This document provides reviews of 3 research papers on distributed systems. The reviews were created by following a structured format including the paper title, authors, main idea, results, impact, evidence, prior work, and ideas for future work. For the first paper, the summary discusses analyzing the cost and resource optimization of running real-life applications on an open source cloud. The second paper proposes a software testing framework called IVRIDIO to provide test-first performance as a cloud service. The third paper presents a formal approach to developing fault tolerant distributed systems using refinement techniques.

1. Assignment on
Paper Review of 3 Research Papers
Distributed Systems
CSE 601
Submitted by
Nadia Nahar BIT0327
Submitted to
Dr. Kazi Muheymin-Us-Sakib
Associate Professor
Institute of Information Technology
University of Dhaka
Submission Date
7th November, 2013

2. From the given papers, my selected papers are –
i. An Empirical Investigation of Cost-Resource Optimization for running Real-Life
Applications in Open Source Cloud
ii. IVRIDIO: Design of a Software Testing Framework to Provide Test-first Performance
as a Service
iii. Developing Fault Tolerant Distributed Systems by Refinement
I have thoroughly read these papers and created review for each of them. For creation of the
reviews I looked for answers of several questions and combined them in my paper reviews.
The question were –
1. What is the problem?
2. Why this is a problem?
3. How author(s) solved the problem?
4. What that author has achieved?
5. What the author did not mention/consider?
After finding the answers of these questions, I have created several sections to include these
answers and the sections are –
1. Paper Title
2. Author(s)
3. Idea (question 1)
4. Result(s) (question 4)
5. Impact (question 2)
6. Evidence (question 3)
7. Prior Work
8. Criticism (question 5)
9. Ideas of Future Work
In the next section, the reviews of the selected papers are presented.

3. Paper Review #1
Paper Title
An Empirical Investigation of Cost-Resource Optimization for running Real-Life
Applications in Open Source Cloud
Author(s)
1. Asif Imran
2. Alim Ul Gias
3. Kazi Sakib
The authors’ affiliation:
Institute of Information Technology, University of Dhaka, Bangladesh
Idea
The main idea of the paper is to find out the cost-resource benefits of executing real
life applications on open source cloud platform. These benefits are measured by evaluating
the empirical data.
Result(s)
The research has revealed that the cloud features play key role to improve cost-resource
performance. The experiment says, memory utilization is seen to increase by
22.16% and direct-indirect cost savings sum up to 40% in the cloud.
Impact
The extent to which cloud can reduce memory wastage and save costs is an important
issue of research. No analysis was held before to measure performance of open source cloud
using practical scenario and data. So, the analysis of empirical data to measure the
performance of open source cloud is needed, which was the intention of the paper being
reviewed.
Evidence
The comparison of five real life applications running on the cloud with traditional
systems was done to find out the cost-resource performance. The metrics they used to
identify the performance are Resource Utilization (RU), Resource Wastage (RW) and Direct-
Indirect costs (DC and IC).
Prior Work
Previously, researches were held to identify performance in terms of CPU and
memory utilization for commercial cloud services. The processor performance for executing

4. heterogeneous and complex scientific calculations was also identified. Reduced maintenance
cost of proprietary cloud services like Amazon Elastic Cloud Compute and Microsoft Azure
were illustrated previously on a paper, “Cost-Benefit Analysis of Cloud Computing versus
Desktop Grids”. And so, similar analyses for open source cloud computing were necessary to
expedite the power of open source cloud applications. This paper is intended to do so by
following the previous works on cloud computing.
Criticism
The paper is well-presented in the sense that it has maintained clarity of presentation.
The information is well structured and the entire paper follows the exact formation to attract
reader. The significance of the idea behind the paper is also good. However, the paper could
become more understandable if the difference between the open source cloud and proprietary
cloud services were cleared. The paper mentioned that empirical analysis on proprietary
cloud services was held and so similar analysis was needed for open source cloud. But why
similar analysis was needed for open source cloud and how analysis on open source cloud
was distinguishable from analysis on proprietary cloud services is unclear. So my suggestion
is to include a single line justification of this, which will enhance the paper understandability.
Otherwise, the paper has a good quality and the authors are appreciated for their nice
presentability.
Ideas for further work
In this paper the research only concentrates on IaaS (Infrastructure as a service). So it
is mentioned here that there is scope to consider PaaS (Platform as a service) and SaaS
(Software as a service) for further researches. But recently some other cloud architectures
also joined the cloud service model, which are NaaS (Network as a service) and STaaS
(Software Testing as a service). As a result researches need to be done on these architectures
also.

5. Paper Review #2
Paper Title
IVRIDIO: Design of a Software Testing Framework to Provide Test-first Performance
as a Service
Author(s)
1. Alim Ul Gias
2. Asif Imran
3. Rayhanur Rahman
4. Kazi Sakib
The authors’ affiliation:
Institute of Information Technology, University of Dhaka
Ramna, Dhaka–1000
Idea
The main idea of this paper is to propose the design of a software testing framework
to provide Test-first Performance (TFP) as a Service to be incorporated in the existing cloud
testing frameworks.
Result(s)
This paper introduces a testing framework named IVRIDIO which offers TFPaaS
(Test-first Performance as a Service). IVRIDIO adds PTFPC (Plugin for TFP in the Cloud) in
the client-end which allows the developers to execute critical test suites along with the
regular development phase. The framework also incorporates the paradigm Convention over
Configuration (CoC) to illustrate its effectiveness in reducing overall test effort.
Impact
The impact of his paper can be great on the cloud as well as the software testing
domain. Though many research papers addressed cloud testing frameworks, none specifically
addressed TFP to be provided as a candidate to be executed in the cloud. This paper proposes
the design of a cloud testing framework IVRIDIO which extends D-Cloud and AGARIC, the
existing testing frameworks. This framework IVRIDIO can be of a great significance as it
allows the developers to execute critical test suites along with the regular development phase,
provides a fast feedback, reduces overall test effort and most of all decreases test costs.

6. Evidence
The framework is demonstrated briefly in the paper which focuses on the structural
components of the framework as well as the interaction between them. The structure and
interactions are demonstrated using visual drawings and also in written format. All the tasks
done by each component are discussed broadly and so the whole scenario of TFP in IVRIDIO
(from client’s IDE to Test Run Center) is illustrated in the paper.
Prior Work
Previously, Lian Yu et al. developed a prototype of Software Testing as a Service
(STaaS) on cloud. Then, Cloud-based Performance Testing System (CPTS) for web services
has been presented. Subsequently, some software testing frameworks like D-Cloud,
AGARIC, Cloud9 were introduced to enrich the cloud services. But none of these
frameworks focused on TFP which is an important testing paradigm. And so these prior
works led to this paper which filled this vacuum by offering IVRIDIO.
Criticism
The research paper is really a good piece of work. The significance of its idea is high
and worthwhile. The presentation of the idea is also good and well-structured. The
demonstration of the design of the framework is neat and clear. The only point I would like to
suggest is that, the title of the paper does not contain the broad domain of the problem. Only
this slight modification is required. Otherwise, the paper holds a good quality and a great
significance.
Ideas for further work
A future challenge evolving from this research is to design a plugin which will
support different cloud based testing services. The plugin proposed in IVRIDIO is
specifically designed to support TFP. However, issues will arise if someone wants to avail the
service offering another testing methodology. Thus the further research challenge is to design
a generic architecture for a plugin of this kind.

7. Paper Review #3
Paper Title
Developing Fault Tolerant Distributed Systems by Refinement
Author(s)
Elena Troubitsyna
Åbo Akademi University, Dept. of IT
Turku, Finland
Idea
The purpose of this paper is to present a formal approach to developing distributed
fault tolerant systems. For this, it used the action system formalism as design framework and
followed the technique of stepwise refinement.
Result(s)
This paper demonstrates how to specify and refine fault tolerant components and
complex distributed systems composed of them. A general pattern for abstract specification
and refinement of fault tolerant components (the idealized fault tolerant components) is
presented here. Overall, it specifies a fault-tolerant distributed system composed of idealized
fault tolerant components as a parallel composition of action systems specifying them.
Impact
Modern distributed systems are usually complex component based systems. Since
each of the system components, in general, might have several failure modes, ensuring fault
tolerance of distributed systems is a challenging task. The development process proposed in
this paper guarantees correctness of the system’s fault tolerance mechanisms. The proposed
approach also provides a scalable solution for verification of complex distributed systems and
in particularly their fault tolerance, at the same time, it facilitates formal model-driven
development and helps to master complexity of modern distributed systems. In the large
distributed systems ensuring the properties, mentioned in the paper, at the architectural level
brings benefits of early problem discovery. Hence it allows us to arrive at a more robust
design and speed up the development cycle.
Evidence
The demonstration of developing distributed fault tolerant systems includes
mathematical notations. The proposed approach is showed by stepwise formalization and
validation.

8. Prior Work
A similar structure of formal specification of a fault tolerant system was proposed by
Arora and Kulkarni. However, they used model checker to verify their design. Hence their
approach is prone to the state explosion problem and is hard to apply to massively distributed
systems. The approach of this paper is free from this limitation. The study of systems
composed from idealized fault tolerant components has also been undertaken by Brito et al.
However, they aimed at verification of already developed architectures rather than designing
a system according to the principles set by the notion of iFTC (idealized fault-tolerant
component).
Criticism
The research paper contains high level terminologies without specifying exact
meaning of them. The presentation of the paper is also weak and incapable to keep the
attention of the reader. The mathematical notions are presented without proper theoretical
introduction. Though the significance of the idea is high and can be of great effect for the
domain, the overall paper arrangement is not much good. As a result regular reader may face
difficulty to conceive its idea.
Ideas for further work
As a future work it would be interesting to extend the proposed technique to reason
about fault tolerant distributed systems with dynamic reconfiguration.