Goldie Gabrani

SLA Based Scheduler for Cloud for Storage & Computational Services

The proposed work gives novel technique to exploit the cloud for the purpose of qualitative improvement in services being rendered. In this context a service level agreement is executed between the service provider and the user incorporating stipulations regarding terms & conditions relevant with the transactions. Agreement envisages the provisioning of services to the users at the appointed time and charges commensurate with the market scenario. Besides the interests of the service providers are equally watched upon by allowing negotiations and settlements at par with prevailing market trends, thus ensuring benefits to both the service providers and the users. The scheduler algorithm allows re-provisioning of resources on the cloud in the event of failures. The focus of the model is to provide fair deal to the users and consumers, enhanced quality of service as well as generation of optimal revenue.

Negotiation based advance reservation priority grid scheduler with a penal clause for execution failures

The utility computing in a grid demands much more adaptability and dynamism when certain levels of commitments are to be complied with. Users with distinct priorities are categorized on the basis of the types of organizations or applications they belong to and can submit multiple jobs with varying specific needs. The interests of consumers and the resource service providers must be equally watched upon, with focus on commercials too. At the same time, the quality of the service must be ascertained to a committed level, thus enforcing an agreement. The authors propose an algorithm for a negotiation based scheduler that dynamically analyses and assesses the incoming jobs in terms of priorities and requirements, reserves them to resources after negotiations and match-making between the resource providers and the users. The jobs thus reserved are allocated resources for future. The performance evaluation of the scheduler for various parameters was done through simulations. The results were found to be optimal after incorporating advance reservation with dynamic priority control over job selection, involving the impact of situations confronting resource failures introducing economic policies and penalties.

An Economy Based Storage Grid Federation with a Decentralized Scheduling Technique

With the evolution of Grid technology, it has become possible to connect resources (service providers or consumers) topologically spread apart falling in different administrative domains. The users or applications submitting multiple storage requests are managed through scheduling arrangements in a centralized manner. The distribution and allocation of the storage requests therefore occurs oblivious of the efficiency and engagements of the underlying resources. The authors propose a model of Storage Grid Federation motivated by concept of Grid Federation constituted of computational resources formulated through peer level coupling. The algorithm is designed for the resource management with a decentralized approach incorporating policies capable of resource monitoring through entities like Storage Federation Agents so as to yield to higher autonomy. The preferences of the consumers are well accessed such as budget, negotiation policies, a specific need of storage and protocol of the organization they belong to. The design includes penal clause and discounts in accordance to a service level agreement while offering Storage Grid services at specified time in future. The goal is achieved through creating number of Storage Federation Agents that are competent to cater to local Resource Service Provider policies. The SFA resource managers are competent to ensure load balancing before distribution of storage requests. The proposed algorithm allows negotiation with SFAs in addition to the local and remote SFAs falling in different domains and hence improving upon its previous federation model based on computational resources only.

Machine Learning Based Predictive Model for Risk Assessment of Employee Attrition

Every organization today is challenged with the issues of employee attrition. Attrition is the reduction in the employee base of an organization. This could be because of voluntary resignation or expulsion by the higher management. It becomes important for the company to be prepared for the loss of human power in whom company has invested and from whose help it has earned revenue. Thus, it is a profitable idea to predict the risk involved with uneven attritions so that management can take preventive measures and wise decisions for the benefit of the organization. In this paper, a model based on Machine Learning techniques that predicts the employee attrition has been designed. The model is implemented and is thoroughly analyzed for the full profile of companies. It has been shown that the model can be effectively used to maximize the employee retention.

Cloud providing on demand storage services enforcing techno commercial service level agreement for an improved QoS

Advancements in the application service provisioning have created stringent scenarios in the world of cloud. The framework now needs to comply with specific requirements of consumers with specific demands. The diversified needs of storage and other related services in the dynamically changing needs of consumers. Such application requirements are becoming frequent as a consequence to high performance computing and large data/information retrievals. At the same time, understandings between the Cloud and its consumers have to be more transparent and should work under pre-defined terms. This leads to a committed Quality of Service delivered which directly effects revenues. To bring in transparency and durability between the service providers and service seekers, there is a need of a Techno Commercial Service Level Agreement that fixes responsibility of both parties. In this work, authors propose and implement an algorithm for meta scheduling that envisages enforcement of SLA apart from managing the distinct cloud applications such as storage. The work assesses the changes brought in terms of revenues, QoS enhancement, recoveries, building long term ties etc. The results have been found to be encouraging while testing the scheduler proposed with stringent service level conditions well. The agreement imposes clear by laws and penalties to the bi-part irate redressed on service provider cloud & the consumers. The frame work is tested and achieves objectives & goals with specific features suitable for storage cloud.

Effort estimation models using evolutionary learning algorithms for software development

Software effort estimation is a complicated task being carried out by software developers as very little information is available to them in the early phases of software development. The information collected about various attributes of software needs to be subjective, which otherwise can lead to uncertainty. Inaccurate software effort estimation can be disastrous as both underestimation and overestimation may result in schedule overruns and incorrect estimation of budget. This paper focuses on the comparative study of various non-algorithmic techniques used for estimating the software effort by empirical evaluation of five different evolutionary learning algorithms. The accuracy of these algorithms is found out and the behavior of these algorithms is analyzed with respect to the size and the type of data. All the five techniques are applied on three different datasets and various paramenters such as MMRE, PRED(25), PRED(50), PRED(75) are calculated. The proposed results are compared to other machine learning methods like SVR, ANFIS etc. The results show that evolutionary learning algorithms give more accurate results than machine learning algorithms.

Handwritten statement analysis using neural networks

This paper aims to scan a handwritten statement sample in the form of an image then uses the neural network to identify the graphological and statement analysis marker s that can give an insight into the psyche of the person who wrote the sample. It is a tried and tested fact that at least 70% of the communication that we engage in is nonverbal in nature. It is from this fact that the sciences of Micro-expression reading, body language analysis etc originated. A similar pattern is observed in the the linguistics of the choice of words of a person and his handwriting. Graphology (Handwriting Analysis) and Statement Analysis can also be a tool to gain an insight in to the psyche of the person. For example if a person uses the word never as a substitute for no in a yes or no there is a high probability of a suspicion of deception in the context. Similar techniques are used by experts of Neuro Linguistic programming worldwide for a variety of applications. They are used to interrogate criminals and hasten the process of investigation. They can act as human lie detectors. From a written statement they can highlight the possible statement that need cross-verification as they can spot graphological and statement analysis markers in them. But such people are few in number and it takes years and years of training to be at that skill level. Hence a software (Matlab application) is then programmed to replicate this behaviour.

Artificial Intelligence Based Recommender Systems: A Survey

In recent years, Artificial Intelligence (AI) techniques like (a) fuzzy sets, (b) Artificial Neural Networks (ANNs), (c) Artificial Immune Systems (AIS) (d) Swarm Intelligence (SI), and (e) Evolutionary Computing (EC) are used to improve recommendation accuracy as well as mitigate the current challenges like Scalability, Sparsity, Cold-start etc. Aim of the survey is to incorporate the recommender system in light of the AI techniques. Various AI techniques are presented and recommender system’s challenges are also presented. Moreover, we have tried to study the ability of AI techniques to deal with the above mentioned challenges while designing recommender systems. Furthermore, pros and cons of AI techniques are discussed in detail.

A dynamic two-level priority based authentication system for job scheduling in a heterogeneous grid environment

The Grid environment is highly dynamic in nature where a variety of users from all over the world try to access the resources that are again distributed all over. Both the users and resources are highly varied. The users can be both valid and invalid, can have single job or multiple jobs, and can have jobs that can have different processing requirements, may be an old loyal user or may be a new one. Similarly the resources can have different characteristics like the processing speed, number of processing elements each resource has, whether the resources have joined or left the grid and so on. In such an environment if an un-authorized user is allowed to access the resource, there is a security threat to the grid. Even if one unauthorized user is allowed to access the grid; the whole grid becomes susceptible to many security threats. So resources should only be available to authorized users. Moreover, as number of users and their jobs increases, they face a lot of competition amongst themselves to get access to the resources. In order to solve the conflicts the resources must be allocated to the users by some arbitration mechanism. In this paper we propose a grid framework that focusses on solving these two major problems namely authentication and arbitration by means of using a robust authentication mechanism and by assigning priority to users and their jobs. In order to achieve secure resource allocation to valid user jobs for maximum resource utilization along with minimizing waiting time and elapsed time of jobs, a secure resource allocation scheme using ECC algorithm with two-level priority has been proposed. The results have been compared with a non-priority based system and our proposed system shows substantial improvement in both waiting time and elapsed time of jobs with better resource utilization.

A dynamic priority metascheduler for an SLA-based storage grid

Technological advancements in various fields of engineering generate large quantum of data, distributed geographically. Managing this data requires huge storage resources that can either be dedicated or can be shared. The dedicated storage being expensive and therefore, having shared storage is economical to provide the resources, as and when the demand is generated. In this paper, the authors propose a model of a special type of grid called storage grid that offers on demand storage services giving different types of storages to cater to different requirements of the users. It first prioritizes the requests of various users and then reserves the storage space and the type of storage for them in advance where an agreement is executed between the storage grid and the user commonly known as Service Level Agreement (SLA). The SLA postulates the terms and conditions governing provisioning of services on demand, pricing of services arising out of negotiations in the background of market value and a penal clause for levy of penalty that can be defrayed by allowing a discount in the price. The proposed storage grid consists of a metascheduler that manages the storage resources in terms of availability and allocation, taking care of the pricing and priority of the users hitherto embodied in the SLA. The performance evaluations were conducted through simulations for the storage requests acceptance and failures subject to reservation criterion.