Haleh Khojasteh

Towards an efficient rendezvous protocol for a cognitive PAN

In this paper we describe a rendezvous protocol that enables nodes to quickly and reliably discover an existing channel-hopping cognitive PAN. The proposed protocol does not rely on the presence of a dedicated controller, the availability of a common control channel, or clock synchronization between the cognitive PAN and the new node. Furthermore, the CPAN need not suspend its operation for the discovery procedure to succeed. The performance of the proposed protocol is evaluated under fixed and adaptive superframe duration.

Characterizing energy consumption of IaaS clouds in non-saturated operation

To save energy, physical machines (servers) in cloud data centers are partitioned in different pools, depending on whether they are kept on at all times and/or whether they have virtual machines instantiated. Partitioning (pooling) of servers affects the power consumption of the data center but also the performance and responsiveness to user requests. In this paper we examine the behavior of pool management scheme in different operating regions which correspond to linear operation, transition to saturation, and saturation, in particular the tradeoff between performance and power consumption. In addition, our results show that the offered load (analogous to the concept often used in networking) does not offer complete characterization of data center operation; instead, the impact of task arrival rate and task service time must be considered separately.

CSCD: A simple channel scan protocol to discover and join a cognitive PAN

The paper describes a simple channel scan protocol through which a station can quickly discover and join an existing piconet following a specified channel-hopping MAC protocol piconet in its vicinity. The proposed MAC protocol is simple and does not need a dedicated control channel. Moreover, the cognitive piconet can operate normally throughout the search, unlike Bluetooth where piconet operation is temporarily suspended during the inquiry/inquiry scan procedure. The performance of the proposed protocol is evaluated through discrete event simulation.

Prioritization of Overflow Tasks to Improve Performance of Mobile Cloud

Mobile devices may offload their applications to a virtual machine running on a cloud host. This application may fork new tasks which require virtual machines of their own on the same physical machine. Achieving satisfactory performance level in such a scenario requires flexible resource allocation mechanisms in the cloud data center. In this paper we present two such mechanisms which use prioritization: one in which forked tasks are given full priority over newly arrived tasks, and another in which a threshold is established to control the priority so that full priority is given to the forked tasks if their number exceeds a predefined threshold. We analyze the performance of both mechanisms using a Markovian multiserver queueing system with two priority levels to model the resource allocation process, and a multi-dimensional Markov system based on a Birth-Death queueing system with finite population, to model virtual machine provisioning. Our performance results indicate that the threshold-based priority scheme not only performs better, but can also be tuned to achieve the desired performance level.

Integration of an IEEE 802.15.4 RFID network with mobile readers with a 802.11 WLAN

In this paper, we propose a coexistence scheme that allows radio-frequency identification readers or sinks to collect data from radio-frequency identification tags or sensors that use IEEE 802.15.4 wireless personal area network standard and send it to the access point via an IEEE 802.11 wireless local area network. The proposed scheme uses time scheduling and bridging via readers to allow for simultaneous operation of heterogeneous wireless networks that operate in the industrial, scientific, and medical frequency band at 2.4GHz. A simple sleep management approach is developed that allows tags to reduce their power consumption as well as collision rate. To provide timely reaction to sudden changes in tag population or reader availability, we devise a scheme where the scheduling parameters are adjustable and readers are mobile. We evaluate the performance of our solution in four scenarios and confirm the flexibility of the proposed approach. Copyright

A two-tier integrated RFID/sensor network with a WiFi WLAN

We investigate the coexistence of Wi-Fi (WLAN) and RFID/sensor networks that use IEEE 802.15.4 standard in the Industrial, Scientific and Medical (ISM) frequency band. To minimize the interference between the networks, a time scheduling scheme is combined with a sleeping scheme for RFID tags/sensors to further reduce the collision rate and energy consumption. We evaluate the performance of our solution under two different scenarios using discrete event simulation.

Task admission control policy in cloud server pools based on task arrival dynamics

In this paper, we propose two task admission control algorithms that utilize random task filtering: a lightweight algorithm based on long-term estimates of average utilization and offered load, and a more complex algorithm based on instantaneous utilization. Detailed performance evaluation confirms that both algorithms are able to ensure that the system remains in the stable operating region. We have also found that more aggressive filtering tends to decrease task blocking rate and delay. Copyright

Task filtering as a task admission control policy in cloud server pools

In this paper we propose a task admission control scheme in a cloud system. We develop two task admission control algorithms which are based on a random task filtering policy. First, we present a lightweight task admission control algorithm which is appropriate for the cloud systems with smooth changes of task arrival rate. This algorithm is based on long-term estimation of average utilization and offered load. Second, we introduce a task admission algorithm which is efficient for highly dynamic systems and is based on instantaneous utilization. The performance of the proposed schemes is evaluated against varying intensities of offered load. Performance evaluation of the proposed schemes confirms that both of them are able to ensure that the system is kept in the stable operating region.

Simple solutions may still be best: on the selection of working channels in a channel-hopping cognitive network

Spectrum decision-that is, the selection of a channel for the next hop-is one of the most important factors that affect the performance of a channel-hopping cognitive network. In this paper, we compare the performance of a number of channel selection algorithms through the probability of collisions with primary user transmissions. The results indicate that a simple histogram-based selection algorithm performs the best, except in the somewhat unrealistic scenario where primary users are homogeneous with respect to mean period and mean duty cycle of their channel activity, in which case selecting the next-hop channel among those that have most recently turned idle offers the best performance. Furthermore, histogram-based selection is shown to be quite resilient to errors inherent to channel sensing and is thus a primary candidate for a wide range of applications of channel-hopping cognitive networks. Copyright

Analyzing the impact of provisioning overhead time in cloud computing centers

In this paper, we analyze an efficient pool management model for cloud systems that partitions the servers (physical machines) into a hot and cold pool to improve energy efficiency. Servers are moved them from one pool to the other as needed to fulfill the incoming task requests, which are provisioned on virtual machines running on the servers. The model features two levels of task admission control: one at the global input, the other at each server separately. We examine the behavior of the cloud system in the regions of linear operation, transition to saturation and saturation, from the viewpoint of task rejection rates and energy consumption. Furthermore, we evaluate the sensitivity of task blocking probability and energy consumption to the pool partitioning threshold and the value of mean look up time in hot and cold pools, respectively, in different test scenarios.