=Paper=
{{Paper
|id=Vol-421/paper-12
|storemode=property
|title=External Behavior Modeling Enrichment of Web Services by Transactional Constraints
|pdfUrl=https://ceur-ws.org/Vol-421/paper12.pdf
|volume=Vol-421
|dblpUrl=https://dblp.org/rec/conf/icsoc/Khebizi08
}}
==External Behavior Modeling Enrichment of Web Services by Transactional Constraints==
https://ceur-ws.org/Vol-421/paper12.pdf
External Behavior Modeling Enrichment of Web Services
by Transactional Constraints
Ali Khebizi
Computer Science Doctoral School
8 mai 1945 University - BP 401 Guelma 24000 - Algeria
kheali@hotmail.com
Supervised by: Hassina Seridi
Badji Mokhtar University, BP 12 Annaba 23000, Algeria
LabGed laboratory. seridi@lri-annaba.net
Abstract.
The current description of services basing on their interfaces and their protocols
remains limited and does not include all the engaged interactions properties in a
Web service environment. For a real deployment and a broad adoption of Web
services technology, service protocols must be enriched by other properties
related to interactions nature between services.
In this paper, we present a model for representing transactional effects and
service protocols description improved by injection of the proposed effect
model. The protocol compatibility and the equivalence analysis will be
reviewed in the light of the proposed enhancements.
Keywords: Web service protocols, Transactional effects, Compensation,
Compatibility analysis, Services equivalence.
1 Introduction
The current description of Web services basing on their interfaces and protocols
remains limited and does not represent all the semantics of the interactions involved
during services invocations. Indeed, current service protocol modeling is taking into
account various characteristics which describe its external behavior (such as: order
and time constraints) [1][2]. In addition, the current infrastructure of Web services
(SOAP, WSDL, and UDDI) is enriched by specifications needed to manage
transactions and coordination at Middleware level (such as the Frameworks: WS-
Coordination [3] and WS-Transaction [4]). However consideration and management
of transactional constraints in services protocols are not given the interest they
deserve.
For an actual deployment and broad adoption of this technology, service protocols
modeling require other enhancements. In this paper, we propose an enrichment of the
external behavior of Web services description by taking into account the transactional
constraints. We will also strive to study the conceptual consequences of this
enrichment on compatibility and equivalence analysis.
The paper is structured as follows: In section 2, we will explain the motivations of
this work. A state of the art of transactions management in service protocols is
presented in section 3. In section 4, we will propose a model for representing
transactional effects and we present the involved protocol model. Section 5 will be
devoted to analyzing compatibility and equivalence of protocols enriched with
transactional effects. Finally, we conclude and present our future works in section 6.
2 Motivations
Because of their excessive cost and their relatively long time, transactions in Web
services differ in their effects and their cancellation process from those related to
traditional databases. Therefore, service providers are rather compensating
transactions and affect often a part of the costs associated to customers [1]. The
compensation is provided by the middleware transparently by execution of the
compensation protocol predefined by the service developer [5]. In this context, it is
appropriate to address a profound reflection on the following issues: How to describe
and model transactions effects and their compensation effects? How to model service
protocols taking into account the transactional effects? What impacts will bring the
injection of transactional constraints in service protocols to the analysis of service
compatibility and equivalence?
Only a few studies exist on this issue that requires more in-depth research efforts. In
addition to answering previous questions, the following reasons motivate this work.
a. Services compatibility verification: the compatibility definition of two protocols
(fully or partially), as described in [1], restricts the test criterion to operations order
and to messages polarity. It has been extended to take into account the time
constraints (time, date) [2]. The compatibility of two services protocols must take into
account transactional aspects related to messages and their effects.
b. To infer transactional properties for existing scenarios: given BPEL programs
availability, it is appropriate to be able to extract their transactional properties for
their analysis and their manipulation. In this perspective, a BPEL program is analyzed
to extract its transactional properties. Indeed, elements of management errors and
transactions, such as: , and constitute activities blocks related to transactional properties that must be
recovered and modeled for their possible treatment.
c. Protocol consistency checking (Design Tools): In a compensation situation, it is
imperative to check whether the compensation protocol is consistent with the trigger
one or not? i.e.: the compensation guarantees – indeed- a "semantic cancellation" of
observed effects? Transactions effects modeling and related services protocols
management will provide a sound conceptual framework to check the consistency of a
protocol for compensation with the trigger one.
3 Transactions management in services protocols state of the art
The state of the art of transactions management -at the protocol level- highlights four
approaches that deal in a more or less rigorous way this aspect. In Protocols
languages modeling, WSFL and XLANG languages provide extensions to the
WSDL standard, offering composition and coordination structures of services based
on rules. However, no model is provided for distributed transactions management and
transactions compensation is discussed in relation to data flows manipulation. This
would require a considerable programming effort. In Web transactions protocols,
the current Web services specifications are relaxing the ACID properties and
strengthening mechanisms for compensation [5]. However, the majority of the
proposed specifications don’t deal neither with the concept of transactional effect nor
with the compensations management. Both protocols have dealt with this issue are:
Business Transaction Protocol (BTP) [7] and Tentative Hold Protocol (THP) [8]. In
BTP, transactions effects are covered in three dimensions: provisional effects,
counter-effects and final-effects. However, specification of effects types remains
manual and specific to the engaged coordination. In addition no mechanism to ensure
counter-effects consistency with effects is presented. THP is based on the reservation
and allocation principle of the current transaction resources by manipulating the
concepts: attempt, non-blocking and holds reservation. The cancellation and
compensation process are then significantly reduced. But the protocol remains limited
in managing the transactions effects and their manipulation. Furthermore, customers
have no idea on resources they will need during the activities evolution. Development
environments of business Web services (Enterprise Java and XML transactions)
suffer from the shortcomings due to the lack of conceptual models for representing
and manipulating transactional effects. Therefore, no mechanism can verify that
observed effects in the real world are, really, those desired apart from traditional
testing suites/scenarios. In addition, compensation is discussed in terms of a new
process to execute. The Web Service Transaction model (WSTx) [9] proposes a
WSDL language extension to describe the customer and provider’s transactional
behavior. However, it suffers from a deficit in modeling effects and proposes only a
WSDL operations type classification following transactional criterion.
To address the deficit in effect modeling and compensation management we will
propose, in what follows, a formal model for representing transactional effects which
is injected thereafter in the service protocol model.
4 Modeling effects and their impact on service protocols model
The Table 1 summarizes models characteristics that bring effects representation on
Web services. It presents a comparison on the basis of a set of criteria, highlighting by
this their strengths and weaknesses.
Model OWL-S [10] BPEL Colombo [11]
Criterion
Concepts Ontology, Classes, Activity, Variables, Database Query
Effects ServiceProfil, Scope Compensation Updating, Universal
ServiceModel, Relation
Formel Meta Model Logic description Language Relational Model
Concept of effect Yes No Yes
Concept of State Yes No Yes
Compensation handling No Yes No
Formel Model for effects No No No
Table 1: Different models representing effects Comparison
Based on the perception of effects such as query for updating the database, the
Colombo model [11] offers advantages in effect and state concepts mastery. However,
it is still failing in the management of compensating transactions and does not allow
comparative effects manipulation.
We will adapt Colombo principle model for representing transactional effects. Indeed,
in our model, effects and their compensation effects are considered as requests to
update the database. Thus, a message of a service protocol will impact on the real
world of a customer by executing a request to update database by type: Insert (R),
Delete (R) or Modify (R), where R is the record of the database reflecting the impact
of the message on customer world. The transactional effects managing problem is
reduced accordingly to that of handling query, as shown in Table 2.
Transactional effect management problem The corresponding Query management
Problem
Checking equivalence and difference of effects Comparison of updating query
Finding the compensation effect for compensating an Search a query for cancellation after
effect elementary effects for complex effect
Finding decomposition
Queries decomposition
Cumulated effects for a complete execution path Sequence of queries
Cumulated effects for compensation Search a query sequence for compensation
Checking transactional effects protocols equivalence Comparing equivalence of query sequences
Checking transactional effects protocols compatibility Comparing sequences of query
Table 2: Transformation of effects management problem to updating queries problem
Taking into account transactional effects allows a rich representation of interactions
reality in Web services. Indeed, a message will be characterized, in addition to its
polarity by effects. It creates in the customer world, as well as compensation effects
involved. Compensation effects are represented jointly with observed effects, in order
to express the fact that service providers implement charges that differ even if effects
are the same. This performance reflects the reality on the diversity of logic
compensation which is specific to each provider.
New structure of message for service protocols enriched by the transactional
effects: According to our model, at each message is associated a request to update the
database and its corresponding complaint related to compensation effects. The new
structure of a message is described as follows: m (p, e, e'), where:
m: Refers to the message and its polarity p (+,-) as the message is input or output [1]
e: All effects observed in the customer world. This is a request to update the database.
e': All effects of compensation to defeat semantically the effects e. This is a request
for updating the database to cancel the effects e while applying charges imposed by
the supplier and relating to the transaction cancellation.
This modeling express in a formal way (relational queries) the effects of transactions
and compensation effects for each message of service protocol.
Formal model of service protocols enriched by the transactional effects:
Integrating the new structure of the message in the basic model of service protocols
[1], will result in an overhaul of protocols model modeled with deterministic finite
state machine. The new model protocol (transactional effects protocol) is described
by the tuple: P = (S, s0, F, M, R, Sb) where:
S: A finite set of states; s0 Є S is the initial state of the protocol;
Sb: state of the database associated to each state of protocol;
F: The set of final states machine, with F ⊂ S; M: a finite set of messages, we
associate to a message m two types of effects e and e', which correspond, respectively,
the requests Ri and Rj for the database updating.
R ⊂ (S x Sb) x M: Transitions set. Each involves a state source, which is associated a
2
database state, to a target state with its database state, following the message receipt.
It should be noted: R((s, sb),(s', s'b),m) instead of ((s,sb), (s',s'b),m) Є R.
In addition, an effect function is defined, for each message allowing the transition
from one state to another (with their database states), combines effects (in terms of
requests) and compensations effects (corresponding requests).
5 Transactional effects protocols’ compatibility and equivalence
analysis
Service protocols compatibility and equivalence analysis as specified in [1] [2] should
be revised in the light of proposed enhancements. Indeed, transactional effects
representation will be exceeded qualitative analysis, beyond its simple syntax and
structural aspect. It will be richer because it is based on semantics of transactions seen
in terms of messages effects in the real world. In addition, modeling compensating
effects in conjunction with observed effects representing a message by related
attributes (e, e') expresses perfectly the real situations in which suppliers combine
compensation effects for each observed effect.
Transactional effects protocols compatibility:
Transactional effects protocols compatibility differs from that of basic protocols, due
to effects induced by messages. Indeed, two compatible protocols in the basic model
[1] may not be in the new context. In this sense, an interaction between two services
protocols is allowed only if observed effects will be compatible. By compatibility
effects, we are presuming that complaints updating request at the databases have the
same type (Delete, Insert or Modify). This condition implies an interaction path
concept redefinition to be extended to query type, as follows:
((State1.State2).Message.QueryType)*
This extension will ensure -when analyzing- verifying the compatibility of updating
query and will promote a richer specification of interaction protocols between the
candidates. Thus, two service protocols may be compatible only if queries -or
sequences of queries- associated to messages would be compliant.
Transactional effects protocols equivalence:
After studying various scenarios, we concluded that transactional effect protocols
equivalence is conditional on final states equivalence of the two databases witch is
considered on the basis of sequence equivalence of query updating.
We have identified two equivalence types for queries sequences: strict equivalence
and converging equivalence.
Strict Equivalence: For each message m of a protocol P1 corresponds to the
corresponding message in the protocol P2, exactly one query that it is equivalent: i.e.
it has the same type (Insert, Delete, and Modify).
Bases states’ converged equivalence: In this case, we are interested in the queries
sequence of the complete execution paths. For each query sequence associated to a
complete execution path of a protocol P1 corresponds to the same path in P2, another
sequence of equivalent queries. This leads to a convergence of updates inducing
databases final states which are equivalent.
The two equivalence types induce two equivalence classes for transactional effects
protocols: strict equivalency Class and bases states’ converged equivalence Class.
The second equivalence class is of particular interest because it expresses a way of
achieving differently from the service providers while leading to identical databases.
6 Conclusion
In this paper, we highlighted the interest of transactional constraints modeling. These
constraints have been perceived as effects affecting the customer world and have been
analyzed in the context of their compensation. We proposed a model based on query
for updating databases for representing transactional effects. The enriched service
protocol model was presented and formalized. The second contribution is on the
compatibility analysis formalization and study of transactional effects service
protocols equivalence.
As future work, we plan to identify the compatibility types and to study the
algorithmic aspect. We intend, moreover, the proposal for a set of operators handling
transactional effects.
References
1. B. Benatallah and al : Representing, Analysing and Managing web Service Protocols.
Data Knowledge Ingineering. 58 (3): 327-357, 2006
2. J. Ponge and al: Fine-Grained Compatibility and Replaceability Analysis of Timed
Web Service Protocols. ER 2007: 599-614
3. F. Cabrera and al. Web Service coordination (WS-coordination), August 2005
4. F. Cabrera and al. Web Service transaction (WS-transaction), January 2004.
http://dev2dev.bea.com/pub/a/2004/01/ws-transaction.html/
5. Gustavo Alonso, Fabio Casati, Hurumi Kuno,Vijay Machiraju : Web services
concepts Architectures and applications, Edition Springer Verlag Berlin 2004
6. Web Services Business Process Execution Language Version 2.0 OASIS Standard, 11
April 2007, http://docs.oasis-open.org/wsbpel/2.0/OS/wsbpel-v2.0-OS.html/
7. OASIS Business Transaction Protocol (BTP),
http://oasis-open.org/committees/business-transactions/
8. J. Roberts , K. Srinivasan. Tentative hold Protocol Part 1: White Paper, W3C Note 28
November 2001, http://www.w3.org/TR/tenthold-1/
9. T. Mikalsen and al, Transactional attitudes: Reliable composition of autonomous
Web Services, WDMS 2002.
10. D. Martin and al, OWL-S: Semantic Markup for Web Services, W3C Submission
November 2004, http://www.w3.org/Submission/2004/SUBM-OWL-S-20041122/
11. D. Berardi and al, Automatic composition of Transition-Based Semantic Web
Services with Messagging, Proceeding of 31st VLDB conference, Trondheim,
Norway, 2005