Asif Imran

An empirical investigation of cost-resource optimization for running real-life applications in open source cloud

Open source cloud technology can optimally increase resource utilization and reduce costs in many organizations to support and execute their applications. The extent to which cloud can reduce memory wastage and save costs is an important issue of research. This paper aims to address this issue through the evaluation of an organizations cost-resource benefits by executing real life applications on open source private cloud platform and evaluating the empirical data. The variability in performance of the cloud and traditional systems is shown by defining metrics and provisioning inputs to obtain results. Through the experimentation, memory utilization is seen to increase by 22.16% in the cloud. Direct-indirect cost savings sum up to 40%. The cloud features playing key role to this improved cost-resource performance are revealed.

A peer to peer resource provisioning scheme for cloud computing environment using multi attribute utility theory

Resource provisioning is critical for cloud computing because it manages the virtual machines (VM) and allocated resources. Traditional resource provisioning schemes make decisions based on centralized configuration and global calculation of resource allocation. However these provisioning frameworks become hindered in large deployment followed by Service Level Objectives (SLO) violation and those are exposed to single point failure as well. The proposed provisioning scheme is based on Peer to Peer (P2P) architecture with no lone decision maker. Each node in the data center makes its own provisioning decision regarding VM allocation and migration which are resolved with Multi Attribute Utility Theory (MAUT) methods. Simulation experiments demonstrate that in comparison with centralized schemes, the proposed scheme generates 60.27% less SLO violations and 83.58% less VM migrations on average. Such outcome proves that the system can manage its resources better compared to centralized scheme.

IVRIDIO: Design of a software testing framework to provide Test-first Performance as a service

Test-first Performance (TFP) is a testing paradigm that focuses on performance testing from the early stage of development. For performance oriented applications like a web service, TFP approach can reduce the overall cost of software testing. Given this potential benefit, TFP is yet to be incorporated in existing cloud testing frameworks. This paper proposes the design of a testing framework which introduces TFP as a Service (TFPaaS) named as IVRIDIO. It includes the Plugin for TFP in the Cloud (PTFPC) that will provide instant feedbacks to fix critical performance issues. To simplify the TFPaaS availing procedure, the Convention over Configuration (CoC) design paradigm has been applied. A configurable project template is designed using the CoC to maintain TFP test cases. Furthermore, necessary directions are given to prototype the framework as a testbed for related research.

Combinatorial Color Space Models for Skin Detection in Sub-continental Human Images

Among different color models HSV, HLS, YIQ, YCbCr, YUV, etc. have been most popular for skin detection. Most of the research done in the field of skin detection has been trained and tested on human images of African, Mongolian and Anglo-Saxon ethnic origins, skin colors of Indian sub-continentals have not been focused separately. Combinatorial algorithms, without affecting asymptotic complexity can be developed using the skin detection concepts of these color models for boosting detection performance. In this paper a comparative study of different combinatorial skin detection algorithms have been made. For training and testing 200 images (skin and non skin) containing pictures of sub-continental male and females have been used to measure the performance of the combinatorial approaches, and considerable development in success rate with True Positive of 99.5% and True Negative of 93.3% have been observed.

ProvIntSec: a provenance cognition blueprint ensuring integrity and security for real life open source cloud

The distributed nature and growing demand for open source cloud makes the system an ideal target for malicious attacks and unauthorised file transfers. Requirements of provenance cognition scheme can come forward to solve the problem. However, such mechanisms of provenance detection has been considered to a limited extent for open source cloud computing. ProvIntSec is a novel mechanism that ensures effective collection of provenance information from a large pool of virtual machine (VM) instances on open source cloud platform. ProvIntSec captures critical system journals from VM instances and pattern matches those with predefined signatures to detect the presence of malicious activities. In addition, ProvIntSec identifies the Linux process trees to determine unauthorised file movements across different nodes. The experiments were executed in OpenStack Essex cloud environment running on real life system, and standard metrics were used to calculate the results. The obtained results show average precision values of 92.81% and 81.24% for malware detection and unauthorised file transfers respectively. At the same time, cumulative performance gains of 0.3991 and 8.77 are obtained. Upon comparison of the obtained results with benchmarks, ProvIntSec shows desirable gain in performance.

Time-Waved Monitoring and Emergent Self Adaption of Software Components in Open Source Cloud

Optimized resource utilization and low cost of service has enabled the cloud to become a popular service in todays world. However, rapid scaling, continuous attacks from hackers, dynamic resource provisioning and distributed nature has made it a complex system to manually monitor and manage by system administrators. This paper proposes an effective time-waved framework for monitoring the cloud and reporting undesirable activities with minimum time delay. Next, it presents a mechanism to self-adapt the attacked modules through allocation of healthy ancillary resources. Performance analysis of the proposed framework yields desirable time complexities of 17.0, 26.6, 27.3 and 18.6 seconds for 4 types of attacks tested here. Also, replacing paralyzed cloud virtual machines (vm) with healthy ones requires 8.4 seconds on average, resulting in desirable performance. The experimentation on open source platform show that the proposed schemes enable better monitoring of cloud services.

Watchword-oriented and time-stamped algorithms for tamper-proof cloud provenance cognition

Provenance is derivative journal information about the origin and activities of system data and processes. For a highly dynamic system like the cloud, provenance can be accurately detected and securely used in cloud digital forensic investigation activities. This paper proposes watchword oriented provenance cognition algorithm for the cloud environment. Additionally time-stamp based buffer verifying algorithm is proposed for securing the access to the detected cloud provenance. Performance analysis of the novel algorithms proposed here yields a desirable detection rate of 89.33% and miss rate of 8.66%. The securing algorithm successfully rejects 64% of malicious requests, yielding a cumulative frequency of 21.43 for MR.

A case-based framework for self-healing paralysed components in Distributed Software applications

Self-healing is the ability of the software to detect faulty modules at execution time and replace or recover those without affecting other components. This paper proposes a framework for self-healing of Distributed Software System (DSS). Monitoring component is used to detect and record failures of DSS. Healing system will replace the paralysed components with healthy ones which will be initiated from the information given by Monitoring system to the proposed Reviver process. A failed case table is required to match the real life failures with it for identification of the solution. Distance between the failed case table and the recorded failures need to be calculated using exclusive OR since the solution closest to the fail can then be determined. Afterwards, the minimum distance between those is used to resolve the failure. This recovery is achieved through replacement of the faulty modules with redundant components in the DSS. Performance evaluation shows a desirable time consumption of less than the standard 0.7 seconds for component replacement in all the experimental iterations.