1. Cloud Computing Overview and Architecture
Cloud Computing, CS
Cloud Computing Overview and Architecture

2. Course Content Overview and Cloud Computing Architecture
IaaS, PaaS, and SaaS
Data Management in the Cloud
Hardware and OS Considerations for the Cloud, MS Azure
End-to-End Design, and Operations Management
Case Study:
Google Cloud Computing: Google App Engine
Amazon Cloud Computing: EC2, AWS, S3, SQS, SimpleDB
Yahoo Cloud Computing
IBM Cloud Computing
OpenStack
Mobile and Cloud Computing

3. Textbooks
Cloud Computing Architected, 2011, John Rhoton & Risto Haukioja, Recursive Press, ISBN-13:
Cloud Computing Explained, 2nd edition, John Rhoton, Recursive Press, ISBN-13:
The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machine, by Luiz André Barroso and Urs Hölzle, Google Inc., ISBN (Internet free download)
Programming Google App Engine, Dan Sanderson, O’Reily | Google Press, 2010, ISBN:
Programming Amazon EC2, Jurg van Vliet & Flavia Paganelli, O’Reilly, 2011, ISBN:

4. Outline Background Cloud Computing
Characteristics, Attributes, Terms, Terminology
Hardware: Data Center and Thin Client
Software: Virtualization and Services
Why now? What’s the challenge ahead?
Mobile Applications and Social Networking
How are they related to Cloud Computing?
Impact on Consumers and Enterprises
Summary and Conclusion
I can pretty much read this one straight through and provide details in following slides.

5. Background
Here’s an overview of cloud computing. Basically the main bulk of computing power resides in the cloud, accessible on-demand through network. There may be a separate software/service provider or a cloud user that channels/aggregates/provisions the desired services to end users.
Source: “Above the Clouds: A Berkeley View of Cloud Computing” Feb. 4, 2009

6. Background: Cloud Computing in a Nutshell
Applications delivered as services over the Internet
Hardware and systems software in the datacenters that provide those services.
Cloud: datacenter hardware and software
Centralization of IT infrastructure made available to end users, software providers, and enterprises over the Internet
Service Models:
Software as a Service (SaaS, e.g. salesforce.com)
Platform as a Service (PaaS, e.g. Google’s App Engine)
Infrastructure as a Service (IaaS, e.g. Amazon Web Service)
Why?
Easier to manage and Larger economies of scale
In general, cloud computing can mean 1. applications/services delivered over the internet 2. HW and Systems SW in DC that enable the services.
Features: Centralization- think of Office, OS, AV updates. Roughly, there are 3 service models.

7. Background: Cloud Computing Economy
Economies of scale in hardware acquisition cost (5-7X improvement), software license cost, electricity cost and IT management cost (7X improvement)
Enable SaaS providers that don’t own IT infrastructure
Semiconductor
Industry
Software
Industry
Analogy
Foundries
(TSMC, UMC)
Cloud
Computing
According to the Berkeley research paper, CC saves. Interesting analogy between semi and software industries.
Enable
Enable
Fabless
Chip Design
(e.g. nVidia…)
Datacenter-less
SaaS provider
Source: “Above the Clouds: A Berkeley View of Cloud Computing” Feb. 4, 2009

8. Background: Players in Cloud Computing Space
End User
Enterprise User
Consumer
SaaS User
Cloud Service User
Enterprise IT Department
Web Application Developer
SaaS Provider
Cloud Service Provider
Mega Data Center Operator
Cloud Application Platform Provider
A closer look at CC with each component
Mega Data Center
Hardware Provider
Software Provider

9. Background: MEGA Data Center Building Blocks
Server vs. Rack vs. Container
High Performance vs. Low Power CPU
Enterprise Grade vs. Commodity Hardware
GPU or FPGA?
Data Center Networking
Eliminate Layer-3 routers within data center
Network load balancing
Storage System
Unification of data and storage networking
Flash Memory or even DRAM storage
Since economies of scale are in play, larger DC is in favor. So the HW blocks can be enumerated.
How do you design for best performance?

10. Background: Challenge 1 - Power Efficiency
Why do you build Mega Data Centers at the scales of 100k to one million machines?
Energy cost becomes critical
Cooling System
Think Green
New Metric: Amount of work done/joule
Intelligent Power Management
Centralized or distributed power source
Fine-granularity HW component turn on/off: memory bank, CPU core, CPU, disk, etc.
Modularity and Flexibility
Easily Expandable/Removable Systems
Physical Design Concerns
MSFT, Google built in Washington and Oregon, Apple in NC

11. Background: Challenge 2 - Data Center Management
Design for manageability
Discovery: Establish dependencies among system components. If Switch 32 dies, how many services are affected?
Diagnosis: Identify the root cause behind a functionality or performance failure
Remediation: Take corrective actions to minimize the mean time to repair
Complexity of Managing of Real and Virtual Computing, Networking, Storage and Other Devices

12. Cloud Computing: Characteristics
Cloud Computing: is a style of computing in which dynamically scalable and often virtualized resources are provided as a service over the internet
Virtual: Software, databases, Web Servers, Operating Systems, Storage, and networking as virtual servers
On-demand: ability to add and subtract processors, memory, network bandwidth, storage as needed
Users need not have knowledge of, expertise in, or control over the technology infrastructure in the “cloud” that supports them
Software to utilize HW to the max

13. Cloud Computing: Attributes
Off-Premise: The service is hosted & delivered from a location that belongs to a service provider over the Internet
Elasticity: leverage the inherent scalability of service provider by providing elastic provisioning mechanism so that resources can be scaled up/down quickly as needed
Flexible Billing: Fine-grained metering of resource usage combined with on-demand service provisioning
Virtualization: Cloud services offered through an abstracted infrastructure
Service Delivery: Typically the service offers programmatic API in addition to user interface
Software to utilize HW to the max

14. Cloud Computing: Attributes
Universal Access (Resource Democratization): Pooled resources are available to anyone authorized to utilize them
Simplified Management: through automatic provisioning to meet scalability needs, user self-service, and programmatically accessible resources to integrate into the Enterprise Management Framework
Affordable Resources: cost is reduced dramatically because: (1) No capital cost. (2) Economy-of-scale
Multi-tenancy: mechanisms to isolate tenants (organizations) from each other while pooling resources across customers
Service-level Agreement (SLA): Sets the expectations with the customer
Software to utilize HW to the max

15. Cloud Computing: Terms
Service-Oriented Architecture (SOA): Decompose IT landscape of an Enterprise into unassociated and loosely coupled functional services. An enterprise that uses SOA is better positioned to leverage Cloud Computing
Grid Computing: Use of many interconnected computers to solve a problem through highly parallel computation
Web 2.0: Is associated with Cloud Computing even though there are no intrinsic connections. Web 2.0 refers to a class of services (platform for web-based collaborative communities) that may be delivered in many different ways, while Cloud Computing refers to a means of delivering the service
Software to utilize HW to the max

16. Cloud Computing: Since 1999 - “Web 2.0”
– Self publishing: LiveJournal/Blogger.com (1999), YouTube
(2005)
– Metadata & feeds: RSS (1999), Atom (2003)
– Collaboration: Wikipedia (2001), Google docs (2006)
– Tagging: del.icio.us (2003), digg.com (2004), flickr (2004)
– Ajax: Gmail (2004), Google Maps (2005)
– Long Tail: Amazon (1994), Google AdSense (2005)
– API: Yahoo!, GData, Amazon… (who doesn't?)
– Aggregation: Google News (2002), Google Reader (2005)
– Social: LinkedIn (2003), MySpace (2003), Facebook (2004)
– Microblogging: Twitter (2006), Plurk (2008)

17. Cloud Computing: Terminology
Software to utilize HW to the max

18. Cloud Computing: Hardware – Datacenter & Thin Client
Cloud computing is the modern, if not ultimate, form of Server-Based Computing (SBC)
Thin client has been used successfully for the last 20 years
Thin client is typically web-based desktop management and support, using HTTP or FTP, and as a result the administration of the thin client infrastructure can be outsources as a cloud service
With thin client and outsourcing IT to external service provider, client data are handled professionally and data availability is guaranteed (SLA)
Software to utilize HW to the max

19. Cloud Computing: Hardware – Datacenter Classification
Tier I datacenters:
Single path for power and cooling distribution, without redundant components
Non-redundant components
Expected availability of %
Tier II:
Adds redundant components to this design (N + 1), improving availability
Expected availability of %
Tier III:
Multiple power and cooling distribution paths but only one active path.
They also have redundant components and are concurrently maintainable, that is, they provide redundancy even during maintenance, usually with an N + 2 setup
Concurrent maintainable infrastructure
Expected availability of %
Software to utilize HW to the max

20. Cloud Computing: Hardware – Datacenter Classification
Tier IV
Meets or exceed tier 1, tier 2, and tier 3 requirements
Two active power and cooling distribution paths, redundant components in each path
Tolerate any single equipment failure without impacting the load
Expected availability of %
Software to utilize HW to the max

21. Cloud Computing: Hardware - Main Components
Power
UPS
PDU
Cooling
CRAC
Free Cooling
What about efficiency?
The amount of computational work performed divided by the total energy used in the process
Software to utilize HW to the max

22. Cloud Computing: Hardware – Datacenter, Where Energy Go
Software to utilize HW to the max

23. Cloud Computing: Hardware – Datacenters Tour
Google
Microsoft HP
IBM, Facebook, Apple, do your own search
Software to utilize HW to the max

24. Cloud Computing: Software – Virtualization and Services
Mapping of Users’ Hardware Requirements
Example: Dual core 2.4GHz CPU, 3GB 800MHz Bus-Speed Memory, Gigabit Ethernet and Two Terabytes of 7200 RPM Hard Drives
Other Environment Variables, Peripherals, Boot Options
Resource Allocation and Provisioning
Virtualization for GPU, PS3, FPGA engine, etc.
Virtualization-based fault tolerance:
Backup and Reliability
Software to utilize HW to the max
Example:
Machine1 Machine Machine3 Machine4
VM VM1(B) VM VM2(B)
VM3 (B) VM VM4 (B) VM4
VM5 (B) VM VM5 (B) VM5

25. Cloud Computing: Software – Virtualization and Services
Hypervisor, Virtual Machine Manager - VMM, is one popular technique allowing multiple Operating Systems, guest OS, to run concurrently on a host computer
Hypervisor describes the interface provided by the cloud computing functionality known as IaaS
There are two types of hypervisors:
Type-1 (bare metal) hypervisors: run directly on the host hardware. It controls both the hardware as well as the guest OS
Type-2 (hosted) hypervisors: run within the conventional OS as a distinct 2nd layer, and the guest OS runs as a 3rd layer
Software to utilize HW to the max

26. Cloud Computing: Software – Virtualization and Services
Software to utilize HW to the max

27. Cloud Computing: Software – Virtualization and Services
Software to utilize HW to the max
Types of Cloud Service

28. Cloud Computing: Why Cloud Computing Now? (1)
Market Pull
Software on the premise  SaaS
Information technology (IT) on the premise  IT service as a rented utility (Electricity)
IT should not and will not be a core competence for most corporations (Nicholas Carr’s “IT does not matter” and “The Big Switch”)
PaaS and IaaS
Lowering up-front and day-to-day IT cost:
Pay only as much as actual resource usage
Total Cost of Ownership including SW maintenance

29. Cloud Computing: Why Cloud Computing Now? (2)
Technology Push
Broadband network connectivity getting faster and more reliable
Low Cost Powerful Processors and Advanced Virtualization Software
Sufficient trust in infrastructure providers
Google is already a critical service for most of the world, and it is in the cloud!
Comparison:
Mainframe systems
Trust is a very important issue. Remember Google mail hiccup this summer?
So private cloud has its market. Counter to mega data center.
Technology has advanced so much. We’re not going back to mainframe days (time-slicing) although the picture may look similar.

31. Mobile Apps & Social Networking: Innovative Apps of Cloud Computing
Mobile Applications and Social Networking are greatly empowered and enriched by Cloud Computing.
Popular mobile devices have to be low power, low priced, thus need to shift most computing tasks to the cloud.
Cloud-backed consumer device:
Thin phone + desktop virtualization
Mobile Internet Device (MID)
E-Books and other Gadgets
Think smartphone Apps and Facebook. Facebook Apps exist in cloud. Smartphones, despite the increase in CPU speed, are not powerful enough for serious computations.

32. Mobile Apps & Social Networking: Social Networking
Facebook, Twitter, Orkut, Linked-In
Viral Effect and Stickiness
Attract huge investment from big players
Online Gaming and Wireless Access
Solution by Cloud Computing
Global Movement and Reach
Kaixin Business Model
Virtual Economy and Social Issues
China Health Ministry to treat 'Net addiction as disorder

33. Mobile Apps & Social Networking: Social Networking in Enterprises
Cisco Systems
The Human Effect Networking Campaign
Acquisition of Five Across, Pure Digital
IBM
Lotus Connections - Social software for business
Home page, Blogs, Profiles, Wikis, Communities...
Google and Others
Performance review based on network circles
Development of Social Networking style Corporate Communications Tools

34. Summary and Conclusion
Cloud Computing
More Public and Private Clouds
Hardware: MB/Server/Rack – ASUS position
Software Challenges
Mobile Apps and Social Networking
Development of innovative applications
Always connected thin clients
We are in for a long ride of a huge wave
It’s probably just the first inning so far.

35. END