Transcript Cloud
NETE4631
Network Information Systems (NISs):
Cloud Computing
Suronapee, PhD
[email protected]
1
Outline
What is Cloud Computing?
Why Cloud Computing?
Cloud Basics and Terminology
Virtualization
Cloud service models
Cloud deployment models
Is Cloud Computing for Me?
Moving application to the cloud
Cloud Challenges and Suitable Applications
2
Big “Cloud”
What people think about the cloud can be?
a datacenter hosting many servers
Services accessible remotely through the internet
Employee of some company, each produces a work output
We use cloud to represent things that we are trying to
abstract (black box)
3
What is Cloud Computing?
“a computing model for enabling convenient,
on-demand network access to a shared pool of
configurable computing resources”
(A definition by NIST, 2011)
Resources can be…
Data Center
networks, servers, storage, applications, and services
Resources are…
4
rapidly provisioned and released with minimal management
effort or service provider interaction.”
Key players of the cloud
Cloud providers
Operates datacenter
Computing resources (server)
Networking resources (switches)
Storage resources (memory devices)
Service providers
Offers software that run in data centers
Cloud users
5
Consumers or enterprises that uses services
running in data centers
What features define cloud services?
Large-scale computing and storage systems
Networking within a data center, across the data center
and to the end users (cellular or WiFi)
Software that provides a GUI, security and privacy, billing
and charging, etc
Overall characteristics
Keyword is “On demand” in
6
Time – change their resource requests for a short period of time
Scale – Scale up and scale down your solutions
History of Cloud
From centralized -> distributed -> cloud computing
The term “Cloud”, Google CEO, 2006
Important steps
7
1961 – computing can be sold like a utility, John MacCharty
1999 – paid application delivered through internet, Saleforce.com
2002 – AWS started, 2006 - Elastic computing cloud (commercial)
2009 – Google apps (cloud-based app) provided to enterprise
Why Cloud Computing
Abstraction for Hardware and IT Infrastructure
There is no upfront cost
A lot of infrastructure to host services
Distributed computing at a massive scale
Mass storage
Always on and On Demand
Scale up and scale down your solutions
Cost is scaled appropriately (Pay per use model)
Convenience and Collaboration
8
Scale automatically
Minimum management efforts (self service)
Can we count on the cloud?
Major concern: shared facility
Consider when using cloud storage than personal storage
9
Pros: cheaper and convenient
Cons: performance,, security and privacy concerns
Analogy: buying your car or riding in the bus
Gartner’s Hype Cycle
10
Usage Patterns
Windows Azure Overview, Microsoft Research
11
Scenario - Reservation System
Requirement
Scalability
Availability
Serve access capacity
Access when required.
Solution
12
Clone system to handle peak load
(hybrid cloud)
Why cloud?
On demand side
Cheaper
Convenience
CapEx to OpEx (pay per use)
Ease of use, flexibility to scale
On supply side
Economic of scale
The average cost to pay is much
lower
Feasibility to scale
13
The average demand of a server
in a data center < 20%
Virtualization Technology
Virtualization is…
the creation of a virtual (rather than actual) version of
something,
Something can be a server, hardware, operating system (OS),
storage device, or network resources, etc.
Server Virtualization
Two type of usage
Server consolidation
High performance computing
Desktop Virtualization
14
Why Virtualization?
High performance computing
All require same physical space.
All require same power.
The system is not fully utilized.
Reconfiguration and setup in a
shared and control manner
Types of Virtualization:
Hardware Virtualization
Operating system Virtualization
Everything runs from the network using a kind of virtual disk.
Nothing is pre-installed or permanently loaded on the local device
Server Virtualization
VMM is directly installed on the hardware system
For different OS plus their applications on the hardware system
VMM is directly installed on the Server system
For creating multiple servers on the demand basis
Storage Virtualization
16
Grouping the physical storage from multiple network storage
For back-up and recovery purposes.
http://www.javatpoint.com/windows-virtualization
17
How Virtualization works
Hypervisor
A low level program that provides system resources access to virtual
machines
Hypervisor is also called Virtual Machine Monitor (VMM)
One level higher than the supervisory program for a hardware and/or
hardware & OS
Two types of Hypervisor
18
Type 1 or full virtualization (Oracle VM,Vmware ESX)
Type 2 or para-virtualization (Xen)
Types of Hypervisors
19
Types of Hypervsiors : Reproduced Figure 5.1 from
Sosinsky, B., Cloud Computing Bible, 2011.
Benefits of Virtualization
Virtualization supports “Abstraction”
Virtualization supports Load balancing
Map logical address (id) for extendable and scalable physical
resources
A load balancer listens to service request and uses a
scheduling algorithm to assign the request
Keeps the record of a request’s session to support a failover
for the request
VM Migration
20
Porting applications between physical machines
Abstraction: Virtualization Infrastructure
Websphere
Exchange
File/Print
APP
APP
APP
APP
APP
APP
OS
OS
OS
OS
OS
OS
APP
APP
APP
OS
OS
OS
Virtual Infrastructure
CPU
Pool
Memory
Pool
Storage
Pool
Interconnect
Pool
Abstraction: Virtualization Infrastructure
Websphere
Exchange
File/Print
APP
APP
APP
APP
APP
APP
OS
OS
OS
OS
OS
OS
APP
APP
APP
OS
OS
OS
Virtual Infrastructure
CPU
Pool
Memory
Pool
Storage
Pool
Interconnect
Pool
Load Balancing: Pods, Aggregation,
& Failover
23
VM Migration
Layer 2 Network
vSphere
local
vSphere
local
สามารถการย้าย virtual machine โดยการย้ายไปยัง่
อีก Storage หนึ่ง
Infrastructure as a Service (IaaS)
Offering IT infrastructure
Virtual Machines (VM)
VM abstract underlying hardware i
Very few cloud provider provide an OS
Other resources such as storage, processing, etc.
Benefits
25
Eliminates the need for every organization to maintain the IT
infrastructure (Cloud Providers owns equipment)
Scale automatically and bill only for the actual usage
Amazon Elastic Compute Cloud (EC2)
26
Amazon EC2 Pricing
27
IaaS Deployment Models
Public Cloud
Private Cloud
Infrastructure is for public use
Own by org. selling cloud
Infrastructure is for private use
Full control of all resources and technologies
Hybrid Cloud
Combine multiple clouds (private and public)
Better Security and less cost
However, some works need..
Policies compliance
Synchronization
Public Cloud
- Infrastructure cost
- Full Control
- Security?
Private Cloud
28
Platform as a Service (PaaS)
Offering Platforms for developing scalable applications
Virtual Machines + developments framework and tools
No access to the underlying virtualization or operating system
Benefits
Application (scalable) accessible through the internet
Reduce developers’ burden by additionally supporting the platform
runtime and related application services.
Challenges
29
Vendor lock-in
Google App Engine
30
Windows Azure
Windows Azure –Service hosting and management,
storage, computation, networking
Microsoft SQL Services –Database services and reporting
Microsoft .NetServices –Service-based implementation of
.NET framework
31
Software as a Service (SaaS)
Offering cloud-based applications
Benefits
Delivered over the platform of the web
Abstracts users from all of the underlying components
Cheaper than shrink-wrapped versions
Eliminate the need to install and maintain, compatibility and fast upgrades
Challenges
32
Internet connection and latency
Difficult to translate business models
Cloud Service Models
33
Service Models and their risks
34Modified
from Cloud Computing Impact on future enterprise architectures,
Schekkerman, J.
Moving Application to the Cloud
Deploy Application to the cloud?
Entirely or Part of the application is on the local system and another
part is in the cloud
Business case
Study the feasibility of deployment in terms of costs-benefits analysis
Dimensions to consider to save cost
35
Ongoing operational cost reduction
The value of preserving capital
The value of upsizing on demand
The value of downsizing on demand
The value of agility
The value of reuse
The value of coolness
Operational Cost Reduction
36
Upsize and Downsize on Demand
37
Case Study on Oil & Gas Industry
The migration use-case of an IT System
From a company’s in-house data center
To Amazon EC2
System Overview
38
System Deployed in the Cloud
Infrastructure Costs
Company C paid £104,000 to Company B for the system
in 2005, £19,400 of which was for the system’s
infrastructure.
In addition, C paid B £43,000 per year for system support
and maintenance, £3600 of which is for the running costs
of the system infrastructure.
Over a 5 year period, the total cost of the system
infrastructure is therefore: £19,400 + (5 x £3,600) =
£37000
39
Price Comparision
Amazon EC2 provides an option of using either small or
large server instances depending on the amount of CPU
power and RAM required.
In Cloud, 37% cheaper compared to 1 small and 1 large
server instance
40
Support and Maintenance
In 2005, 218 Support calls have been made regarding the
operation of the system.
The majority were about software problems.
45 calls were related to the system’s infrastructure.
38 calls – backup problems
5 calls – network problems
2 calls – power outages at B
In cloud – 21% eliminated
41
Stakeholders’ Impact Analysis
Stakeholders’ impact analysis is a method of identifying
potential sources of benefits and risks from the
perspectives of multiple stakeholders.
42
Some of the Challenges!!!
Security
Would my data be more secure with Cloud provider?
Interoperability
Significant risk of vendor lock-in –Standardized interfaces not available,
incompatible programming models
Reliability
Use of commodity hardware, prone to failure
Make sure that SLAs provided
Laws and regulations
Privacy, security, and location of data storage
Organizational changes
Changing authorities of IT departments, compliance policies
Cost
Purchase vs. Lease?
Network Latency
How long does it take to send 1 TB over 20Mbps in WAN
43
Following types of application are more
suitable for cloud
Not mission critical.
Not handling core business operations.
Not dealing with sensitive data.
Can tolerate high network latencies and work on low
network bandwidth.
Following industry standards. E.g. using standardized
tools for object to relational mappings.
Do not require detailed customization for each tenant.
Organization involved in porting have full knowledge of
application and its associated domain.
44
References
Armbrust, M., et al., 2010, A View of Cloud
Computing, ACM, 53(4), pp. 50-58.
Zhang, Q., Cheng, L., Boutaba, R., Cloud Coomputing:
state-of-the-art and research challenges, Journal of
Internet Services and Applications, 2010, 1:7-18.
The Future of Cloud Computing: Opportunities for
European Cloud Computing Beyond 2010.
Sosinsky, B., Cloud Computing Bible, Wiley, 2011
VMware Technologies
45