Hoda K. Mohamed

Green cloud computing: Datacenters power management policies and algorithms

Cloud computing is offering utility oriented IT services to users worldwide. Based on a pay per use model, it provides a variety of computing resources, enterprise applications while enabling their hosting from consumer, scientific and business domains through a three layered architecture and different cloud types. The proliferation of cloud computing has resulted in the establishment of large-scale data centers around the world containing thousands of computing nodes which consume huge amounts of energy, contributing to high operational costs and carbon footprints to the environment. Energy consumption is not only determined by hardware efficiency, but it also depends on the resource management system deployed on the infrastructure and the efficiency of applications running in the system. The challenge is addressed in finding cloud computing solutions that not only save energy for the environment but also reduce operational costs. Our Fuzzy based contribution improves power efficiency with around 40 % than other policies.

Automatic documents classification

Automatic document classification is of paramount importance to knowledge management in the information age. Document classification poses many challenges for learning systems since the feature vector used to represent a document must capture some of the complex semantics of natural language. In this paper, we design an automatic document classification system. We investigate the different parameters and design decisions that affect the building of automatic classifiers. The system creates an item vector for each document retrieved and assigns weights for each item. The vectors are selected using combined techniques from stemmer algorithm and natural language processing. Several weighting schema have been used. Documents are classified using neural network (NN). We investigate different cases applied to the NN classifier. Cases are classified according to weighting schema, effect of weighting words in the title, and the number of inputs to the classifier. Analyzing the performance of the classifier according to different cases is illustrated.

Data Mining for Electrical Load Forecasting In Egyptian Electrical Network

The paper presents the design of a model for forecasting long-term electricity load. The model uses data mining techniques. The paper defines the load forecast and the summary of the most important factors affecting the load forecast in Egyptian electricity network. The steps needed for the knowledge discovery process is implemented to the time series data. Preprocessing the data in order to detect the missing value, odd value, outliers and normalize data. The output from the preprocessing step is fed into multiple regression or neural network to predict the coefficient parameters. Comparison between different cases using different techniques is indicated

Towards Reliable Mobile Cloud Computing

Cloud computing has been one of the fastest growing components of the IT industry. It altered the future of the web by having computing, communication, and storage provides as services to internet users. Mobile Cloud Computing (MCC) is currently gaining steam as an extension to cloud computing as it delivers a large variety of cloud application to billions of smartphones and wearable devices. This paper studies reliability for MCC by determining the ability of a system component to function correctly under different scenarios for a specified period of time. Our aim is to be able to estimate and manage uncertainty and risks of failure. The assessment procedures consist of determine Mean Time between Failures (MTBF), Mean Time to Failure (MTTF), and availability percentages for main components in both cloud computing and MCC structures applied on single node OpenStack installation to analyze its performance with different settings governing the behavior of participants. Additionally, we present here several factors with significant impact on the overall cloud system reliability that should be taken into account in order to deliver highly available cloud computing services for mobile consumers.

Energy efficient resource management for Cloud Computing Environment

Cloud computing is a highly scalable and cost - effective infrastructure for running High Performance Computing, enterprise and Web applications. However, the growing demand of Cloud infrastructure has drastically increased the energy consumption of data centers, which has become a critical issue. Hence, energy efficient solutions are required to minimize this energy consumption. The energy efficient solutions aim at lowering the energy usage of data centers because computing applications and data are growing so quickly that increasingly larger servers and disks are needed to process them fast enough within the required time period so here we reduce the energy consumption by an average of 40% over previously introduced methods. So in datacenters, the number of physical machines can be reduced using virtualization by consolidating virtual machines onto shared servers and enabling them to migrate according to migration policy. This paper presents virtual machines migration and selection policies to boost Cloud Computing Environment energy efficiency and performance.

Adaptive Power Panel of Cloud Computing Controlling Cloud Power Consumption

Cloud computing had created a new era of network design, where end-users can get their required services without having to purchase expensive infrastructure or even to care about troubleshooting. Power consumption is a challenge facing the Cloud Providers to operate their Datacenters. One solution to overcome this is the Virtual Machine (VM) migration, which is a technique used to switch under-utilized hosts to sleep mode in order to save power, and to avoid over-utilized hosts from Service Level Agreement (SLA) violation. But still the problem is that the Cloud Service Provider apply a single policy on all nodes. Our proposed solution is an adaptive power panel where different policies can be applied based on both of the nature of the tasks running on hosts, and the Cloud Provider decision.

Image inpainting based on image segmentation and segment classification

We present a new inpainting algorithm that is based on image segmentation and segment classification. First, we employ the mean shift algorithm to segment the input image. Then, we divide the original inpainting problem to be either one of the two problems: Large Segment Inpainting problem or Non-uniform Segments inpainting problem. The reason we do that is that human eye is more discerning to the errors in the structure and texture propagation of a large-uniform regions with less details while it is less discerning to errors in non-uniform regions with more details. We propose a novel algorithm for each one of the problems- Large Segment Inpainting and Non-uniform Segments inpainting- according to the main features of each one. The experimental results show the advantage of our technique which produces output images with better perceived visual quality.

Automatic cloud's cluster sizing: Controlling cloud service level agreement violation

Cloud computing creates new business opportunity for small, medium or even large size companies, as it eliminates the need of investing in a huge datacenter in order to operate the environment. The contract between the cloud service provider (CSP) and the lessee is called service level agreement (SLA), which guarantee the quality of service (QoS) that should be fulfilled from the CSP to the lessee. The problem is that an over utilized node may result to SLA violation; However, virtual machine (VM) migration from utilized node to unutilized node or a node in power saving mode can decrease the SLA violation percentage, but potentially increase the clouds power consumption. Our proposed solution is to automatically modify the number of nodes in each cluster, to find the optimum cluster size based on the current SLA violation percentage, and to be an extension to our previously proposed adaptive power panel, in which the clouds datacenter is divided into clusters, where different VM migration policies are applied.

Image inpainting using vanishing point analysis and scene segmentation

A new image inpainting technique is developed to fit perfectly for special categories of images that contain mainly buildings. This technique handles the need to obtain an image of building free from parking cars along sides of the roads. To do that, one needs to carefully inpaint the roads and the missing parts of images. This can be done by combining vanishing points detections and image segmentation. After detecting vanishing points, it is possible to draw the line dividing the missing regions into two parts: the road part and the building part. Each part should be then inpainted independently using a different source region for each one. A segmentation technique, which is based on color and texture features, is employed to extract a source region for road part inpainting. Simple geometric calculations are employed to detect the source region for building part inpainting.