1. IN4392 Cloud Computing
Introduction to Cloud Computing
Alexandru Iosup
Parallel and Distributed Systems Group
Delft University of Technology
The Netherlands
Our team: Undergrad Thomas de Ruiter, Anand Sawant, Ruben Verboon, …
Grad Siqi Shen, Guo Yong, Nezih Yigitbasi Staff Henk Sips, Dick Epema, Alexandru
Iosup, Otto Visser Collaborators Ion Stoica and the Mesos team (UC Berkeley),
Thomas Fahringer, Radu Prodan, Vlad Nae (U. Innsbruck), Nicolae Tapus, Mihaela
Balint, Vlad Posea (UPB), Derrick Kondo, Emmanuel Jeannot (INRIA), Assaf Schuster,
Mark Silberstein, Orna Ben-Yehuda (Technion), ...
February 21, 2013
1
IN4392 Cloud Computing
4. What is Cloud Computing?
3. A Useful IT Service
“Use only when you want! Pay only for what you use!”
Q: What do you use?
Q: Why not this level?
Q: Why not this level?
February 21, 2013
4
5. Agenda
1. What is Cloud Computing?
2. IaaS Clouds, the Core Idea
3. The IaaS Owner Perspective
4. The IaaS User Perspective
5. Reality Check
6. Conclusion
February 21, 2013
5
6. IaaS Cloud Computing
VENI – @larGe: Massivizing Online Games using Cloud Computing
7. Joe Has an Idea ($$$)
MusicWave
(Source: A. Antoniou, MSc Defense, TU Delft, 2012. Original idea: A. Iosup, 2011.)
8. Solution #1
Buy or Rent
• Big up-front commitment
• Load variability: NOT supported
10%
…
(Source: A. Antoniou, MSc Defense, TU Delft, 2012. Original idea: A. Iosup, 2011.)
9. Solution #2 • NO big up-front commitment
Deploy on IaaS Cloud
• Load variability: supported
Q: So are we just shifting the problem to
somebody else, that is, the IaaS cloud owner?
(Source: A. Antoniou, MSc Defense, TU Delft, 2012. Original idea: V. Nae, 2008.)
10. Inside an IaaS Cloud Data
Center
(Source: A. Antoniou, MSc Defense, TU Delft, 2012. Original idea: A. Iosup, 2011.)
11. Time and Cost Sharing Among
Users
User C
User B
MusicWave
(Source: A. Antoniou, MSc Defense, TU Delft, 2012.)
12. Main Characteristics of IaaS Clouds
1. On-Demand Pay-per-Use
2. Elasticity (cloud concept of Scalability)
3. Resource Pooling
4. Fully automated IT services
5. Quality of Service
February 21, 2013
12
13. Agenda
1. What is Cloud Computing?
2. IaaS Clouds, the Core Idea
3. The IaaS Owner Perspective:
How to Deploy a Cloud?
4. The IaaS User Perspective
5. Reality Check
6. Conclusion
February 21, 2013
13
14. IaaS Cloud Deployment Models
Private Public
On-premises Off-premises
Hybrid
(Source: A. Antoniou, MSc Defense, TU Delft, 2012. Original idea: Mell and Grance, NIST Spec.Pub. 800-145, Sep 2011.)
15. Resource Sharing Models
MusicWave
Grids IaaS Clouds
Space-Sharing Time-Sharing
Q: Which one is better?
MusicWave OtherApp MusicWave OtherApp
OtherApp
Host OS Host OS
February 21, 2013
15
16. Virtualization
Applications Applications
Guest OS Guest OS
MusicWave
Virtual Resources OtherApp OtherApp
Virtual Resources
Q: What is the problem?
Q: What to do now?
VM Instance VM Instance
Virtualization
Host OS
February 21, 2013
16
17. Virtualization and The Full IaaS
Stack
Applications Applications
Applications
Guest OS
Guest OS Guest OS
Virtual Resources
Virtual Resources Virtual Resources
VM Instance Instance
VM VM Instance
Virtual Machine Manager Virtual Machine Manager
Virtual Infrastructure Manager
February 21, 2013
Physical 17
Infrastructure
18. The Virtual Machine Lifecycle
Q: Is this fair?
February 21, 2013
18
(Source: A. Antoniou, MSc Defense, TU Delft, 2012.)
19. Use Case:
Amazon Elastic Compute Cloud (EC2)
• Prominent IaaS provider
• Datacenters all over the world
Instance Capacity US$/hour
• Many VM instance types
m1.small 0.10
• Per-hour charging m1.large 0.38
c1.xlarge 0.76
February 21, 2013
19
20. Agenda
1. What is Cloud Computing?
2. IaaS Clouds, the Core Idea
3. The IaaS Owner Perspective
4. The IaaS User Perspective:
How to Use Clouds? How to Choose Clouds?
5. Reality Check
6. Conclusion
February 21, 2013
20
21. Workload
MusicWave OtherApp
OtherApp
OtherApp
Load = 4
OtherApp
MusicWave
Time
RunTime= 6
February 21, 2013
21
22. Use Case: Workloads of Zynga
(Massively Social
SellingGaming)
in-game virtual
goods:
“Zynga made est.
$270M in 2009 from.”
http://techcrunch.com/2010/0
5/03/zynga-revenue/
Sources: CNN, Zynga.
“Zynga made more than $600M in 2010
from selling in-game virtual goods.”
S. Greengard, CACM, Apr 2011
February 21, 2013
22
Source: InsideSocialGames.com
23. Use Case: Workloads of Zynga
(Massively Social
Gaming)
Load
• Load can grow very quickly
February 21, 2013
23
25. Provisioning and Allocation of Resources
Q: What is the interplay between provisioning and allocation?
Provisioning Allocation
Load
Time
February 21, 2013
25
26. Provisioning and Allocation Policies
Q: How many policies exist? Q: How to select a policy?
Provisioning Allocation
When? From where? When?Where?
How many? etc.
Load
Which type?
etc.
Time
February 21, 2013
26
(Source: A. Antoniou, MSc Defense, TU Delft, 2012.)
27. Use Case:
Two Provisioning Policies,
Compared
Startup
OnDemand
February 21, 2013
Villegas, Antoniou, Sadjadi, Iosup. An Analysis of Provisioning and
27
Allocation Policies for Infrastructure-as-a-Service Clouds,
28. Use Case:
Two Provisioning Policies,
Compared
Metrics for comparison
• Job Slowdown (JSD ): Ratio of actual runtime in the
cloud and the runtime in a dedicated non-virtualized
environment
Q: Charged cost vs Total RunTime?
• Charged Cost (C c )
• Utility (U )
February 21, 2013
Villegas, Antoniou, Sadjadi, Iosup. An Analysis of Provisioning and
28
Allocation Policies for Infrastructure-as-a-Service Clouds,
29. Use Case:
Two Provisioning Policies,
Compared
Workloads
Uniform Increasing Bursty
February 21, 2013
Villegas, Antoniou, Sadjadi, Iosup. An Analysis of Provisioning and
29
Allocation Policies for Infrastructure-as-a-Service Clouds,
30. Use Case:
Two Provisioning Policies,
Compared
Environments
System Hardware VIM Hypervisor Max VMs
DAS4/Delft 20 Dual quad- 64
core 2.4 GHz
24 GB RAM
2x1 TB storage
FIU 7 Pentium 4 3.0 7
GHz
5 GB RAM
340 GB Storage
Amazon EC2 unkown/various - 20
February 21, 2013
Villegas, Antoniou, Sadjadi, Iosup. An Analysis of Provisioning and
30
Allocation Policies for Infrastructure-as-a-Service Clouds,
31. Use Case:
Many Provisioning Policies,
Compared
Job Slowdown (JSD)
Q: Why is OnDemand worse than Startup?
A: waiting for machines to boot
February 21, 2013
31
32. Use Case:
Many Provisioning Policies,
Compared
Charged Cost (C c )
Q: Why is OnDemand worse than Startup?
A: VM thrashing
Q: Why no OnDemand on Amazon EC2?
February 21, 2013
32
34. Agenda
1. What is Cloud Computing?
2. IaaS Clouds, the Core Idea
3. The IaaS Owner Perspective
4. The IaaS User Perspective
5. Reality Check:
Who Uses Public Commercial Clouds?
6. Conclusion
February 21, 2013
34
35. The Real IaaS Cloud
VS
http://www.flickr.com/photos/dimitrisotiropoulos/4204766418/ Tropical Cyclone Nargis (NASA, ISSS, 04/29/08)
• “The path to abundance” • “The killer cyclone”
• On-demand capacity • Not so great performance
• Cheap for short-term tasks for scientific applications
• Great for web apps (EIP, web (compute- or data-intensive)
crawl, DB ops, I/O)
February 21, 2013
35
36. February 21, 2013
36
(Source: http://www.cca08.org/files/slides/w_vogel.pdf)
37. Zynga zCloud: Hybrid Self-
Hosted/EC2 large scale
• After Zynga had
• More efficient self-hosted servers
• Run at high utilization
• Use EC2 for unexpected demand
February 21, 2013
(Sources: http://seekingalpha.com/article/609141-how-amazon-s-aws-can-attract-ugly-economics and 37
http://www.undertheradarblog.com/blog/3-reasons-zynga-is-moving-to-a-private-cloud/)
38. Other Cloud Customers
• 218 virtual CPUs
• 9TB/2TB block/S3 storage
• 6.5TB/2TB I/O per month
February 21, 2013
38
(Source: http://markbuhagiar.com/technical/businessinthecloud/)
39. Agenda
1. What is Cloud Computing?
2. IaaS Clouds, the Core Idea
3. The IaaS Owner Perspective
4. The IaaS User Perspective
5. Reality Check
6. Conclusion
February 21, 2013
39
40. Conclusion Take-Home Message
• Cloud Computing = IaaS + PaaS + SaaS
• Core idea = lease vs self-own
• On-Demand, Pay-per-Use, Elastic, Pooled, Automated, QoS
• The Owner Perspective
• Time-Sharing
• Virtualization
• The User Perspective
• Variable workloads
• Provisioning and Allocation policies
• Reality Check: 100s of users
http://www.flickr.com/photos/dimitrisotiropoulos/4204766418/
February 21, 2013
40
41. Thank you for your attention!
Questions? Suggestions?
Observations?
More Info:
- http://www.st.ewi.tudelft.nl/~iosup/research.html
- http://www.st.ewi.tudelft.nl/~iosup/research_cloud.html
- http://www.pds.ewi.tudelft.nl/
Do not hesitate to
contact me…
Alexandru Iosup
A.Iosup@tudelft.nl
http://www.pds.ewi.tudelft.nl/~iosup/ (or google “iosup”)
Parallel and Distributed Systems Group
Delft University of Technology
February 21, 2013
41
Editor's Notes
1 minute Cloud = Infinite stack of computers at your disposal Fine-grained, flexible, dynamic, and cost-effective source of computers for any applications
introduce the problems that have been addressed in this work, by using an example. Consider a startup company which has a new idea, to deploy a web-based music creation tool, that offers the ability to the users of this tool to compose music and possibly collaborate or sell their creations through this platform. The company wants to bring MusicWave to the market. The idea needs to be deployed on infrastructure. So what are their choices in deploying MusicWave?
(1 minute) One solution would be to buy or rent computing power, machines. Two problems with this approach: It requires a big upfront commitment. The company has to spend a lot of money to buy or to rent this equipment, And to subsequently set-it up. They also need to have technical expertise The second problem is the load variability. The number of users of MusicWave will experience load variation over time. As they acquire more Users, their machines will become less able to handle with the load that is experienced, and this can potentially Result in significant delays which will lead to a worse user experience. Moreover, based on experience we know that the average load for such kind of workloads is really small, about 10%, So, the company will pay for resources that are idle most of the time, But at other moments they cannot handle the experienced load.
(seconds) The second solution would be to deploy their application on the cloud. Without having any upfront commitment, they could almost instantly get computing power from the cloud, so that they can Serve the current amount of users on Musicwave. Additionally, as the number of users increases, they could easily make the decision to acquire More computing power to serve all the users. And at moments when the load decreases, they could just as easily make the Decision of releasing resources.
Now lets just say that MusicWave is indeed launched on a Cloud. What you can see here, is a cloud datacenter, which generates the computing power offered to cloud users, with thousands of machines. Now musicWave will be hosted on only a small part of this datacenter, as an example the machines in green here.
But this doesn’t mean that the MusicWave will have these resources for dedicated use! In Cloud systems, the resources are time-shared between multiple users. This sharing allows the cloud provider to better utilize their machines and divide the cost of running these systems between multiple users, and it Subsequently passes the cost savings to the cloud users, by offering non-expensive Computing power
(40 seconds) There are also different deployment models. An organization might setup a cloud ontop of their private infrastructure and intended for private use, this is what we call a private cloud On the other hand, we have public clouds that can be used by anyone who has internet access. Also an organization might decide To use some resources from a public cloud, when it’s private cloud resources are not enough, that’s what we call hybrid cloud.
