This presentation had place at the 4th International Workshop on a Research Agenda for Maintenance and Evolution of Service-Oriented Systems. We proposed an adaptive approach to manage context information that can affect the accomplishment of SOA governance objectives
1. MESOA - Sept. 17, 2010
Timişoara - Romania
Context-Driven Adaptive Monitoring for
Supporting SOA Governance
Norha M. Villegas
Hausi A. Müller
Rigi Group - Computer Science Department
University of Victoria, Canada
y ,
nvillega@cs.uvic.ca hausi@cs.uvic.ca
http://webhome.csc.uvic.ca/~nvillega/ http://webhome.cs.uvic.ca/~hausi/
2. 2
MESOA - Sept. 17/2010
Research Questions
Service-oriented applications are highly dependent on
environmental information
How can we represent the context that can affect the behavior of
a service-oriented system?
How can we change this representation at run-time as required
by changes in business objectives and the environment?
How can we manage the relevant context under changing
requirements?
3. 3
MESOA - Sept. 17/2010
Outline
1. The research 2. Challenges in
problem and the dynamic
3. O contribution
Our t ib ti
promise of service- monitoring for SOA
oriented computing governance
6. Ongoing research
4.
4 Application 5.
5 Our Approach
and summary
4. 4
MESOA - Sept. 17/2010
Dynamic Monitoring of Service-Oriented Systems
5. 5
MESOA - Sept. 17/2010
Service-Oriented Computing: its
visionary promise
Creating dynamic Spanning
and agile business organizations and
processes business platforms
By assembling components into loosely
coupled network of services
Papazoglou et al., Service-oriented computing: state of the art and research challenges. IEEE
Computer, 2007
6. 6
MESOA - Sept. 17/2010
Complex Dynamics in Service Composition
• Service composition must understand and respect or is
affected by:
y
▫ Policies
▫ Performance levels Context
▫ Security requirements Monitoring is
▫ Service level agreements Key!
▫ User’s matters of concern
Dynamic service composition is affected by environmental information
that characterize the situation of users, services, computational
, , p
platforms and businesses
7. 7
MESOA - Sept. 17/2010
Some MESOA Research Challenges
• Design for context-awareness
• Automated governance Context is
• Run-time monitoring
R i i i
• Dynamic service discovering
everywhere!
• Dynamic service composition
• Adaptive maintenance and evolution practices
• Mapping between business process and services
• Run-time V&V (compliance with design and run-time constraints)
Lewis et al., “MESOA 2009: Proceedings of the 3rd international workshop on a research agenda for maintenance
MESOA
and evolution of service-oriented systems,” CMU/SEI, Tech. Rep. CMU/SEI-2010-SR-004, 2010.
8. 8
MESOA - Sept. 17/2010
Dynamic Context Monitoring for SOA Governance:
Research Challenges
SOA governance needs to manage context information to
ensure the execution and evolution of service-oriented
systems
Relevant context must be Monitoring requirements are
specified at design-time but dynamic, the relevant context
managed at run-time is not the same over time
Distributive
Business
B i User’s
U ’ New
N service i Different
Diff t
nature of
objectives concerns interactions SLAs
services
9. 9
MESOA - Sept. 17/2010
Dynamic Monitoring Instrumentation
Service-oriented monitoring infrastructures are
required to adapt themselves to address dynamic
monitoring objectives according to changes in business-
g j g g
level objectives and processes, user’s concerns and the
environment
10. 10
MESOA - Sept. 17/2010
Optimizing SOA Governance
SOA governance requires effective mechanisms to
control the accomplishment of governance objectives
l h li h f bj i
under changing environmental conditions
Context- Adaptive Dynamic
Awareness Sytems Monitoring
11. 11
MESOA - Sept. 17/2010
Our Contribution: Dynamic Monitoring
for SOA Governance
A feature based model to represent relevant
feature-based
context and monitoring requirements
Feedback loops are first level
An adaptive components
control-based
reference
Monitoring objectives keep
architecture to
relevance with respect to
implement governance objectives
dynamic context
monitoringg
New monitoring strategies
infrastructures defined at run-time
12. 12
MESOA - Sept. 17/2010
Through Run-Time Re-Negotiation of SLAs
13. 13
MESOA - Sept. 17/2010
Run-Time Re-Negotiation of SLA
R Ti R N ti ti f SLAs
Minimum 10 trans/sec summer
transaction
rate 5 trans/sec the remainder of the year
SLA: Minimum transaction rate
SLO: throughput
14. 14
MESOA - Sept. 17/2010
Run-Time Re-Negotiation of SLAs
The dynamic re-negotiation of
y g
SLAs should be performed
according to current situations:
g
Changes in the hotel occupancy
g p y
rate
17. 17
MESOA - Sept. 17/2010
Context-Aware Governance Feedback
Loops
Müller et al., “SOA governance optimizes the business and evolution of service-oriented systems,” In proceedings of
MESOA 2009 CMU/SEI, Tech. Rep. CMU/SEI-2010-SR-004, 2010.
18. 18
MESOA - Sept. 17/2010
Feature-Based Context Meta-Model
1. Our context taxonomy
Minimum 10 trans/sec summer
transaction
rate 5 trans/sec the remainder of the year
N. M. Villegas and H. A. Müller, “Managing dynamic context to optimize smart interactions and services”, Springer
LNCS, vol. 6400, 2010, to appear.
19. 19
MESOA - Sept. 17/2010
Feature-Based Context Meta-Model
2. Features to guide context modeling and
management
N. M. Villegas and H. A. Müller, “Managing dynamic context to optimize smart interactions and services”, Springer
LNCS, vol. 6400, 2010, to appear.
21. 21
MESOA - Sept. 17/2010
Control-Based S i Component
C t l B d Service C t
Reference Architecture
22. 22
MESOA - Sept. 17/2010
Control-Based S i Component
C t l B d Service C t
Reference Architecture
23. 23
MESOA - Sept. 17/2010
Context-Model Controller Architecture
Inference Inference Definition
of context of of
Context control monitoring monitoring Monitoring
Model objectives
j strategies
g p
plans Plan
24. 24
MESOA - Sept. 17/2010
Context Monitor Architecture
C t t M it A hit t
Deployment Correlation
Monitoring Plan
g of the Gathering Handling of context Symptoms
y p
strategy
t t facts
f t
25. 25
MESOA - Sept. 17/2010
Context Sensing Architecture
C t t S i A hit t
Context sensing Context Sensor Pre- Context
request gathering discovery processing observation
27. 27
MESOA - Sept. 17/2010
Case Application:
Concrete
architecture for
the monitoring
infrastructure
required by the
case application
28. 28
Rigi Research Group
Related Work
• Current SOA monitoring approaches are not context aware (e.g., IBM-
IBM
WSLA)
• There is not many evidence of monitoring approaches able to change
dynamic monitoring strategies at run-time (e.g., approaches surveyed by
Bianco et al., Villegas and Müller, and the Petals European Project)
• A
Aspect-oriented approaches provide partial d
t i t d h id ti l dynamicity (
i it (e.g., D
Dynamo of
f
Baresi and Guinea’s)
▫ Monitoring assertions are generated at compilation-time or load-time
▫ Once at run-time, logic monitoring is statically integrated into business logic
, g g y g g
• Current approaches focuses on requirements for the adaptation of the
managed system rather than on the adaptation of the monitoring strategy
itself
it lf
29. 29
MESOA - Sept. 17/2010
Ongoing Research
Definition of case studies from industry to validate the approach
(e.g., IBM-CAS)
Design-time representation
Evaluation of the feature-based model
approach to represent context and Run-time modification
monitoring requirements
Support f user modification
S t for difi ti
Apache Tuscany
Evaluation of available technology to
support the modification of IBM WebSphere application server
monitoring plans at run-time
FraSCAti
Evaluation of existing context
g
COSMOS
management frameworks
30. 30
MESOA - Sept. 17/2010
Summary
• We proposed a f
d feature-based reference model to guide the representation of
b d f d l d h f
context information and monitoring requirements for SOA governance
▫ Represented at design-time
▫ Managed at run-time
• We proposed a control-based reference architecture to guide the
implementation self-adaptive monitoring infrastructures
▫ Based on feedback loops to control the adaptation process
▫ Able to implement dynamic monitoring strategies deployable at run time
run-time
according to SOA governance objectives
• The next steps focuses on
▫ The validation of models to represent context at design time and then
design-time
manage it at run-time
▫ The validation of available technologies for implementing dynamic
monitoring strategies
▫ The implementation and validation of the monitoring infrastructure to be
applicable in industry
31. 31
Rigi Research Group
Thank you!
h k !
Questions and discussion
University of Victoria, aerial view