In recent years, the number of CPU cores in a multi-core processor keeps increasing. To leverage the increasing hardware resource, programmers need to develop parallelized software programs. One promising approach to ...
详细信息
In recent years, the number of CPU cores in a multi-core processor keeps increasing. To leverage the increasing hardware resource, programmers need to develop parallelized software programs. One promising approach to parallelizing high-performance applications is pipeline parallelism, which divides a task into a serial of subtasks and then maps these subtasks to a group of CPU cores, making the communication scheme between the subtasks running on different cores a critical component for the parallelized programs. One widely-used implementation of the communication scheme is software-based, lock-free first-in-first-out queues that move data between different subtasks. The primary design goal of the prior lock-free queues was higher throughput, such that the technique of batching data was heavily used in their enqueue and dequeue operations. Unfortunately, a lock-free queue with batching heavily depends on the assumption that data arrive at a constant rate, and the queue is in an equilibrium state. Experimentally we found that the equilibrium state of a queue rarely happens in real, high-performance use cases (e.g., 10Gbps & x002B;network applications) because data arriving rate fluctuates sharply. As a result, existing queues suffer from performance degradation when used in real applications on multi-core processors. In this paper, we present EQueue, a lock-free queue to handle this robustness issue in existing queues. EQueue is lock-free, efficient, and robust. EQueue can adaptively (1) shrink its queue size when data arriving rate is low to keep its memory footprint small to utilize CPU cache better, and (2) enlarge its queue size to avoid overflow when data arriving rate is in burstiness. Experimental results show that when used in high-performance applications, EQueue can always perform an enqueue/dequeue operation in less than 50 CPU cycles, which outperforms FastForward and MCRingBuffer, two state-of-the-art queues, by factors 3 and 2, respectively.
In this paper, we propose a distributed power-efficient data gathering and aggregation algorithm (DPEG), in which a node, according to its residual energy and the strength of signal received from its neighboring nodes...
详细信息
A river model is a semi-distributed hydrological model and it includes many processes such as flow routing, irrigation diversion, overbank flow, ground water interaction for simulating flows a river system for water r...
详细信息
ISBN:
(纸本)9780987214355
A river model is a semi-distributed hydrological model and it includes many processes such as flow routing, irrigation diversion, overbank flow, ground water interaction for simulating flows a river system for water resources planning and management. A number of calibration parameters are introduced in such models to represent various processes using simplified mathematical equations. Traditionally, a river model is calibrated using a reach-by-reach calibration approach starting from the top of the system cascading down to the end of the system. While the reach-by-reach approach is suitable for obtaining optimum model performance at a single river reach with high quality observed data, it does have the limitation of error propagation from upstream to downstream reaches if poor quality data are used in the calibration. A system-wide calibration approach has recently been developed for river system modelling in large river basins. Comparing with traditional reach-by-reach calibration, this new method optimises parameters of all river reaches within a region simultaneously using a weighted global objective function. The results of its application of this new approach in the Murray-Darling basin, Australia have shown its potential to overcome over-fitting and improve fitness of each individual gauge. However, due to the system-wide optimization of multiple reach parameters in a region, the search space and computational time required for system calibration increase exponentially with the increase of number of parameters. This limits the number of parameters that can be optimised and thus, the size of the region. To potentially overcome this limitation, a parallelcomputing enabled shuffled complex evolution (SCE) optimisation tool has been developed. A series of comparison studies have been conducted to evaluate the performance of this approach over normal SCE. These are: 1) comparison of computation time and performance for the same number of parameters;2) comparison o
TACO is a template library that implements higher-order parallel operations on distributed object sets by means of reusable topology classes and C++ function templates. We discuss an experimental application that expl...
详细信息
TACO is a template library that implements higher-order parallel operations on distributed object sets by means of reusable topology classes and C++ function templates. We discuss an experimental application that exploits TACO's distributed object groups and collective operations for computing the similarity between groups of molecular sequences, a computationally intensive core problem in molecular biology research. In particular we show how TACO's distributed collections can be conveniently combined with well known concepts found in the C++ standard template library (STL) to solve matching and sorting problems effectively on distributed hardware platforms. The resulting implementation is concise and gives excellent parallel performance on PC- and workstation clusters.
Proper trust management in cloud computing environments can significantly assist their widespread adoption. Trust can act as a countermeasure to the several security threats that the cloud faces. This paper provides a...
详细信息
ISBN:
(纸本)9781509041534
Proper trust management in cloud computing environments can significantly assist their widespread adoption. Trust can act as a countermeasure to the several security threats that the cloud faces. This paper provides an overview of how trust has been applied in cloud computing. Trust models suggested in contemporary literature for cloud computing systems are presented. Furthermore, a critical comparison, based on the set of the main characteristics of an appropriate trust management method, is included.
As data continues to grow rapidly, NoSQL clusters have been increasingly adopted to address the storage and processing demands of these large amounts of data. In parallel, cloud computing is also increasingly being ad...
详细信息
ISBN:
(纸本)9781467379359
As data continues to grow rapidly, NoSQL clusters have been increasingly adopted to address the storage and processing demands of these large amounts of data. In parallel, cloud computing is also increasingly being adopted due to its flexibility, cost efficiency and scalability. However, evaluating and modelling NoSQL clusters present many challenges. In this work, we explore these challenges by performing a series of experiments with various configurations. The intuition is that this process is laborious and expensive and the goal of our experiments is to confirm this intuition and to identify the factors that impact the performance of a Big Data cluster. Our experiments mostly focus on three factors: data compression, data schema and cluster topology. We performed a number of experiments based on these factors and measured and compared the response times of the resulting configurations. Eventually, the outcomes of our study are encapsulated in a performance model that predicts the cluster's response time as a function of the incoming workload and evaluates the cluster's performance less costly and faster. This systematic and effortless evaluation method will facilitate the selection and migration to a better cluster as the performance and budget goals change. We use HBase as the large data processing cluster and we conduct our experiments on traffic data from a large city and on a distributed community cloud infrastructure.
Welcome to the NSF/TCPP Workshop on parallel and distributedcomputing Education (EduPar-20) proceedings. The EduPar-20 workshop, held in conjunction with the IEEE internationalparallel and computingsymposium (IPDPS...
详细信息
ISBN:
(数字)9781728174457
ISBN:
(纸本)9781728174570
Welcome to the NSF/TCPP Workshop on parallel and distributedcomputing Education (EduPar-20) proceedings. The EduPar-20 workshop, held in conjunction with the IEEE internationalparallel and computingsymposium (IPDPS), is devoted to the development and assessment of educational and curricular innovations and resources for undergraduate and graduate education in parallel and distributedcomputing (PDC). EduPar brings together individuals from academia, industry, and other educational and research institutes to explore new ideas, challenges, and experiences related to PDC pedagogy and curricula. The workshop is designed in coordination with the IEEE TCPP curriculum initiative on parallel and distributedcomputing (http://***/~tcpp/curriculum) for computer science and computer engineering undergraduates, and is supported by the NSF and the NSF-supported Center for parallel and distributedcomputing Curriculum Development and Educational Resources (CDER).
Grid computing "the next generation of the internet" has become an important topic in distributedcomputing. This paper explores how relational data can be exploited using Grid computing. This paper provides...
详细信息
Grid computing "the next generation of the internet" has become an important topic in distributedcomputing. This paper explores how relational data can be exploited using Grid computing. This paper provides a better understanding of Grid computing and reviews the potential effects of relational data within the Grid will have on future distributedcomputing. The experiences are taken from two ongoing projects: open grid services architecture - database access and integration (OGSA-DAI) and the European DataGrid's relational grid monitoring architecture (R-GMA). Both projects have early working implementations which can be exploited for commercial or scientific use.
暂无评论