Osman Ghazali

A trust computing mechanism for cloud computing with multilevel thresholding

Cloud computing has been considered the new computing paradigm that would make the computing a utility. With the widespread acceptance of cloud computing, computing resources including computational power, storage, development platform and applications will be available as services. Hitherto computing resources were considered capital items and customers used to pay for these resources as for any other product. Cloud computing changes all these and consumers will pay only for what is consumed similar to other services or utilities. When cloud computing becomes more popular, different vendors would provide different services at different levels of qualities and prices. Hence customers will have to identify the differences in quality offered by different service providers and need to select the right service provider based on their requirements. A trust management system will match the service providers and the customers based on the requirements customers and the capabilities of service providers. In this paper, the authors propose a trust management system along with a trust evolution mechanism that can be used to measure the performance of services based on the response time of systems. The proposed mechanism formulates trust scores for different service level requirements based on assurance (confidence) levels. Hence this mechanism is suitable for managing multiple service levels against single trust score for all the service levels. Also the proposed mechanism is adaptive as it continuously modifies the scores based on the performance of the system. The proposed mechanism has been tested in a simulated environment and the results are found to be satisfactory.

Modeling of cloud system using Erlang formulas

Cloud computing has been considered as the 5th utility after electricity, water, gas and telephony. When the cloud computing matures, there will be multiple vendors offering different services at different Quality of Services and at different prices. This would necessitate new tools and mechanisms for analyzing the performance of the system for matching the offerings with requirements. In this paper the authors have modeled the cloud system using queuing theory, specifically Erlang formulas. Four different cloud utility models of various complexities have been presented and analyzed using simulations. The simulations results have also been presented along with an in depth discussion.

A qualitative comparison evaluation of the greedy forwarding strategies in Mobile Ad Hoc Network

The Mobile Ad Hoc Network (MANET) is wireless network which provides communication among wireless mobile hosts without the need of any standing network infrastructure. In such networks, and to facilitate communication between participating nodes, every node has to offer routing services. Routing in MANET is responsible for selecting and forwarding packets along optimal paths. Finding an optimal route is a crucial task in MANET where routes tend to be multi-hop. Many routing protocols have been proposed in literature. However, few of them are efficient when the network is sparse and highly dynamic. Position-based routing and forwarding provides the opportunity for improving the efficiency and performance of the existing MANET routing strategies. This research work presents an extensive overview of geographic forwarding techniques in MANET. In particular, it focuses on the presentation of the basic operation mode of geographic forwarding, which is greedy forwarding. Meanwhile, this research work presents a qualitative evaluation of the most current and popular greedy forwarding strategies used with position based routing protocols. Furthermore, the findings have been used to conclude the most appropriate unicast forwarding policy for future research efforts.

Cloud computing for rural ICT development: Opportunities and challenges

According to the United Nations, almost 50 percent of world population and more than 70 percent of that of developing countries are living in rural areas. Majority of this population live in poverty especially in North and sub Saharan Africa. Rural population face unique problems compared to their urban counterparts in terms of access to technologies, infrastructure, markets and social services. Experts have identified ICT as the tool for overcoming the inefficiencies in traditional methods for the empowerment of rural masses. In this paper, we study how cloud computing can be used to meet the ICT requirements for rural development in terms of opportunities and challenges of implementing and using the new technology.

Evalvid-RASV: Shaped VBR rate adaptation stored video system

Video traffic is a variable bit rates (VBR) data source in nature and it generates highly bursty traffic. Recent implementations mostly buffer the media source in order to regenerate it in the form of constant bit rates (CBR). Consequently, it will add more delays to the system and thus unable to support the original nature of the video data. Inspired by the works of Hamdi et al. [1] and Lie and Klaue [2], we developed Evalvid-RASV. This system is working on the VBR concept (open-loop video coding), but it is “shaped” so that it will not produce uncompromised bursty traffic without additional delay. With the knowledge of video characteristics in advance, Evalvid-RASV was developed to utilize the information resulting a better algorithm. In addition, we implemented the system in Evalvid-RA environment. It is an environment which is able to perform rate adaptation to the media data source and has an integrated video performance evaluation tools, especially user-perceived video quality. Our experiments have shown that Evalvid-RASV outperforms open-loop VBR in term Peak Signal Noise Ratio (PSNR) value and acceptable delay time.

Friendliness of DCCP towards TCP over large delay link networks

The need for TCP friendly transport protocol is vital for the future Internet. With this motivation, IETF has defined a new transport protocol, DCCP, to be a friendly transport protocol when coexist with other congestion-controlled transport protocol such as TCP. DCCP is an unreliable transport protocol which is designed to be able to carry unreliable time sensitive data traffic such as audio or video streaming with built-in congestion control. Therefore, DCCP can carry streaming data same as UDP but this new transport protocol can adjust the transmission data rate according to the condition of the network. In case of congestion, the sending data rate will be slowed down for the network to function well so that other transport protocol data can have the chance to utilize the network bandwidth simultaneously. In this paper, we investigate the friendliness of DCCP towards another transport protocol TCP over large delay link networks when carrying constant bit rate data. The result showed that DCCP is TCP friendly and it can coexist with TCP harmonically over large delay link networks.

Enhancing DCCP congestion control mechanism for long delay link

Most of the multimedia applications use the User Datagram Protocol (UDP) as a transport layer protocol because it is suitable for the delivery of multimedia data over the Internet. However, the use of UDP could endanger the stability of the network because there is no congestion control applied. To a certain extent, the network can collapse if too many applications deliberately use this protocol. Subsequently, instead of using the UDP, the applications have choices to use the Datagram Congestion Control Protocol (DCCP), which has a built-in congestion control that can provide a better network. Nevertheless, the congestion control mechanism in the Congestion Control Identifier (CCID)-2 TCP-like can cause problems when delivering multimedia data over a long delay link. To alleviate the problems, such as longer time taken for achieving maximum throughput, and throughput fluctuation during a congestion avoidance phase, two approaches have been used, i.e. setting of an appropriate slow-start threshold value and adjusting congestion window during a congestion avoidance phase. “TCP-like Threshold Window (TCP-like TW)” has been developed and modeled in the Network Simulator 2 (ns-2). For a long delay link, the TCP-like TW congestion control mechanism is able to minimize the time taken to achieve the maximum throughput. It also can smooth the fluctuation of throughput after achieving the maximum throughput.

Pause Time Optimal Setting for AODV Protocol on RPGM Mobility Model in MANETs

For the last few years, a number of routing protocols have been proposed and implemented for wireless mobile Ad hoc network. The motivation behind this paper is to discover and study the pause time effects on Ad hoc on Demand Distance Vector (AODV) protocol and find out the node pause time optimal setting for this protocol where Reference Point Group Mobility (RPGM) model uses as a reference model. In order to come across the best performance of a particular routing protocol, there a need to examine a number of parameters with different performance and analyze the optimal setting of that protocol and its network configuration environment. This experiment, the speed is fixed with 20 ms in all scenarios while the pause time is varying from scenario to another to observe the optimal setting of the pause time on protocol’s performance in this configuration. The outcome of the experiment are analyzed with different parameters such as varying number of nodes, increasing connections, increasing pause time and discussed the effects of the pause time. The results have shown that the value of the pause time can be affecting the performance of the protocol. In the experiment, we found that the lower pause time give better performance of the protocol. However, this paper is a part of ongoing research on AODV protocol in link failure. Thus, it is important to figure out the factors which can be involved the performance of the protocol.

Impact of Weight Values in Hybrid and Adaptive FEC Mechanism over Wireless Network

Forward Error Correction (FEC) techniques have been adopted with Automatic Repeat reQuest (ARQ) to overcome packet losses and avoid network congestion in various wireless network conditions. Two factors have been used which are queue length and packet retransmission times. According to these two factors, two different weight values are used to generate consistence and smooth value of queue length and packet retransmission times. Firstly, qweight is calculated with current queue length to generate average queue length. Secondly, sweight is calculated with current packet retransmission times to determine average packet retransmission times. Thus, this paper presents the experiments to investigate the reasonable qweight and sweight values, in order to determine the appropriate average number of these two parameters. The results show when the value of qweight and sweight is 0.1, high number of FEC packets and high value of Peak Signal Noise Ratio (PSNR) is generated. When the experiment is run with different values of qweight and static sweight, the result shows that smaller value of qweight is better in order to produce best quality of video.

A Memoryless Trust Computing Mechanism for Cloud Computing

Trust management systems play an important role in identifying the quality of service in distributed systems. In cloud computing, trust systems can be used to identify service providers who would meet the requirements of customers. Several trust computing mechanisms have been proposed in literature based on various trust metrics. Most of these systems compute the trust scores incrementally from the previous values. This is major vulnerability that can be exploited by adversaries to attack the system forcing the trust scores towards extreme values. In this paper, the authors present a memoryless trust computing mechanism which is immune to such attacks. The proposed mechanism does not depend on the previous trust scores hence it cannot be forced towards extreme values by repeated requests. The simulations experiments conducted show that the trust scores computed using the proposed mechanism are more representative and stable in the face of attacks compared to other systems.