Doruk Bozdag

Mobility-based communication in wireless sensor networks

Wireless sensor networks are proposed to deliver in situ observations at low cost over long periods of time. Among numerous challenges faced while designing WSNs and protocols, maintaining connectivity and maximizing the network lifetime stand out as critical considerations. Mobile platforms equipped with communication devices can be leveraged to overcome these two problems. In this article existing proposals that use mobility in WSNs are summarized. Furthermore, a new approach to compute mobile platform trajectories is introduced. These solutions are also compared considering various metrics and design goals.

Partitioning based mobile element scheduling in wireless sensor networks

In recent studies, using mobile elements (MEs) as mechanical carriers of data has been shown to be an effective way of prolonging sensor network life time and relaying information in partitioned networks. As the data generation rates of sensors may vary, some sensors need to be visited more frequently than others. In this paper, a partitioning-based algorithm is presented that schedules the movements of MEs in a sensor network such that there is no data loss due to buffer overflow. Simulation results show that the proposed Partitioning Based Scheduling (PBS) algorithm performs well in terms of reducing the minimum required ME speed to prevent data loss, providing high predictability in inter-visit durations, and minimizing the data loss rate for the cases when the ME is constrained to move slower than the minimum required ME speed.

Benchmarking short sequence mapping tools

BackgroundThe development of next-generation sequencing instruments has led to the generation of millions of short sequences in a single run. The process of aligning these reads to a reference genome is time consuming and demands the development of fast and accurate alignment tools. However, the current proposed tools make different compromises between the accuracy and the speed of mapping. Moreover, many important aspects are overlooked while comparing the performance of a newly developed tool to the state of the art. Therefore, there is a need for an objective evaluation method that covers all the aspects. In this work, we introduce a benchmarking suite to extensively analyze sequencing tools with respect to various aspects and provide an objective comparison.ResultsWe applied our benchmarking tests on 9 well known mapping tools, namely, Bowtie, Bowtie2, BWA, SOAP2, MAQ, RMAP, GSNAP, Novoalign, and mrsFAST (mrFAST) using synthetic data and real RNA-Seq data. MAQ and RMAP are based on building hash tables for the reads, whereas the remaining tools are based on indexing the reference genome. The benchmarking tests reveal the strengths and weaknesses of each tool. The results show that no single tool outperforms all others in all metrics. However, Bowtie maintained the best throughput for most of the tests while BWA performed better for longer read lengths. The benchmarking tests are not restricted to the mentioned tools and can be further applied to others.ConclusionThe mapping process is still a hard problem that is affected by many factors. In this work, we provided a benchmarking suite that reveals and evaluates the different factors affecting the mapping process. Still, there is no tool that outperforms all of the others in all the tests. Therefore, the end user should clearly specify his needs in order to choose the tool that provides the best results.

Hypergraph-based Dynamic Load Balancing for Adaptive Scientific Computations

Adaptive scientific computations require that periodic repartitioning (load balancing) occur dynamically to maintain load balance. Hypergraph partitioning is a successful model for minimizing communication volume in scientific computations, and partitioning software for the static case is widely available. In this paper, we present a new hypergraph model for the dynamic case, where we minimize the sum of communication in the application plus the migration cost to move data, thereby reducing total execution time. The new model can be solved using hypergraph partitioning with faced vertices. We describe an implementation of a parallel multilevel repartitioning algorithm within the Zoltan load-balancing toolkit, which to our knowledge is the first code for dynamic load balancing based on hypergraph partitioning. Finally, we present experimental results that demonstrate the effectiveness of our approach on a Linux cluster with up to 64 processors. Our new algorithm compares favorably to the widely used ParMETIS partitioning software in terms of quality, and would have reduced total execution time in most of our test cases.

A repartitioning hypergraph model for dynamic load balancing

In parallel adaptive applications, the computational structure of the applications changes over time, leading to load imbalances even though the initial load distributions were balanced. To restore balance and to keep communication volume low in further iterations of the applications, dynamic load balancing (repartitioning) of the changed computational structure is required. Repartitioning differs from static load balancing (partitioning) due to the additional requirement of minimizing migration cost to move data from an existing partition to a new partition. In this paper, we present a novel repartitioning hypergraph model for dynamic load balancing that accounts for both communication volume in the application and migration cost to move data, in order to minimize the overall cost. The use of a hypergraph-based model allows us to accurately model communication costs rather than approximate them with graph-based models. We show that the new model can be realized using hypergraph partitioning with fixed vertices and describe our parallel multilevel implementation within the Zoltan load balancing toolkit. To the best of our knowledge, this is the first implementation for dynamic load balancing based on hypergraph partitioning. To demonstrate the effectiveness of our approach, we conducted experiments on a Linux cluster with 1024 processors. The results show that, in terms of reducing total cost, our new model compares favorably to the graph-based dynamic load balancing approaches, and multilevel approaches improve the repartitioning quality significantly.

Mobile element based differentiated message delivery in wireless sensor networks

In recent years, mobile elements (MEs) have been proposed as mechanical carriers of data to prolong the lifetime of sensor networks and to overcome network partitioning problem. A scheduling approach is proposed in Y. Gu et al., (2005) for MEs to collect periodically generated data, also called regular messages (RMs), from nearby sensor nodes with no buffer overflow. However, increased delay in message delivery with ME-based communication compared to multi-hop communication may not be tolerated in some cases. Some messages can be more urgent than others due to critical values of the sensed data. Such messages maybe required to be delivered to the ME within a specified deadline. In this paper, this new problem of differentiated message delivery (DMD) considering both regular and urgent message collection is addressed. The proposed solution incorporates multi-hop communication into the ME scheduling problem. The investigated performance metrics are the minimum required ME speed to prevent data loss and guarantee the maximum tolerated urgent message delay, as well as urgent and regular message loss rates for a given ME speed. The proposed solution is shown to perform well in terms of these metrics in various network scenarios. Furthermore, comparisons with existing ME scheduling algorithms show that the proposed solution meets the urgent message delivery requirement with a reasonable increase in ME speed

A task duplication based bottom-up scheduling algorithm for heterogeneous environments

We propose a new duplication-based DAG scheduling algorithm for heterogeneous computing environments. Contrary to the traditional approaches, proposed algorithm traverses the DAG in a bottom-up fashion while taking advantage of task duplication and task insertion. Experimental results on random DAGs and three different application DAGs show that the makespans generated by the proposed DBUS algorithm are much better than those generated by the existing algorithms, HEFT, HCPFD and HCNF.

Compaction of Schedules and a Two-Stage Approach for Duplication-Based DAG Scheduling

Many DAG scheduling algorithms generate schedules that require prohibitively large number of processors. To address this problem, we propose a generic algorithm, SC, to minimize the processor requirement of any given valid schedule. SC preserves the schedule length of the original schedule and reduces processor count by merging processor schedules and removing redundant duplicate tasks. To the best of our knowledge, this is the first algorithm to address this highly unexplored aspect of DAG scheduling. On average, SC reduced the processor requirement 91, 82, and 72 percent for schedules generated by PLW, TCSD, and CPFD algorithms, respectively. SC algorithm has a low complexity (O{N}3) compared to most duplication-based algorithms. Moreover, it decouples processor economization from schedule length minimization problem. To take advantage of these features of SC, we also propose a scheduling algorithm SDS, having the same time complexity as SC. Our experiments demonstrate that schedules generated by SDS are only 3 percent longer than CPFD (O{N}4), one of the best algorithms in that respect. SDS and SC together form a two-stage scheduling algorithm that produces schedules with high quality and low processor requirement, and has lower complexity than the comparable algorithms that produce similar high-quality results.

A framework for scalable greedy coloring on distributed-memory parallel computers

We present a scalable framework for parallelizing greedy graph coloring algorithms on distributed-memory computers. The framework unifies several existing algorithms and blends a variety of techniques for creating or facilitating concurrency. The latter techniques include exploiting features of the initial data distribution, the use of speculative coloring and randomization, and a BSP-style organization of computation and communication. We experimentally study the performance of several specialized algorithms designed using the framework and implemented using MPI. The experiments are conducted on two different platforms and the test cases include large-size synthetic graphs as well as real graphs drawn from various application areas. Computational results show that implementations that yield good speedup while at the same time using about the same number of colors as a sequential greedy algorithm can be achieved by setting parameters of the framework in accordance with the size and structure of the graph being colored. Our implementation is freely available as part of the Zoltan parallel data management and load-balancing library.

Comparative analysis of biclustering algorithms

Biclustering is a very popular method to identify hidden co-regulation patterns among genes. There are numerous biclustering algorithms designed to undertake this challenging task, however, a thorough comparison between these algorithms is even harder to accomplish due to lack of a ground truth and large variety in the search strategies and objectives of the algorithms. In this paper, we address this less studied, yet important problem and formally analyze several biclustering algorithms in terms of the bicluster patterns they attempt to discover. We systematically formulate the requirements for well-known patterns and show the constraints imposed by biclustering algorithms that determine their capacity to identify such patterns. We also give experimental results from a carefully designed testbed to evaluate the power of the employed search strategies. Furthermore, on a set of real datasets, we report the biological relevance of clusters identified by each algorithm.