This study introduces a framework for service oriented cloud computing with the focus on its service quality aspects in which most demanded by cloud users. It considered two dominant sub-layers such that functional and runtime; against cloud characteristics. SEQUAL model and recent literature were used to derive the quality constructs in cloud environment whilst the opinions of the industry experts were adding more to the blend. The proposed model gives proper identification of service quality expectations and also applicable for different business or geographical contexts. The validity of it for Sri Lankan context were evaluated in this study by using questionnaire based survey while PLS-SEM (partial least squares-structural equation modeling) technique was used to evaluate the outcome. Further the research findings shows the significance of functional layer is rather expected by user than runtime layer and prioritizes quality factors of each layer; in which Service time, Information and data security, Cost benefit, Service Transparency, SLA (Service Level Agreement) are some of them.
2. OVERVIEW
• Introduction
• Objective
• Literature Review
• Conceptual Model
• Collection of data
• Data Analysis
• Conclusion
• Limitation
• Recommendation
• Future Avenue
3. INTRODUCTION
Cloud computing is a model for enabling ubiquitous, convenient, on-
demand network access to a shared pool of configurable computing.
(The NIST Definition of Cloud Computing September, 2011.)
Service-orientation is a way of thinking in terms of services and service-
based development and the outcomes of services.
5. OBJECTIVE
• Identifying cloud service quality requirements at functional level
and Runtime level
• Identifying cloud computing attributes which are supported for
value creation
• Identifying most effective quality requirements at each phases
(Functional and runtime )
• Based on that, develop a service oriented quality requirement
Framework for cloud computing
6. LITERATURE REVIEW
Models / Frameworks Focus study
[S. Negaeshet.al., 2003]) SERVQUAL requirement model
Yu’s quality measurement
model for web service [Yu et.
al.,2006]
QoWs is categorized in to two, Business and run time and
separate quality factors has identified under each factor.
IAAS (Infrastructure as
services) Requirement of
cloud users [Rochwerger et.
al.,2009]
Focused on main quality requirements for the cloud services
at IaaS
7. LITERATURE REVIEW
Models / Frameworks Focus study
NIST model for cloud
characteristics
On demand Self- Service, Broad network access, Resource pooling,
Rapid elasticity, Measured services.
Yu’s Dimensional model for
Deploying & managing web
services [Yu et.al., 2006]l
Arsanjani ‘s layered SOA
architecture [Arsanjani A.,2004]
Interoperability, Web, Security/ Privacy, Quality of web, service
Management
IT M Framework on Cloud
Computing Environment
[Arabalidousti F.,et al.,2014]
Service management;
Security, resiliency , performance & consumption
Governance of IT
Brandis’s dimensional model for
cloud governance [Brandis
K.et.al.,2013]
Strategic alignment, Value delivery, Risk management, Resource
management, and Performance measurement
9. DIMENSIONS OF SOCC
(SERVICE ORIENTED CLOUD COMPUTING)
Dimension Indicators Variable name
Interoperable service architecture
(INT)
systematic interoperability INT1
semantic interoperability INT2
Cloud service management (CSM) Service provisioning management CPM3
Business and operational support
management
CSM4
Service measurement (SM) Service billing BIL5
Service monitoring SMN6
On demand self service (OND) Service provisioning capability SPR7
Shared Resource pooling (SRP) Multi-tenant model MLT8
capability of assigning Different
physical and virtual resources
RES9
10. DIMENSIONS OF SOCC
Broad network access (BNA) Access over the networks NET10
Access over client platforms CLP11
Rapid elasticity (RE) Number of versions released V12
Availability at any given time AVI13
11. FUNCTIONAL LEVEL QUALITY
REQUIREMENT INDICATORS
Dimension Factors Variable Name
User friendliness
(F_UF) - The physical features of the
system, such as whether the system is
appealing and looks good.
Attractiveness of the application F_UF1
Consistency F_UF2
Understandability F_UF3
Reliability
(F_REL) - focusing on whether the
system is right, useful, and dependable
Relevance F_REL1
Dependability F_REL2
Cost benefit F_REL3
Accuracy F_REL4
12. FUNCTIONAL LEVEL QUALITY
REQUIREMENT INDICATORS
Responsiveness (F_RES) - The
readiness of the service to provide service
Service time F_RES1
Assurance (F_AS) - The knowledge and
courtesy expressed in the system and its
ability to inspire trust and confidence in its
safety
Service Transparency (SLAs) F_AS1
Reputation F_AS2
Information security F_AS3
Completeness F_AS4
Sufficiency F_AS5
User orientation (F_UO) -
individualized attention
Customization of application F_UO1
13. QUALITY REQUIREMENT
INDICATORS AT RUN TIME
Dimension Factor Variable Name
User friendliness
(R_UF)
Throughput R_UF1
Number of active sessions
(concurrency level)
R_UF2
Resource allocation R_UF3
Redundancy R_UF4
Reliability (R_REL) Dependability R_REL1
Recoverability R_REL2
Responsiveness
(R_RES)
Response time R_RES1
Assurance (R_ASS) Availability R_ASS1
Accessibility R_ASS2
Data security R_ASS3
Technical support service R_ASS4
User orientation (R_UO) Integrity R_UO1
14. DATA COLLECTION
Pilot survey - 10 respondents
• Sample : Graduates from 2009 Batch from the department of Industrial
Management who has the industry exposure and the knowledge.
Main survey – 53 respondents
• Online questionnaire
• Focus on IT companies
Likert scale
Response Superior
Somewhat
satisfactory
About
average
Somewhat
unsatisfactory
Very
poor
Scale
value
5 4 3 2 1
16. DATA ANALYSIS
Areas of Expertise Respondents %
Software Engineering 43 81.13%
Network systems and data communications
analysis 11 20.75%
Network and computer systems administration 13 24.53%
Computer support specialization 22 41.51%
Business Analysis 7 13.21%
Database administration 8 15.09%
Software Quality Assurance 12 22.64%
Experience Level respondents %
< =1 year 8 15.09%
2- 4 years 26 49.06%
5 - 7 years 15 28.30%
above 7 Years 4 7.55%
17. DATA ANALYSIS
Software
Engineering
37%
Network
systems and
data
communicatio
ns analysis
10%
Network and
computer
systems
administration
11%
Computer
support
specialization
19%
Business
Analysis
6%
Database
administration
7%
Software
Quality
Assuarence
10%
Expertised
< =1 year
15%
2- 4 years
49%
5- 7 years
28%
above
7
Years
8%
Experience
18. DATA ANALYSIS –
SERVICE PROVIDERS
Cloud service provider Response %
Microsoft Azure 28 52.83%
IBM 13 24.53%
Google App Engine 21 39.62%
SAP HANA 9 16.98%
Amazon Web Services
(AWS) 5 9.43%
CISCO 3 5.66%
Mango Apps 2 3.77%
Other 2 3.77%
Microsoft
Azure
34%
IBM
16%
Google App
Engine
25%
SAP HANA
11%
Amazon Web
Services
(AWS)
6%
CISCO
4%
Mango Apps
2%
Other
2%
cloud service providers
19. DATA ANALYSIS
Service Type Response
Software as a service 52
Platform as a service 47
Infrastructure as a service 16
Other 0
Software as a
service
45%Platform as a
service
41%
Infrastructure
as a service
14%
Other
0%
Service type
20. STATISTICAL ANALYSIS
• Hypothesis starts with a causal model
• The model is tested against the obtained
data
• The operationalization then allow
testing the relationship between the
concepts,
Confirmatory
Modeling
• SEM (Structural Equation Modeling)
21. SEM TECHNIQUE : PLS (PARTIAL
LEAST SQUARES METHOD )
• Out of the SEM techniques Partial Least Squares (PLS) is the well-
established technique for estimating path coefficients in structural
models and has been widely used in various research studies [Gefen
et. al., 2000]
• Path model consists of three components:
• Structural model -
• inner model (Graphical model)
• Measurement model -
• The connections between LVs and MVs are referred to as measurement or
outer model
• Weighting scheme -
• estimation of the inner weights of the PLS algorithm
28. T- VALUES
Confidence level = 90%
Significance level: 0.1
The hypothesized paths of the constructs are considered to be significant if
t-value is greater than 1.6747
All t values > 1.67
T Statistics (|O/STERR|)
Cloud Dimension -> Service Quality requirement 1.8223
Service Quality requirement -> Functional
requirement 60.523
Service Quality requirement -> Runtime
Requirement 22.002
Cloud Dimension -> Functional requirement 1.8115
Cloud Dimension -> Runtime Requirement 1.8363
29. MAIN PATH
COEFFICIENT
significance P value <0.01
If path coefficient value is > 0.01 , the model is accepted
Therefore null hypothesis is rejected
H1: Runtime Level quality requirements
are positively related to Service
Quality requirements of users
1.3253 Accepted
H2: Functional Level quality requirements are
positively related to Service Quality
requirements of users
2.4538 Accepted
H3:
main
Service Quality requirements are positively
related with SOCC service dimensions.
0.1064 Accepted
30. PATH COEFFICIENT OF
DERIVED MODEL
Quality
requirement
Functional
level quality
requirement
Quality
requirement
at runtime
SOCC
dimensions P = 0.1064
P=2.4538
P= 1.3253
Significance P value <0.01
Significance T value <1.674
T= 22.002
T= 1.823
T= 60.523
T= 1.8115
T= 1.8363
31. ANOTHER FINDINGS
Individual behaviour of each Y variables against SOCC dimension
The individual T values show strong relationship with each functional and
runtime
Strength
>Functional Runtime
32. CONCLUSION
The impact of functional level quality is more higher than Runtime
level quality requirements, when determining service quality of cloud
services
Service quality of
cloud services
Functional
level quality
requirements
Runtime level
quality
requirements
33. CONCLUSION
The importance of quality attributes of cloud services at functional level.
Accuracy
Relevance (service alignment
Attractiveness of the
application
Sufficiency
Cost benefit
Service Transparency SLA
(Support Service)
Completeness
Information security
Service time
Functional level
Quality
requirement
Priority
34. CONCLUSION (CONT.)
The importance of quality attributes of cloud services at
Runtime level.
Response time
Throughput
Availability
Integrity
Accessibility
Recoverability
Data security
Quality
Requirements
at Runtime
Priority
35. CONCLUSION
• The relationship between Functional level quality
requirements and SOCC dimensions is stronger than ,
• the relationship between Quality requirements at runtime level
and SOCC dimensions.
• Therefore the functional level quality requirements can make
great impact on overall service quality of cloud services
36. LIMITATION
• The perception of service quality requirement can be differed
from its spectrum of the domain where the cloud services are
used in
• The collection of data is more bias in geographical aspects
because of the unfeasibility of achieving
• The uncertainty of the total population of cloud users in Sri
Lanka
37. RECOMMENDATION
• This framework can be used as benchmark to assess the service
discrepancies of cloud services in both perspectives; Provider
and the user
• Minimize the service gap by standardizing the quality as to
improve the satisfaction level of users.
• The outcome of the research may valid for next 5 years; since
the avenue of the Service oriented cloud technology has
continuous improvements in advanced.
38. FUTURE AVENUE
• Lead the analysis on domain wise such as development,
manufacturing, service sector…etc.. .Then the prioritization of
variables can be attributed based on domain
• The improving the framework in the perspective of IT governance
which includes assessment of continuous performance and
Continuous Quality management and improvement
• Thereby it is possible to develop ontology for the purpose of quality
assurance of cloud implementations
Ensure that enterprise IT performance and conformance measurement and reporting are transparent, with stakeholders approving the goals and metrics and the necessary remedial actions and SLA objectives between the service provider and the cloud service provider.
weighting scheme -
Estimate each LV as a weighted sum of its neighbouring LVs
Bootstrapping procedure was adopted with 200 bootstrap samples in 100 iterations to obtain the statistical significance of path coefficient estimates. Figure 4-10 shows the path coefficient estimates and the level of statistical significance in our empirical model. The bootstrapping technique is used in this research. At the 0.01 significance level, the hypothesized paths of the constructs are considered to be significant (t-value is greater than 1.6747), according to the calculated data.