Dynamic Service Selection &

Dynamic Service Selection &
Analyze

Weina

Ma

June 6
th
, 2013

Paper 1

•
“Web Service Discovery with Implicit
QoS

Filtering”

•
Problem

–
Traditional WS discovery supported by WSDL and
UDDI API is insufficient

–
UDDI API is keyword
-
based query

•
This paper proposed a framework for WS
selection based on
QoS

properties
collected
by a distributed
agent
-
based

system.

Framework

•
Personal Agent (PA) accepts requests from client and send goal to Register.

•
Matching Agent provides list of candidate services which satisfy client’s
goal. It keeps history of requests and provides information searched for
the similar service before.

•
Service Mediator (SM) collects statistics about WS invocations.

•
PA

ranks the services and invokes the best one.

Conclusion

•
Agent
-
based technology is extremely suitable
for WS performance evaluation of cloud
environment.

•
Matching Agent can learn from experience.

Paper 2

•
“Web Service Competition: A New Approach to
Service Selection”

•
Problem

–
Service description is expressed inaccurate, globally
and statically, no depending on client’s needs and
contexts

•
This paper proposes a service selection approach,
using generic service representative technology,
to compare services based on their performance
in a specific client context
.

Methodology

Methodology Continue

Conclusion

•
Security and privacy challenges between
service provider and representative.

•
Increase network traffic.

•
Centralized competition desk fits comparing
performance of algorism, but not real
computing and traffic performance because of
the difference between WS running
environment.

Paper 3

•
“A Simple Approach for Testing Web Services
Based Applications”

•
Problem

–
A Web Application might invoke multiple web services
located on different servers with no design, source
code, or interface available.

•
This paper presents how to test web services
when building composited web application.

–
Two
-
level abstraction model to represent a web
application.

–
Three sets of test cases are generated automatically.

Web Application Representation

•
Task Precedence Graph (TPG)

•
Each

node represents a web service, like Main Component (MC), Hotel Reservation (HR), Car
Reservation (CR) and Weather Prediction (WP).

•
An

edge means the flow of actions (transitions)

•
Each

edge is labeled with an action and its timing constrain.

Single Web Service Representation

•
Time
d Labeled Transition System (TLTS)

•
TLTS represents

transitions from one state to another.

•
Node

means a state

•
Each edge is labeled with an action and its timing constrain.

Test Cases Generation

•
The first set of test cases are generated based
on boundary value testing analysis of WSDL

•
The second set of test cases are all possible
actions in each individual service, by
traversing all paths going from initial state of
TLTS to be fulfilled.

•
The third set of test cases are generated by
traversing all paths going from the initial state
of the TPG.

Web Service Testing Framework

Conclusion

•
Web service testing at stage of building
composited web application rather than
during invocation time.

•
Full
-
coverage test cases generated
automatically.


Paper 4

•
“Runtime Monitoring of Web Service Conversations”

•
This paper proposes to use
runtime monitoring

of
conversations

between partners as a means of
checking behavioral correctness
of the entire web
application.

•
Describe a specification language,
UML 2.0 Sequence
Diagrams (SDs)
, for dynamic analysis of web services.

•
Focus on monitoring finitely terminating behaviors and
capturing safety and
liveness

properties.


Sequence Diagrams

•
Sequence Diagrams (SDs) is to model
behavioral scenarios by describing sequences
of messages communicated between different
objects over time.

•
BasicSD
, par, alt, strict
seq
, weak
seq
, loop,
consider and ignore are terminal symbols.

•
E is a set of SD messages.



Formalizing Sequence Diagrams

•
NFA (Nondeterministic Finite Automaton) is a
finite state machine where from each state and a
given input symbol the automaton may jump into
several possible next states.

•
DFA (deterministic finite automaton) is a finite
that accepts/rejects finite strings of symbols and
only produces a unique computation (or run) of
the automaton for each input string.

•
An NFA A receive a trace of sequence diagram as
input and changes its current state according to
its transition relation

Architecture of Framework

•
User create UML SDs with help

of
PropertyM
.

•
Monitoring
M

converts NFA to DFA.

•
Message
M obtains interaction events from SCAM and directs to
MonitoringM
, in turn, update the state of every active monitor automaton,
until an error has been found or all partners terminate.

Conclusion

•
Not require any code instrument

•
Not significantly affect the performance of
monitored system

•
Requiring a distributed monitoring framework