www conference 2012 - web-engineering - cloudgenius
TRANSCRIPT
CloudGenius: Decision Support for
Web Server Migration to the Cloud
Michael Menzel, Rajiv Ranjan
KIT, UNSW
WWW Conference – Web Engineering II
Lyon, 2012
Agenda
1. Motivation
2. CloudGenius Framework
3. CumulusGenius Tool
4. Experiments & Evaluation
5. Conclusion
6. Future Work
MOTIVATION
Web Servers in the Cloud
• ... to gain Cloud features
– Elasticity (slashdot)
– Pay-per-use
– Global distribution
– ...
• What to be done?
• Where to go? (Cloud compute service)
Options for Realization
• 3 Options to migrate a Web server
– Convert it into a Cloud-compatible VM image
– Rebuild on a basic VM image
– Adopt prepared Web server VM image
Converted
VM
Image
Basic
VM
Image
Prepared
VM
Image
Less effort
Higher Customizability
Web Server Migration Problem
Web server
Cloud Provider A
Service
A
Service
B
Provider
evaluation
Service
evaluation
Web server
requirements
Web server
goals
Service
decisionCloud Provider B
Servic
e C
Image
decision
Image
evaluation
Image A
Image C
Image B?
?
Composite
decision
VM imagesinfluences
choose
CLOUDGENIUS FRAMEWORK
Elements of the
CloudGenius Framework
Set
Goals/Preferenc
es
Evaluate Images
& Services
Select Image &
Service
Deploy, Customi
ze... ...
Engineer
CloudGenius
Cloud
Model
requirements
Multi-Criteria Decision-
Making Method (AHP)criteria
1. Alternative 2
(0.8966)
2. Alternative 1
(0.1211)
3. ...
Cyclic process Model Evaluation methods
+ +
CloudGenius
Migration Process (condensed)
Set
Goals/Preferen
ces
Evaluate
Images &
Services
Select Image &
Service
Deploy, Custom
ize... ...
Engineer
CloudGenius
Cloud
Model
CloudGenius
Model of Cloud Landscape
• VM Images & Compute Services have attributes
• Attributes are basis for criteria and requirements
• VM Images and Services are related
Model holds Data
Evaluation Methods
Leverage (MC2)2 Framework [1]
appropriate alternatives
Alternative n
Alternative 2
Alternative 1
requirements
Multi-Criteria
Decision-Making
Method (AHP)criteria
1. Alternative 2 (0.8966)
2. Alternative 1 (0.1211)
3. ...
(MC2)2 allows to create evaluation
methods with given criteria and
requirements
Resulting evaluation methods
filter and evaluate alternatives
We settle for AHP
for normalized evaluations
[1] Menzel, M., Schönherr, M., Nimis, J., & Tai, S. (2010). (MC2)2: A Generic Decision-
Making Framework and its Application to Cloud Computing. In Procs. International
Conference on Cloud Computing and Virtualization (CCV 2010), Singapore.
Evaluate VM Images VM
Image
Attributes [2]
[2] S. Kalepu, S. Krishnaswamy, and S. Loke. Verity: A QoS Metric for Selecting Web Services and Providers. In Web Information
Systems Engineering Workshops, 2003. Proceedings. Fourth International Conference on, pages 131-139. IEEE, 2003.
Evaluate Compute Services
Attributes [3]
[3] S. Kalepu, S. Krishnaswamy, and S. Loke. Verity: A QoS Metric for Selecting Web Services and Providers. In Web Information
Systems Engineering Workshops, 2003. Proceedings. Fourth International Conference on, pages 131-139. IEEE, 2003.
Define Goals/Preferences
Assign weights in pairwise comparisons (per level)
Evaluate Combinations
{ } { }x = { }Weighted
Not all Combinations are viable!
AMI
VM
ImageVM
Image
Evaluated set of combinations
CUMULUSGENIUS
IMPLEMENTATION
CumulusGenius
• Implementation of the model, evaluation
methods in Java [2]
• Basis for Experiments and future Tools
• jClouds for deployments on EC2
[4] available as java library: http://code.google.com/p/cumulusgenius
EXPERIMENTS & EVALUATION
Experimental Setup
• Employed CumulusGenius Implementation
• Generated Database of VM Images &
Compute Services
– Attribute values in plausible ranges
– Every combination viable
• All Criteria have same weight
• 20 Runs with growing Database size
AMI & Service Evaluation
• Service evaluation
has higher effort
• AMI & Service
evaluation
not growing linearly
Experiment Results
(avg. 20 runs)
• Non-linearly growing computation time
Evaluation
• Currently 10,000 AMIs on Amazon alone!
• Filtering important
• Fast evaluation algorithm
– Parallelization
– Heuristics such as Genetic Algorithms
Conclusion
• Framework for Migration of Web servers
– Cylcic Process
– Model
– Evaluation Methods
• Implementation CumulusGenius
– Java library
• Experiments regarding computation time
– Non-linear growing
• Improve attribute list
– Talk to experts (ongoing: German Telekom)
– Public prototype, evaluate feedback
• Apply & evaluate in real life migration
scenarios (prototype w/ GUI)
• Expand database of Cloud landscape
– Scan existing VM images for data
– Integrate existing databases
(cloudmarket, bitnami)
• Support more complex system setups
Future Work
Contact Me
For Questions, Discussions,
or Initiating Research Exchange:
Michael Menzel
Research Center for Information Technology (FZI)
Karlsruhe Institute of Technology (KIT)
Englerstr. 11
76131 Karlsruhe
Email:
Slides
• Made available on
http://www.slideshare.net/mugglmenzel
next week
• Made available on www2012 Website
MERCI FOR YOUR ATTENTION!
Questions, Comments, Discussion
DETAILS
Web server Migration Process
(Guidance)
Web server Migration Process – ctd.
Process supports
evolutionary Migration
Select Target Setup
Execute Migration
Set Preferences
Incorporate experience
Success!R
Model
Evaluate VM Images VM
Image
Evaluate Compute Services
Combining AMIs & Services
CumulusGenius: Web Frontend
CumulusGenius Suggester
Images Services
currentness of data?
Aotearoa Evaluation
ComponentUser
GWT
Frontend
CumulusGenius
LogicUser
Preferences
Data Collector
own
benchmarks
User
Ratings
jClouds
Deployments
Apache in the Cloud?
Prepare & Plan!