Context, Quality and Relevance: Dependencies and Impacts on RESTful Web Services Design

Context, Quality and Relevance:
Dependencies and Impacts on RESTful
Web Services Design
Hong-Linh Truong1, Schahram Dustdar1 , Andrea Maurino2 ,
Marco Comerio2
1
Distributed Systems Group, Vienna University of Technology
2
Department of Informatics, Systems and Communication
University of Milano – Bicocca

truong@infosys.tuwien.ac.at
http://www.infosys.tuwien.ac.at/Staff/truong

Outline

 Motivation
 Quality, context and relevance in data-intensive
services
 Access to quality and context associated with
data resources
 APIs for specifying and obtained data quality
 Open issues
 Conclusions and future work

ComposableWeb's 10, ICWE 2010, Vienna 2

Motivation: compose data intensive
RESTful services
 Composition of Yahoo! Boss News Search, Google
News Search , and Flickr
 recent news and high-qualified images, but free-
of charge, related to "Haiti earthquake"


If the composer is aware of context
and quality parameters
 Possible mappings of context and quality
requirements

but it is a tedious task and we are not sure we have a
correct mapping

Motivation: access to large data
resources
 Retrieve big datasets from RESTful services for further
extraction, transform or mashup activities

http://www.undata-api.org/


Motivation: access to large data
resources

 Without QoD: get datasets and perform mashup
 With QoD support:
 http://localhost:8080/UNDataService/data/query/Population
annual growth rate (percent)?crq.qod:
 dataelementcompleteness= 0.8654708520179372,
datasetcompleteness=0.7356502242152466;
 http://localhost:8080/UNDataService/data/query/Adult literacy
rate (percent)?crq.qod:
 dataelementcompleteness=0.5874439461883408,datasetcompleteness
=0.04349775784753363
→ Should we retrieve the data and perform data mashup?

Context, quality and relevance for
data-intensive services
 Data-intensive services provide data resources whose
quality and context could not easily described at the
level of service as a whole
 We all know: missing context and quality leads to
irrelevant results
 But several questions:
 how can the composition developer recognize context and
quality description associated with data resources?
 how context and quality parameters can be mapped and
passed to a service via REST APIs?
 how to obtain context and quality description associated with
data resources so that further activities can be made?


Related work
 Well-researched QoS and context WS
 QoS-based and context-oriented Web service design
 QoS-based/NFP-based Web service discovery and
ranking
 User experiences and context models for
personalized Web services .
 However, existing works mainly deal with QoS and
context information
 QoD other data concerns are the key to data
resources
 Data resources have different data concerns
 Recent work on quality of mashup

Context, quality, and relevance
dependencies in data service
composition

 Data resources: (semi-)structured data, unstructured
documents, images, zipped datasets

Impact on the lack context and
quality – at data resource level
 Difficult to select relevant resources and to provide
general description of context and quality for the service
managing the resource.
 The service has to build its own quality and context
description
 difficult if the service developer is not the data
resource provider
 It is not clear if the resource will meet the consumer
quality description
 lead to the information overloading
 challenge for the service to determine the quality of
the data resource.

Impact on the lack context and data
quality – at data service level

 Not sure if the service can be used.
 Not sure if resources provided can be used.
 The composite service has to implement its own
service selection mechanisms.
 The composite service is not sure about the quality of
services and resources to be composed.
 The composite service has to determine the quality
and context of the resources at its side using its own
knowledge.


Access to data quality and context
associated with data resources
 Data quality and context associated with data
resources:
 Different with representative quality and context
associated with a service as a whole
 Current works:
 Describe context and QoS associated with a service
or a service operation but not with data resources
 We focus on mechanisms/convention to access
quality and context associated with data
resources
 Not how to determine quality and context metrics

Data quality and context support in
current RESTful WS design
 Many RESTful services do not provide data quality and
context information
 In RESTful service description, WADL or ATOM
 Allow specify individual resource but there is no
description for context and quality information.
 At the REST APIs level, there is no convention for
specifying and obtaining context and quality information.
→ provide context and quality description exchange
protocols among the interactions of service
compositions, services and resources


Data quality and context support in
current RESTful WS design
 RESTful WS design: PUT, GET, POST, DELETE data
resources but no support for non-functional properties
(e.g., context/quality)
 Possible option:

→ difficult to understand and manage the relationship between
the resource and its quality/context description
 We assume: non-functional properties are a part of data
resources → using the same operation for the same
resource id but with different parameters

Proposed design principles –
convention
 Focus on specifying and obtaining context and
quality associated with data resources
provided by services
 Static and dynamic context and quality binding
 Static: context and quality can be published with
resource description (e.g., WADL, ATOM)
 Dynamic: context and quality information obtained at
runtime
 We test this principle with
 WADL, a RESTful service provides the Google Flu
Trend data, a RESTful service wrapping the UN
datasets service

Representations for Context and
Quality
 We do not develop/propose new context and quality models
 Assume that the community will rely on existing models
 We use custom models from [APSCC 2009: 87-94]
{
...
"crq.qod": {
"crq.uptodateness": "up to dateness",
...
"crq.dataelementcompleteness":"data element completeness",
"crq.datasetcompleteness":"data set completeness",
"crq.domainspecificqod": [
{
"name": "name of the metric",
"value": "value of the metric"
},
]
},
…
}

Static quality and context information
for data resources WADL
specify context and quality schemas and their types using
grammars/include and doc/title
<grammars>
<include
href="http://www.infosys.tuwien.ac.at/prototype/SOD1/restfuldesign/schemas/crq-
quality.xsd">
<doc title="Quality">
grammar describing quality
</doc>
</include>
<include
href="http://www.infosys.tuwien.ac.at/prototype/SOD1/restfuldesign/schemas/crq-
context.xsd">
<doc title="Context">
grammar describing context
</doc>
</include>
</grammars>


for data resources with WADL
specify the published, static quality and context information using
representation
<representation id="QualityPresentation" element="crqQuality:crq.qod"
mediaType="application/json">
<doc title="QoDDescription">
{"crq.qod": {
"crq.datasetcompleteness": 0.10256410256410256,
"crq.consistency": 1,
"crq.uptodateness":5
}}
</doc>
</representation>
<representation id="ContextPresentation" element="crqContext:crq.context"
mediaType="application/json">
<doc title="ContextDescription">
{"crq.context": {
"crq.location": "Europe",
"crq.datapermission":"free"
}}
</doc>
</representation>


for data resource discovery with
ATOM-alike model
<cdc:category label="Demographic and socioeconomic statistics"/>
<cdc:id>http://undata-api.appspot.com/data/query/Adult literacy rate (percent)
</cdc:id>
<cdc:title>UN Data on Adult literacy rate</cdc:title>
<cdc:entry>
<cdc:category term="DaaSConcerns" scheme="...daasconcern-v0.2.xsd"/>
<cdc:content type="application/xml">
<tns:qod xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:tns="http:/
/www.infosys.tuwien.ac.at/SOD1/daasconcerns/custom" xsi:type="tns:QoD">
<tns:crq.dataelementcompleteness>0.5874439461883408</tns:crq.dataelementcomple
teness>
<tns:crq.datasetcompleteness>0.04349775784753363</tns:crq.datasetcompleteness>
</tns:qod>
</cdc:content>
<cdc:published>2010-06-23T22:23:30.557Z</cdc:published>
</cdc:entry>


Specifying and obtaining data context
and quality using parameters in REST
APIs
 Specifying requests by using utilizing query parameters
the form of crq.metricName=value
GET/resource?crq.accuracy="0.5"&crq.location=’’Europe”

 Obtaining contex and quality by using context and
quality parameters without specifying value conditions
curl http://localhost:8080/UNDataService/data/query/Population annual growth rate
(percent)?crq.qod
{”crq.qod” : {
”crq.dataelementcompleteness ”: 0.8654708520179372,
”crq.datasetcompleteness”: 0.7356502242152466,
...
}}

 Context and quality schema can be also accessed via
parameters

Open issues
 Data context and quality of data resources is a part of
the contract/licensing of the data resource?
 The use of parameters can be achieved but
 Context and quality metrics and representations are
diverse → how do we compare quality/context metrics
defined by different services
 We propose only convention/principle. To convince its
use we need:
 Use accepted terms, evaluate via real applications
 Perform possible different mappings
 Data concerns must more than data quality and context
 e.g., data privacy


Conclusions and future work
 Analyzed the importance of data context and quality
support in RESTful data-intensive services
 Limitations of current RESTful Web service design with
the respect to data context, quality and relevance
 Propose conventions to specify and obtain context and
quality associated with data resources
 Future plans
 Examine possible mappings and focus on validation
 Composition of context and quality of data resources
 Link data concern evaluation to data concern
publishing for data services


Thanks for your attention!

Hong-Linh Truong
Distributed Systems Group
Vienna University of Technology
Austria

truong@infosys.tuwien.ac.at
http://www.infosys.tuwien.ac.at


Context, Quality and Relevance: Dependencies and Impacts on RESTful Web Services Design

More Related Content

Viewers also liked (9)

Similar to Context, Quality and Relevance: Dependencies and Impacts on RESTful Web Services Design (20)

More from Hong-Linh Truong (20)

Recently uploaded (20)

Context, Quality and Relevance: Dependencies and Impacts on RESTful Web Services Design