Search NASA⌕ Search

SEARCH · Search NASA

Results for “COMMUNICATION NETWORKS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,279 records · Page 71

Molecular Communication over Gas Stream Channels using Portable Mass Spectrometry

The synthetic generation/coding and transmission of olfactory information over a gas stream or an odor network is a new and unexplored field. Application areas vary from the entertainment or advertisement industry to security and telemedicine. However, current technological limitations frustrate the accurate reproduction of decoded and transmitted olfactory data. This study describes the development, testing, and characterization of a novel odor emitter (OE) that is used to investigate the generation-encoding of gaseous standards with odorous characteristics with a regulatable way, for scent transmission purposes. The calibration and the responses of a developed OE were examined using a portable quadrupole mass spectrometer (MS). Experiments were undertaken for a range of volatile organic compounds (VOCs) at different temperatures and flow rates. Individual compounds and mixtures were tested to investigate periodic and dynamic transmission characteristics within two different size tubular containers for distances up to 3 m. Olfactory information transmission is demonstrated using MS as the main molecular sensor for odor detection and monitoring and for the first time spatial encryption of olfactory information is shown. .

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A general-purpose method for Pareto optimal placement of flow rate and concentration sensors in networked systems – With application to wastewater treatment plants

The advent of affordable computing, low-cost sensor hardware, and high-speed and reliable communications have spurred installation of ubiquitous sensors in complex engineered systems. However, ensuring reliable data quality remains a challenge. Exploitation of redundancy among sensor signals can help improving the precision of measured variables, detecting the presence of gross errors, and identifying faulty sensors. The cost of sensor ownership, maintenance efforts in particular, can still be cost-prohibitive however. Maximizing the ability to assess and control data quality while minimizing the cost of ownership thus requires a careful sensor placement. To solve this challenge, in this work we develop a generally applicable method to solve the multi-objective sensor placement problem in systems governed by linear and bilinear balance equations. Importantly, the method computes all Pareto-optimal sensor layouts with conventional computational resources and requires no information about the expected sensor quality.

42 ENGINEERING↗

SBIR Phase II Final Report - Multi-Protocol Energy Management Gateway for Home-Area Networks

Significant amounts of electricity, natural gas, and heating oil are wasted by homeowners due to inefficient operation and inadequate maintenance of heating, ventilation, and air conditioning (HVAC) equipment. Coincident’s work under this award reduces energy waste, saves consumers money, and reduces carbon emissions. It does so in three ways: First, Coincident’s approach replaces the traditional thermostat with a wireless network of sensors and controllers that measure temperature, humidity and occupancy in multiple rooms in the house. The “Internet of Things” is a technology trend holding the promise of ubiquitous inexpensive sensors. The reality, however, is that energy and HVAC monitoring and management is a patchwork of incompatible protocols and expensive proprietary technologies. Coincident’s multi-protocol architecture, developed in part under this award tackles this problem and brings low cost interoperable sensor and control devices to market. Second, the Coincident system eliminates hard-to-program and rigid thermostat schedules and instead provides automatic operation of heating and cooling by combining individual temperature and comfort preferences with energy-saving targets, real-time utility use information, weather data, and room utilization patterns. Energy efficiency technology must be appealing to consumers otherwise it will not be used. The Coincident user interface has engaging features such as remote control from any smart phone or web browser and per-room performance breakdowns. Expected energy savings resulting from more efficient operation of heating and air conditioning equipment are in the range of 10-20%. Third, the Coincident system provides heating and air-conditioning contractors with fine-grained performance data for every residence they support (subject to customer privacy controls). This data is integrated from diverse networks within the residence and includes HVAC performance and fuel use data. This information allows the partner to validate energy savings and identify potential system faults (whether from installation problems or maintenance issues). When combined with professional installation as part of high-efficiency HVAC upgrade, energy savings levels of 20-30% can be achieved. Economic feasibility of energy efficiency technology is one of the key challenges addressed in this award. The Coincident system is engineered to be delivered at a disruptive price point, making the system financially feasible for new and retrofit homes of all types and sizes. The Coincident system is intended to be sold through the HVAC professional—the industry most capable of improving HVAC efficiency. Providing HVAC contractors with detailed home performance data motivates them to sell the product, provides them with maintenance and upgrade revenue opportunities, and therefore delivers customer savings and environmental benefits. Having demonstrated technical and financial feasibility, Coincident has won additional grants and awards, participated in pilot projects, started partnership discussions with several HVAC equipment vendors, and has lined up several large channel partners ready to participate in large pilot rollouts.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Scalable multiscale modeling of platelets with 100 million particles

Here, we developed the core components of the AI-aided multiple time stepping algorithm for multiscale modeling of cell dynamics. This algorithm was implemented and analyzed on two supercomputer architectures with an application of simulating the aggregation of 250 platelets, or 102 million particles. To scale on these computers with complex memory and network architectures with GPUs, we devised a biomechanics-informed task mapping scheme to optimize load imbalance, communications, and memory utilization. Our simulations, scaling well up to 192 nodes on a Summit-like supercomputer with a peak speed of 11 petaflops, achieved a rate of 423 μs/day which is 500 times faster than the conventional algorithm using static time step and this has enabled studies of record size blood clots at record spatial–temporal resolutions. Additionally, we discovered the sensitive dependence of the scalability and execution time on the methods of decomposition, CPU–GPU coupling, and task mapping.

97 MATHEMATICS AND COMPUTING↗

PowerMappeR: Power-Optimized Mapping of SNNs onto ReRAM Crossbars coupled via Packet-Switched NoCs

Many recent efforts in developing hardware-accelerated spiking neural networks (SNNs) are characterized by deep co-design between algorithms, architectures, and devices. Architectural advances overcome device constraints by coupling together many small resistive-RAM (ReRAM) crossbars via a network-on-chip (NoC) for neuromorphic component operation. Concurrently, improved SNN training methods increase accuracy and structural sparsity in networks despite growing problem sizes. Finally, compilers leverage these attributes to minimize area and inter-crossbar communication while mapping large SNNs to sophisticated architectures. However, for compiler-driven co-design to realize increasingly complex and profitable optimizations, a compile-time view of power consumption is critical. We present PowerMappeR to express and optimize over mapping-, architecture-, and device-specific power consumption information. By modeling the dynamic power of well-established components, we develop an integer linear programming (ILP)-based, encoding-agnostic, parametric power estimation model. Using this model, we demonstrate practical improvements in area and inter-crossbar communication by 0%–9.5% and 1.4%–5.1%, respectively. We also limit hotspot formation during optimization, achieving comparable or better results in targeted metrics with up to 96.4%–97.1% restriction of hotspot magnitude. Finally, we introduce profile-guided formulations to reduce worst-case and expected-case hotspot magnitude by 40.7%–69.5% and 40.6%–56.3%, respectively. Optimizing worst-case hotspot magnitude incidentally improves expected-case magnitude by 10.85%–33.45%. Reciprocally, optimizing expected-case magnitude incidentally improves worst-case magnitude by 4.33%–39.87%. Validation against hardware simulators confirms that PowerMappeR can decrease dynamic power consumption by 12.6%–27.3%.

Pohl, Devin [ORNL] (ORCID:0009000040149027)↗

Revolutionize Situational Awareness in Emergencies

This report describes an integrated system that provides real-time actionable information to first responders. LANL will integrate three technologies to form an advanced predictive real-time sensor network including compact chemical and wind sensor sin low cost rugged package for outdoor installation; flexible robust communication architecture linking sensors in near-real time to globally accessible servers; and the QUIC code which predicts contamination transport and dispersal in urban environments in near real time.

42 ENGINEERING↗

Bad data packet capture device

An apparatus and method for capturing data packets for analysis on a network computing system includes a sending node and a receiving node connected by a bi-directional communication link. The sending node sends a data transmission to the receiving node on the bi-directional communication link, and the receiving node receives the data transmission and verifies the data transmission to determine valid data and invalid data and verify retransmissions of invalid data as corresponding valid data. A memory device communicates with the receiving node for storing the invalid data and the corresponding valid data. A computing node communicates with the memory device and receives and performs an analysis of the invalid data and the corresponding valid data received from the memory device.

97 MATHEMATICS AND COMPUTING↗

Efficient Certificate Verification for Vehicle-to-Grid Communications

While public charging stations are typically used for Electric Vehicle (EV) such as charging, home microgrids that may act as private charging stations are also expected to be used for meeting the increased EV charging demands in the future. Such home microgrids can be accessible through their smart meters, which makes advanced metering infrastructure (AMI) a viable alternative for vehicle-to-grid (V2G) communications. However, to ensure secure V2G communications using public-keys, smart meters will need to maintain certificate revocation lists (CRLs) not just for the AMI network but also for large number of EVs that may interact with them. For resource-constrained smart meters, this will increase the storage requirements and introduce additional overhead in terms of delay and CRL maintenance. To eliminate this burden, we propose keeping merely non-revoked certificates that belong to EVs, which are usually driven within the vicinity of that particular microgrid. The motivation comes from the fact that it is inefficient to distribute and store a large CRL that has revocation information about all EVs in the whole system as most of these EVs will never come to the geographic vicinity of that home microgrid. The approach ensures that any status changes of these certificates are communicated to the smart meters. We implemented the proposed approach in a realistic V2G communication scenario by using IEEE 802.11s mesh as the underlying AMI infrastructure using ns-3 simulator. The results confirmed that the proposed approach significantly reduces the certificate verification time and the storage requirements on smart meters.

25 ENERGY STORAGE↗

The Citizen Cyberscience Lectures - 1) Mobile phones and Africa: a success story 2) Citizen Problem Solving

Mobile phones and Africa: a success story Dr. Mo Ibrahim, Mo Ibrahim Foundation Citizen Problem Solving Dr. Alpheus Bingham, InnoCentive The Citizen Cyberscience Lectures are hosted by the partners of the Citizen Cyberscience Centre, CERN, The UN Institute of Training and Research and the University of Geneva. The goal of the Lectures is to provide an inspirational forum for participants from the various international organizations and academic institutions in Geneva to explore how information technology is enabling greater citizen participation in tackling global development challenges as well as global scientific research. The first Citizen Cyberscience Lectures will welcome two speakers who have both made major innovative contributions in this area. Dr. Mo Ibrahim, founder of Celtel International, one of Africa’s most successful mobile network operators, will talk about “Mobile phones and Africa: a success story”. Dr. Alpheus Bingham, founder of InnoCentive, a Web-based community that solves industrial R&D; challenges, will discuss “Citizen Problem Solving”. The Citizen Cyberscience Lectures are open and free of charge. Participants from outside CERN must register by sending an email to Yasemin.Hauser@cern.ch BEFORE the 23rd october to be able to access CERN. THE LECTURES Mobile phones and Africa: a success story Dr. Mo Ibrahim, Mo Ibrahim Foundation Abstract The introduction of mobile phones into Africa changed the continent, enabling business and the commercial sector, creating directly and indirectly, millions of jobs. It enriched the social lives of many people. Surprisingly, it supported the emerging civil society and advanced the course of democracy Bio Dr Mo Ibrahim is a global expert in mobile communications with a distinguished academic and business career. In 1998, Dr Ibrahim founded Celtel International to build and operate mobile networks in Africa. Celtel became one of Africa’s most successful companies with operations in 15 countries, covering more than a third of the continent’s population and investing more than US 750 millionin Africa.The company was sold to MTC Kuwaitin 2005 for 3.4billion. In 2006 Dr Ibrahim established the Mo Ibrahim Foundation to support great African leadership. The Foundation focuses on two major initiatives to stimulate debate around, and improve the quality of, governance in Africa. The Ibrahim Prize for Achievement in African Leadership recognises and celebrates excellence; and the Ibrahim Index of African Governance provides civil society with a comprehensive and quantifiable tool to promote government accountability. Dr Ibrahim is also Founding Chairman of Satya Capital Ltd, an investment company focused on opportunities in Africa. Dr Ibrahim has been awarded an Honorary Doctorate by the University of London’s School of Oriental and African Studies, the University of Birmingham and De Montfort University, Leicester as well as an Honorary Fellowship Award from the London Business School. He has also received the Chairman’s Award for Lifetime Achievement from the GSM Association in 2007 and the Economists Innovation Award 2007 for Social & Economic Innovation. In 2008 Dr Ibrahim was presented with the BNP Paribas Prize for Philanthropy, and also listed by TIME magazine as one of the 100 most influential people in the world. Citizen Problem Solving Dr. Alpheus Bingham, InnoCentive Abstract American playwright Damien Runyon (Guys and Dolls) once remarked, "the race is not always to the swift, nor the victory to the strong -- but that IS how you bet." Not only does a system of race handicapping follow from this logic, but the whole notion of expertise and technical qualifications. Such 'credentials' allow one to 'bet' on who might most likely solve a difficult challenge, whether as consultant, contractor or employee. Of course, the approach would differ if one were allowed to bet AFTER the race. When such systems came into broad use, i.e., chat rooms, usenets, innocentive, etc., and were subsequently studied, it was often found that the greatest probability of solution lies in the "long tail" of the function rather than in the head representing formally vetted 'experts.' Insight into a problem is often the intersection of training, experience, metaphor and provocation (think Archimedes). Examples of "citizens" outside a targeted field of expertise providing uniques solutions will illustrate the principles involved. Bio Dr. Alph Bingham is a pioneer in the field of open innovation and an advocate of collaborative approaches to research and development. He is co-founder, and former president and chief executive officer of InnoCentive Inc., a Web-based community that matches companies facing R&D; challenges with scientists who propose solutions. Through InnoCentive, a platform that leverages the ability to connect to a whole planet of people through the Internet, organizations can access individuals – problem solvers – who might never have been found. Alph spent more than 25 years with Eli Lilly and Company, and offers deep experience in pharmaceutical research and development, research acquisitions and collaborations, and R&D; strategic planning. During his career he was instrumental in creating and developing Eli Lilly's portfolio management process as well as establishing the divisions of Research Acquisitions, the Office of Alliance Management and e.Lilly, a business innovation unit, from which various other ventures were spun out that create the advantages of open and networked organizational structures, including: InnoCentive, YourEncore, Inc., Coalesix, Inc., Maaguzi, Inc., Indigo Biosystems, Seriosity, Chorus and Collaborative Drug Discovery, Inc. He currently serves on the Board of Directors of InnoCentive, Inc., and Collaborative Drug Discovery, Inc.; the advisory boards of the Center for Collective Intelligence (MIT), and the Business Innovation Factory, as well as a member of the board of trustees of the Bankinter Foundation for Innovation in Madrid. He has lectured extensively at both national and international events and serves as a Visiting Scholar at the National Center for Supercomputing Application at the University of Illinois at Champaign-Urbana. He is also the former chairman of the Board of Editors of the Research Technology Management Journal. Dr. Bingham was the recipient of the Economist's Fourth Annual Innovation Summit "Business Process Award" for InnoCentive. He was also named as one of Project Management Institute's "Power 50" leaders in October 2005. Dr. Bingham received a Ph.D. in organic chemistry from Stanford University.

99 GENERAL AND MISCELLANEOUS↗

Sparsity-Promoting Optimal Wide-Area Control of Power Networks

Inter-area oscillations in bulk power systems are typically poorly controllable by means of local decentralized control. Recent research efforts have been aimed at developing wide-area control strategies that involve communication of remote signals. In conventional wide-area control, the control structure is fixed a priori typically based on modal criteria. In contrast, here we employ the recently-introduced paradigm of sparsity-promoting optimal control to simultaneously identify the optimal control structure and optimize the closed-loop performance. To induce a sparse control architecture, we regularize the standard quadratic performance index with an ℓ 1 -penalty on the feedback matrix. The quadratic objective functions are inspired by the classic slow coherency theory and are aimed at imitating homogeneous networks without inter-area oscillations. We use the New England power grid model to demonstrate that the proposed combination of the sparsity-promoting control design with the slow coherency objectives performs almost as well as the optimal centralized control while only making use of a single wide-area communication link. In addition to this nominal performance, we also demonstrate that our control strategy yields favorable robustness margins and that it can be used to identify a sparse control architecture for control design via alternative means.

25 ENERGY STORAGE↗

Space-based quantum networking in the presence of a nuclear disturbed environment

Space-based quantum networks provide a means for near-term long-distance transmission of quantum information. Here this article analyzed the performance of a downlink quantum network between a low-Earth-orbit satellite and an observatory operating in less-than-ideal atmospheric conditions. The effects from fog, haze, and a nuclear disturbed environment on the long-range distribution of quantum states were investigated. A density matrix that estimates the quantum state by capturing the effects from increased signal loss and elevated background noise to estimate the state fidelity of the transmitted quantum state was developed. It was found that the nuclear disturbed environment and other atmospheric effects have a degrading effect on the quantum state. These environments impede the ability to perform quantum communications for the duration of the effects. In the case of the nuclear disturbed environment, the nuclear effects subside quickly, and network performance should return to normal by the next satellite pass.

42 ENGINEERING↗

Cyber-Physical Test Platform for Microgrids: Combining Hardware, Hardware-in-the-Loop, and Network-Simulator-in-the-Loop

This paper presents a cyber-physical testbed, developed to investigate the complex interactions between emerging microgrid technologies such as grid-interactive power sources, control systems, and a wide variety of communication platforms and bandwidths. The cyber-physical testbed consists of three major components for testing and validation: real time models of a distribution feeder model with microgrid assets that are integrated into the National Renewable Energy Laboratory's (NREL) power hardware-in-the-loop (PHIL) platform; real-time capable network-simulator-in-the-loop (NSIL) models; and physical hardware including inverters and a simple system controller. Several load profiles and microgrid configurations were tested to examine the effect on system performance with increasing channel delays and router processing delays in the network simulator. Testing demonstrated that the controller's ability to maintain a target grid import power band was severely diminished with increasing network delays and laid the foundation for future testing of more complex cyber-physical systems.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Parallel Agent-Based Simulations on Clusters of GPUs and Multi-Core Processors

An effective latency-hiding mechanism is presented in the parallelization of agent-based model simulations (ABMS) with millions of agents. The mechanism is designed to accommodate the hierarchical organization as well as heterogeneity of current state-of-the-art parallel computing platforms. We use it to explore the computation vs. communication trade-off continuum available with the deep computational and memory hierarchies of extant platforms and present a novel analytical model of the tradeoff. We describe our implementation and report preliminary performance results on two distinct parallel platforms suitable for ABMS: CUDA threads on multiple, networked graphical processing units (GPUs), and pthreads on multi-core processors. Message Passing Interface (MPI) is used for inter-GPU as well as inter-socket communication on a cluster of multiple GPUs and multi-core processors. Results indicate the benefits of our latency-hiding scheme, delivering as much as over 100-fold improvement in runtime for certain benchmark ABMS application scenarios with several million agents. This speed improvement is obtained on our system that is already two to three orders of magnitude faster on one GPU than an equivalent CPU-based execution in a popular simulator in Java. Thus, the overall execution of our current work is over four orders of magnitude faster when executed on multiple GPUs.

97 MATHEMATICS AND COMPUTING↗

Stop managing people, start managing agreements - 15081

One definition of management in business and organizations is: 'the function that coordinates the efforts of people to accomplish goals and objectives using available resources efficiently and effectively'. Unfortunately, what it takes to coordinate (literally, to create order together) is beyond the scope of today's management tools. The top five management tools in the world, according to a Bain 2013 survey, are: Strategic Planning; Customer Relationship Management; Employee Engagement Surveys; Benchmarking; and Balanced Scorecard. These are useful tools in business and organizations, but do not help managers coordinate the complex performance networks of agreements with resource providers, team members, and internal and external user-customers and authorities required to accomplish group or organizational goals. How do we manage such a network? While coordination is increasingly necessary in today's world of work, it is also a much neglected area of serious study because it is, almost literally, an invisible phenomenon. We can observe people inside groups, as well as their activities and results, but coordination is a product of managing what is going on between groups, not inside them. The challenge is making that in-between visible and actionable, not just 'white space'. A solution: define a Team's performance as agreements - and deliveries on the agreements - for products, services, and communications - the things moving between a Team and its immediate circle of resource providers, customers, and authorities relevant to a goal. This approach has been shown to reduce operational costs while also increasing team members' recognition of their interdependent responsibilities. Managers and Team leaders can then turn any Team goal into an opportunity for accomplishment in three steps: 1. Define 'performance' in terms of deliverable products, services, and communications for every goal; 2. Keep performance status current and visible to all team members; and 3. Debrief and update performance status routinely. Teamwork and engagement in goal accomplishment require a focus on Sender-Receiver agreements, productive communication, and timely and accurate status-tracking and feedback. Sustaining the context for accomplishment is not about managing people - it requires supporting coordination for the network of agreements in which people operate. (authors)

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Performance analysis of distributed symmetric sparse matrix vector multiplication algorithm for multi-core architectures

In this article, sparse matrix vector multiply (SpMVM) is an important kernel that frequently arises in high performance computing applications. Due to its low arithmetic intensity, several approaches have been proposed in literature to improve its scalability and efficiency in large scale computations. In this paper, our target systems are high end multi-core architectures and we use messaging passing interface + open multiprocessing hybrid programming model for parallelism. We analyze the performance of recently proposed implementation of the distributed symmetric SpMVM, originally developed for large sparse symmetric matrices arising in ab initio nuclear structure calculations. We also study important features of this implementation and compare with previously reported implementations that do not exploit underlying symmetry. Our SpMVM implementations leverage the hybrid paradigm to efficiently overlap expensive communications with computations. Our main comparison criterion is the "CPU core hours" metric, which is the main measure of resource usage on supercomputers. We analyze the effects of topology-aware mapping heuristic using simplified network load model. Furthermore, we have tested the different SpMVM implementations on two large clusters with 3D Torus and Dragonfly topology. Our results show that the distributed SpMVM implementation that exploits matrix symmetry and hides communication yields the best value for the "CPU core hours" metric and significantly reduces data movement overheads.

97 MATHEMATICS AND COMPUTING↗

Scaling Deep Learning on GPU and Knights Landing clusters

The speed of deep neural networks training has become a big bottleneck of deep learning research and development. For example, training GoogleNet by ImageNet dataset on one Nvidia K20 GPU needs 21 days. To speed up the training process, the current deep learning systems heavily rely on the hardware accelerators. However, these accelerators have limited on-chip memory compared with CPUs. To handle large datasets, they need to fetch data from either CPU memory or remote processors. We use both self-hosted Intel Knights Landing (KNL) clusters and multi-GPU clusters as our target platforms. From an algorithm aspect, current distributed machine learning systems are mainly designed for cloud systems. These methods are asynchronous because of the slow network and high fault-tolerance requirement on cloud systems. We focus on Elastic Averaging SGD (EASGD) to design algorithms for HPC clusters. Original EASGD used round-robin method for communication and updating. The communication is ordered by the machine rank ID, which is inefficient on HPC clusters. First, we redesign four efficient algorithms for HPC systems to improve EASGD's poor scaling on clusters. Async EASGD, Async MEASGD, and Hogwild EASGD are faster \textcolor{black}{than} their existing counterparts (Async SGD, Async MSGD, and Hogwild SGD, resp.) in all the comparisons. Finally, we design Sync EASGD, which ties for the best performance among all the methods while being deterministic. In addition to the algorithmic improvements, we use some system-algorithm codesign techniques to scale up the algorithms. By reducing the percentage of communication from 87% to 14%, our Sync EASGD achieves 5.3x speedup over original EASGD on the same platform. We get 91.5% weak scaling efficiency on 4253 KNL cores, which is higher than the state-of-the-art implementation.

60 APPLIED LIFE SCIENCES↗

A Redundant Communication Approach to Scalable Fault Tolerance in PGAS Programming Models

Recent trends in high-performance computing point towards increasingly large machines with millions of processing, storage, and networking elements. Unfortunately, the reliability of these machines is inversely proportional to their size, resulting in a system-wide mean-time-between-failures (MTBF) ranging from a few days to a few hours. As such, for long-running applications, the ability to efficiently recover from frequent failures is essential. Traditional forms of fault tolerance, such as checkpoint/restart, suffer from performance issues related to limited I/O and memory bandwidth. In this paper, we present a fault-tolerance mechanism that reduces the cost of failure recovery by maintaining shadow data structures and performing redundant remote memory accesses. We present results from a computational chemistry application running at scale to show that our techniques provide applications with a high degree of fault tolerance and low (2%--4%) overhead for 2048 processors.

99 GENERAL AND MISCELLANEOUS↗

Genetic algorithm based task reordering to improve the performance of batch scheduled massively parallel scientific applications

The growth in size of networked high performance computers along with novel accelerator-based node architectures has further emphasized the importance of communication efficiency in high performance computing. The world's largest high performance computers are usually operated as shared user facilities due to the costs of acquisition and operation. Applications are scheduled for execution in a shared environment and are placed on nodes that are not necessarily contiguous on the interconnect. Furthermore, the placement of tasks on the nodes allocated by the scheduler is sub-optimal, leading to performance loss and variability. Here, we investigate the impact of task placement on the performance of two massively parallel application codes on the Titan supercomputer, a turbulent combustion flow solver (S3D) and a molecular dynamics code (LAMMPS). Benchmark studies show a significant deviation from ideal weak scaling and variability in performance. The inter-task communication distance was determined to be one of the significant contributors to the performance degradation and variability. A genetic algorithm-based parallel optimization technique was used to optimize the task ordering. This technique provides an improved placement of the tasks on the nodes, taking into account the application's communication topology and the system interconnect topology. As a result, application benchmarks after task reordering through genetic algorithm show a significant improvement in performance and reduction in variability, therefore enabling the applications to achieve better time to solution and scalability on Titan during production.

97 MATHEMATICS AND COMPUTING↗