David S. L. Wei

Task clustering and scheduling for distributed memory parallel architectures

This paper addresses the problem of scheduling parallel programs represented as directed acyclic task graphs for execution on distributed memory parallel architectures. Because of the high communication overhead in existing parallel machines, a crucial step in scheduling is task clustering, the process of coalescing fine grain tasks into single coarser ones so that the overall execution time is minimized. The task clustering problem is NP-hard, even when the number of processors is unbounded and task duplication is allowed. A simple greedy algorithm is presented for this problem which, for a task graph with arbitrary granularity, produces a schedule whose makespan is at most twice optimal. Indeed, the quality of the schedule improves as the granularity of the task graph becomes larger. For example, if the granularity is at least 1/2, the makespan of the schedule is at most 5/3 times optimal. For a task graph with n tasks and e inter-task communication constraints, the algorithm runs in O(n(n lg n+e)) time, which is n times faster than the currently best known algorithm for this problem. Similar algorithms are developed that produce: (1) optimal schedules for coarse grain graphs; (2) 2-optimal schedules for trees with no task duplication; and (3) optimal schedules for coarse grain trees with no task duplication.

Software implementation strategies for power-conscious systems

A variety of systems with possibly embedded computing power, such as small portable robots, hand-held computers, and automated vehicles, have power supply constraints. Their batteries generally last only for a few hours before being replaced or recharged. It is important that all design efforts are made to conserve power in those systems. Energy consumption in a system can be reduced using a number of techniques, such as low-power electronics, architecture-level power reduction, compiler techniques, to name just a few. However, energy conservation at the application software-level has not yet been explored. In this paper, we show the impact of various software implementation techniques on energy saving. Based on the observation that different instructions of a processor cost different amount of energy, we propose three energy saving strategies, namely (i) assigning live variables to registers, (ii) avoiding repetitive address computations, and (iii) minimizing memory accesses. We also study how a variety of algorithm design and implementation techniques affect energy consumption. In particular, we focus on the following aspects: (i) recursive versus iterative (with stacks and without stacks), (ii) different representations of the same algorithm, (iii) different algorithms – with identical asymptotic complexity – for the same problem, and (iv) different input representations. We demonstrate the energy saving capabilities of these approaches by studying a variety of applications related to power-conscious systems, such as sorting, pattern matching, matrix operations, depth-first search, and dynamic programming. From our experimental results, we conclude that by suitably choosing an algorithm for a problem and applying the energy saving techniques, energy savings in excess of 60% can be achieved.

An efficient anonymous communication protocol for peer-to-peer applications over mobile ad-hoc networks

An efficient anonymous communication protocol, called MANET Anonymous Peer-to-peer Communication Protocol (MAPCP), for P2P applications over mobile ad-hoc networks (MANETs) is proposed in this work. MAPCP employs broadcasts with probabilistic-based flooding control to establish multiple anonymous paths between communication peers. It requires no hop-by-hop encrypt ion/decryption along anonymous paths and, hence, demands lower computational complexity and power consumption than those MANET anonymous routing protocols. Since MAPCP builds multiple paths to multiple peers within a single query phase without using an extra route discovery process, it is more efficient in P2P applications. Through analysis and extensive simulations, we demonstrate that MAPCP always maintains a higher degree of anonymity than a MANET anonymous single-path routing protocol in a hostile environment. Simulation results also show that MAPCP is resilient to passive attacks.

A Cognitive MAC Protocol for QoS Provisioning in Overlaying Ad Hoc Networks

Cognitive Radio (CR) can effectively reuse the same frequency of the existing legacy systems with the help of the adaptivity provided by the software defined radio technique and the intelligence learned by sensing the huge spectrum in the surrounding environment. One fundamental issue for a CR network is how CR users establish an overlaying ad hoc link on licensed and unlicensed bands. On licensed band, the CR user has to detect the presence of the primary user and vacate accordingly to avoid the interference. On the unlicensed band, the medium access shall support the quality of service (QoS) as well as improving the efficiency and fairness for the spectrum usage. In this paper, by moderately reshaping the legacy carrier sense multiple access (CSMA) medium access control (MAC) protocol, we propose a cognitive and distributive MAC protocol to establish a CR ad hoc network with QoS provisioning, high efficiency and fairness. Through the simulations by NS-2, the proposed cognitive MAC protocol can improve throughput by 50% compared to the legacy carrier sense multiple access with collision avoidance (CSMA/CA) MAC protocol, while keeping the dropping rate less than 2% for delay-sensitive applications.

Analysis of the Bluetooth device discovery protocol

Device discovery and connection establishment are fundamental to communication between two Bluetooth (BT) devices. In this paper, we give an analytical model of the time it takes for the master in a piconet to discover one slave. We show that, even in the absence of packet interference, the discovery time can be long in some instances. We have simulated the discovery protocol by actually implementing it to validate the analytical model. By means of simulations, we show how discovery time is affected by (i) the presence of multiple potential slaves, and (ii) changes in the maximum backoff limit. Using simulation studies we observed the effectiveness of two proposed improvements to device discovery, namely, (i) avoiding repetitions of the A and B trains before a train switch, and (ii) eliminating the idea of random backoff, or reducing the backoff limit. We show that discovery time can be reduced by avoiding repetitions of the A and B trains before a train switch. However, complete elimination of the random backoff is not a good idea, as discovery time will be too long when the number of BT devices is large. Instead, choosing a small backoff limit of 250–300 slots is highly effective in reducing discovery time even in the presence of a large number (say, 50) of potential slaves.

Latency Analysis for Dynamic Spectrum Access in Cognitive Radio: Dedicated or Embedded Control Channel?

Dynamic spectrum access (DSA) is the key feature of cognitive radio (CR) networks, but it also poses many new challenges on the medium access control (MAC) design. One of key challenges is the fact that the secondary CR users can only borrow the licensed spectrum from the primary users for a short period of time. Hence, unlike many available multi-channel MAC protocols for ad hoc networks where throughput is the main performance issue, the DSA protocols in CR networks shall place more emphases on the access latency. Hence, one fundamental issue arises: how can the spectrum be dimensioned for control channels in order to minimize the access delay of DSA protocol in CR networks? In this paper, we provide a comparative study in an analytical manner on the latency performance of two DSA protocols: 1) dedicated control channel, and 2) embedded dedicated control channel approaches. Our results show that an optimal ratio of the control channel bandwidth over the total channel bandwidth can be found to minimize the latency of DSA with dedicated control channels. However, the delay performance of DSA with dedicated control channels is more sensitive to the variations of the data lengths than that of DSA with embedded control channels. Hence, we conclude that the way of dimensioning the spectrum for control frames for DSA in CR networks should be adaptive to the variations of the traffic characteristics and the number of users.

A topology control algorithm for constructing power efficient wireless ad hoc networks

In this paper, we present a localized algorithm for constructing power efficient topology for wireless ad hoc networks. Each mobile node determines its own transmission power based only on local information. The proposed algorithm first constructs the constrained Gabriel graph from the given unit disk graph and then reduces the total transmission power by allowing each node individually excises some replaceable links. The constructed topology is sparse, has a constant bounded power stretch factor, and the total transmission power is lower than those obtained from other proposed algorithms. In addition, compared with others, our algorithm requires lower time complexity to generate a solution, and can thus further save the energy for each mobile node. We demonstrate the performance improvements of our algorithm through simulations.

Analysis of packet interference and aggregated throughput in a cluster of Bluetooth piconets under different traffic conditions

In a Bluetooth piconet, the Master essentially controls the channel. Due to an absence of coordination between independent Masters while accessing the wireless medium, devices will encounter high packet interference if several piconets are simultaneously operating in the same area. Since even a headset and a mobile phone can be connected with a Bluetooth link forming a piconet, it may not be unusual to find tens of independent piconets in crowded places like airports, international conferences, shopping malls, and so on. Study of packet interference is important because interference affects the throughput of a piconet. Motivated by the fact that applications will benefit, in terms of higher available data rate in one direction, by using multiple-slot packets in an asymmetric manner, in this paper, we present an analytical model of packet interference in a cluster of piconets using multiple-slot packets. Also, considering that all the portable devices can have a Bluetooth interface and people are highly mobile these days, it will not be uncommon to find a cluster of piconets of both the 79-hop and the 23-hop types in the same area. We then present an analytical model of interference of multiple-slot packets in a heterogeneous cluster of Bluetooth piconets. By a heterogeneous cluster we mean some piconets are of the 23-hop type and the rest are of 79-hop type. We show how the aggregate throughput in a cluster of piconets degrade under various traffic scenarios, such as 1-slot, 3-slot, and 5-slot packets in symmetric and asymmetric modes in synchronous and asynchronous conditions of Master clocks. Our analytic model is based on the idea of probabilistic graphs, where a node denotes a piconet and an edge denotes the probability of interference between two nodes. Though the 23-hop system has been phased out, our work gives a general approach to model packet interference in multiple, frequency-hopping systems that need not be Bluetooth systems.

Coexistence Mechanism Using Dynamic Fragmentation for Interference Mitigation between Wi-Fi Bluetooth

In this paper, we present a non-collaborative coexistence mechanism using dynamic fragmentation that functions very well at the Wi-Fi in the presence of Bluetooth interference. The proposed scheme tries to optimize MAC layer packet length such that the Wi-Fi device has better chance to avoid the interference caused by Bluetooth devices. The mechanism dynamically adjusts the fragmentation level based on the current packet error rate (PER). We developed an analytical model that provides the MAC the necessary information (in terms of PER) to decide the right time for further packet fragmentation. The developed model is also employed to measure the throughput of the Wi-Fi device. Simulations are also performed to validate the developed model. We show that our coexistence mechanism could significantly improve the performance of Wi-Fi in both throughput and transmission delay, though there is only slight performance improvement at the Bluetooth side

Packet interference in a heterogeneous cluster of Bluetooth piconets

In a Bluetooth piconet, the master essentially controls the channel. Due to an absence of coordination between independent masters while accessing the wireless medium, devices will encounter high packet interference if several piconets are simultaneously operating in the same area. Since even a headset and a mobile phone can be connected with a Bluetooth link, forming a piconet, it may not be unusual to find tens of independent piconets in crowded places like airports, international conferences, shopping malls, and so on. The study of packet interference gains importance, because it affects throughput of a piconet. Considering that all the portable devices can have a Bluetooth interface and people are highly mobile these days, it is not uncommon to find a cluster of piconets of both the 79-hop and the 23-hop types in the same area. In this paper, we present an analytical model of packet interference in a heterogeneous cluster of Bluetooth masters. By a heterogeneous cluster, we mean a cluster of piconets consisting of 23-hop and 79-hop types. Our analytic model is based on the idea of probabilistic graphs, where a node denotes a piconet and an edge denotes the probability of interference between two nodes.