allocated

as needed.
 There is no need to purchase peak‐load storage.

 As the project moves from the beta testing stage

to full production, the systems administrators will

add virtual servers as needed.
 Systems administrators can focus on immediate

demand and server Cloudallocation

Platforms to efficiently and
Department of AI and Big Data
20
 ADVANCED ANALYTICS IN CLOUD
 Creating virtual images with the required software,
such as ETL systems, and preprocessing scripts
and statistical and data mining packages.
 Instantiating servers to run parts of the workflow

as needed; for example, based on the type of

source data and it’s configuration
 It might make sense to instantiate 10 virtual

servers for ETL operations that run in parallel—as

the ETL operations execute, they write data to
cloud storage, which is taken as input to pre‐
processing scripts that output data into the proper
format for the data mining application

Cloud Platforms
Department of AI and Big Data
21
 ADVANCED ANALYTICS IN CLOUD
 Allocate storage to store the raw and processed
data; once the data has gone through the pre‐
processing stage, the raw data is deleted.
 Once the predictive models are built, the output of

the preprocessing stage is deleted as well

Cloud Platforms
Department of AI and Big Data
22
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
 Time to deploy new services
 Cost control

 Ability to scale to demand

 Adaptability of resources

Cloud Platforms
Department of AI and Big Data
23
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
Time to deploy new services
 When hardware is dedicated to specific functions,

it can be difficult to find compute and storage

resources for a new initiative.
 Procuring hardware can be time consuming.

Capital expenditures for multiple servers, storage

arrays, and other equipment can require multiple
levels of approval.
 Plans may have to be reviewed and approved from

both a budget and technical perspective.

 Delivery of hardware can take weeks, and in some

cases, months.

Cloud Platforms
Department of AI and Big Data
24
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
Time to deploy new services
 Installing hardware is a multifaceted process. It

needs to be configured according to organizational

standards and incorporated into support systems,
like backup schedules and patch management
systems.
 Some of the most frustrating delays come when a

single piece of hardware, such as a storage

controller, has to be ordered separately and
installed when the server arrives.

Cloud Platforms
Department of AI and Big Data
25
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
Cost Control and Ability to scale as per
demand

Cloud Platforms
Department of AI and Big Data
26
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
Adaptability of Resources
 Through the course of IT’s history, there has been

a trend toward making computing resources more

adaptable.
 In 1960s and 1970s, if you purchased a mainframe

or mini‐computer from IBM, Digital Equipment, or

one of the few other hardware vendors of the day,
you would get “the” operating system (OS) for
that machine.
 In the 1990s, it was not uncommon to run different

OSs on the same type of hardware.

 Cloud computing has moved this trend to the next

stage with the ability to rapidly switch virtual

machine images running on a hardware platform
Cloud Platforms
Department of AI and Big Data
27
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
Adaptability of Resources
 In the cloud, hardware resources are not tightly

coupled to any single platform.

 The same resource that runs an instance of

Windows Server 2008 an hour ago may be running

Ubuntu Linux now.
 A set of servers that were tasked with generating

reports for a data warehouse might be used to

generate customer invoices after that.
 Removing restrictions on the type of software and

radically reducing the time and expertise required

to change OS platforms significantly improves the
adaptability of hardware.
Cloud Platforms
Department of AI and Big Data
28
 ADVANTAGES OF DOING BUSINESS WITH
CLOUD
 The advantages of cloud computing stem from the
ability to deploy new services faster than possible
under more traditional models.
 The ability to control costs at a much fine grained

level of detail than possible before, including the

ability to rapidly scale to needs and the
adaptability of resources to different tasks.
 The movement away from dedicated servers for

single tasks to using cloud resources brings with it

several sources of ROI.

Cloud Platforms
Department of AI and Big Data
29
 SOURCES OF ROI IN CLOUD
Lowering Capital Costs with Cloud Computing
 Reduced need for initial capital outlay
 Reduced upfront cost of IT infrastructure
 Reduced need for building for peak capacity
 No need to purchase dedicated hardware and incur it’s
procurement, operation and maintenance cost just to
meet the peak demand
 More efficient utilization through virtualization
 Same resource can be used for multiple tasks
 Scalabilityand adaptability with disaster recover is
possible due to virtualization

Cloud Platforms
Department of AI and Big Data
30
 SOURCES OF ROI IN CLOUD
Lowering Operational Costs with Cloud
Computing
 On‐demand provisioning
 Reduced marginal cost of systems administration

 Standardization and automation

 Service management reporting

Cloud Platforms
Department of AI and Big Data
31
 SOURCES OF ROI IN CLOUD
On-demand Provisioning
 Submitting a service desk ticket requesting a
virtual machine instance
 Identifying which physical server will host the

virtual machine
 Determining the configuration parameters for the

new instance
 Specifying required support services, such as

backups
 Coordinating with other users on the shared hosts

to avoid common peak demand periods—for

example, running a full backup on one virtual
machine instance while an I/O intensive job is
running on another instance.
Cloud Platforms
Department of AI and Big Data
32
 SOURCES OF ROI IN CLOUD
On-demand Provisioning

Cloud Platforms
Department of AI and Big Data
33
 SOURCES OF ROI IN CLOUD
Reducing Marginal Costs of Systems
Administration
 Installing new applications and packages on
servers
 Patching OSs and applications on each server

 Backing up local storage on each server

 Allocating space to file systems as needed

 Reviewing and purging log files

 Performing security checks, such as running

vulnerability scanners and reviewing results for

each server

Cloud Platforms
Department of AI and Big Data
34
SOURCES OF ROI IN CLOUD

Cloud Platforms
Department of AI and Big Data
35
 SOURCES OF ROI IN CLOUD
Standardization and automation
 Another reason for operations‐related ROI is that
by standardizing on a set of general purpose
virtual machine images, you reduce the overhead
in maintaining them.
 Images are deployed and virtual machine
instances are started using a management
console, so a cloud user who knows how to deploy
a Windows server knows how to deploy a Linux
server or a relational database as well.
 Standardization also enables behind‐the‐scenes

automation that further reduces the demand for

systems administrator expertise
Cloud Platforms
Department of AI and Big Data
36
 SOURCES OF ROI IN CLOUD
Service management reporting
 Number of server hours allocated
 Overall average server utilization

 Average server utilization by hour

 Average server utilization by instance type

 Total storage space used

 Amount of network I/O

Cloud Platforms
Department of AI and Big Data
37
 ASSESSING THE BUSINESS VALUE OF CLOUD
SERVICES
Utilization of all cloud resources
 If resources are underutilized, servers can be
powered down to save on energy costs. IT may
also want to promote the use of the cloud and
publicize availability of resources.

Systems management hours

 Labor can account for significant portions of IT

operating budgets but should be significantly less

for cloud services.

Cloud Platforms
Department of AI and Big Data
38
 ASSESSING THE BUSINESS VALUE OF CLOUD
SERVICES
Virtual machine image use
 All images in a service catalog have to be
maintained. If some images are not used, or used
infrequently, they may be incurring more costs
than they recoup through usage charges.
 Infrequent use or use by only one user can also

indicate specialized or “one off” images.

 These are sometimes necessary to meet business

requirements, but if the number of specialized

images grows, the cost of maintaining them will
increase Charges may need to be adjusted to
recoup the full costs of maintaining specialized
images.
Cloud Platforms
Department of AI and Big Data
39
 ASSESSING THE BUSINESS VALUE OF CLOUD
SERVICES
Time to provision
 This metric can indicate insufficient resources in
the cloud.
 If a sufficient number of servers are not available,

users will have to wait for other jobs to finish in

the cloud before there virtual machine instances
will be provisioned

Cloud Platforms
Department of AI and Big Data
40
 SUMMARY
 Cloud computing offers new ways to deliver
business services. As the two example scenarios
highlighted, different types of business
applications can benefit from deploying in the
cloud.
 The ability to scale compute and storage
resources as needed reduces the need to build for
peak demand.
 This, in turn, reduces the cost of delivering

services while avoiding costly risk mitigation

strategies, such as adding contingency funds to a
project budget to purchase additional hardware to
meet unexpected demand
Cloud Platforms
Department of AI and Big Data
41
 SUMMARY
 Further benefits of cloud computing accrue with
regards to reducing the time to deploy new
services, more ways to control costs, and the
adaptability of resources.
 Cloud can be repurposed rapidly and with minimal

technical expertise, reducing the need for

dedicated servers and their typical low utilization
rates.
 Perhaps the primary driver for the adoption of

cloud computing is the ROI.

 Capital costs are reduced largely due to higher

utilization rates of servers.

Cloud Platforms
Department of AI and Big Data
42
 SUMMARY
 Even more substantial savings can realized with
self‐service management and savings in systems
management.
 With standardized images, automation, and

service management reporting, cloud users can

not only deploy services in the cloud but also
manage them effectively.

Cloud Platforms
Department of AI and Big Data
43
 Q&A
THANK-YOU

Cloud Platforms
Department of AI and Big Data
44