scholarly journals Characterizing and Mitigating Work Time Inflation in Task Parallel Programs

2013 ◽  
Vol 21 (3-4) ◽  
pp. 123-136 ◽  
Author(s):  
Stephen L. Olivier ◽  
Bronis R. de Supinski ◽  
Martin Schulz ◽  
Jan F. Prins

Task parallelism raises the level of abstraction in shared memory parallel programming to simplify the development of complex applications. However, task parallel applications can exhibit poor performance due to thread idleness, scheduling overheads, andwork time inflation– additional time spent by threads in a multithreaded computation beyond the time required to perform the same work in a sequential computation. We identify the contributions of each factor to lost efficiency in various task parallel OpenMP applications and diagnose the causes of work time inflation in those applications. Increased data access latency can cause significant work time inflation in NUMA systems. Our locality framework for task parallel OpenMP programs mitigates this cause of work time inflation. Our extensions to the Qthreads library demonstrate that locality-aware scheduling can improve performance up to 3X compared to the Intel OpenMP task scheduler.

Author(s):  
Mark Endrei ◽  
Chao Jin ◽  
Minh Ngoc Dinh ◽  
David Abramson ◽  
Heidi Poxon ◽  
...  

Rising power costs and constraints are driving a growing focus on the energy efficiency of high performance computing systems. The unique characteristics of a particular system and workload and their effect on performance and energy efficiency are typically difficult for application users to assess and to control. Settings for optimum performance and energy efficiency can also diverge, so we need to identify trade-off options that guide a suitable balance between energy use and performance. We present statistical and machine learning models that only require a small number of runs to make accurate Pareto-optimal trade-off predictions using parameters that users can control. We study model training and validation using several parallel kernels and more complex workloads, including Algebraic Multigrid (AMG), Large-scale Atomic Molecular Massively Parallel Simulator, and Livermore Unstructured Lagrangian Explicit Shock Hydrodynamics. We demonstrate that we can train the models using as few as 12 runs, with prediction error of less than 10%. Our AMG results identify trade-off options that provide up to 45% improvement in energy efficiency for around 10% performance loss. We reduce the sample measurement time required for AMG by 90%, from 13 h to 74 min.


Author(s):  
Jing Chen ◽  
Pirah Noor Soomro ◽  
Mustafa Abduljabbar ◽  
Madhavan Manivannan ◽  
Miquel Pericas

2016 ◽  
Vol 39 (3) ◽  
pp. 334-343 ◽  
Author(s):  
Rafal Cupek ◽  
Kamil Folkert ◽  
Marcin Fojcik ◽  
Tomasz Klopot ◽  
Grzegorz Polaków

Classical control applications with a centralized logic and distributed input/output system are being replaced by dynamic environments of cooperating components. Thus, the OPC (Object Linking and Embedding for Process Control) UA (Unified Architecture) is becoming more popular, because the OPC Data Access substandard is not well suited for distributed systems. Moreover, in many production systems, redundant data servers are preferred, for financial and legal reasons. Providing performance evaluation gives an estimate of the time required (and data samples lost) to switch to a backup data source for redundant OPC UA architecture, depending on the failure detection method, number of variables and redundancy mode.


2017 ◽  
Vol 10 (3) ◽  
pp. 254-261
Author(s):  
Michaël Leblanc ◽  
Claude Lavoie

We experimentally tested the feasibility of a control campaign of purple jewelweed (Impatiens glandulifera), an exotic invasive species in Europe and North America. We evaluated the amount of time and money required to control the plant along riverbanks, with particular attention paid to the recovery of riparian vegetation following hand pulling and bagging. Work time was directly and significantly related to stem density and fresh biomass of the invader, but the relationship was stronger for density. Density and biomass were strongly reduced by the first hand-pulling operation from a mean of 45 to 2 stems m−2 and from a mean of 0.95 kg m−2 to nearly zero, a good performance but not enough to negate the need for a second hand pulling later in the summer. A single hand pulling significantly reduced the cover of purple jewelweed from to 30% to 7%. Riparian vegetation disturbed by the first hand pulling largely recovered during the following 30 d. Expressed over an area of 1 ha, the total amount of time required to control purple jewelweed is 1,400 work hours over 2 yr, or a minimum investment of Can$21,000 (US$17,000). Although controlling a well-established purple jewelweed population is expensive, to properly evaluate the benefits, we must also consider the costs of soil erosion caused by this species.


2020 ◽  
Vol 10 (1) ◽  
pp. 8
Author(s):  
Sultan Alamri

In many developing cities, the improvement of transport infrastructure is usually accompanied by major road construction and maintenance. This paper presents approaches and opportunities using peer-to-peer updating to improve spatial road networks undergoing construction and maintenance, which in turn will improve traffic flow and benefit cities overall. In many cities, the spatial road network requires maintenance, and these works often require traffic detours. With the current GPS (Global Positioning System) services, there is a noticeable delay in the updating of many spatial road networks. Thus, when a driver plans a trip to a certain location (such as Starbucks), his/her usual route may have changed, but the spatial road network in the GPS has not been updated. This can affect the user in many ways. For example, a trip that usually takes five minutes might now take half an hour, taking into account the additional time required to find alternative roads and possibly encountering more unexpected road closures, until the driver reaches his/her destination. This paper addresses this issue by proposing solutions that offer several advantages including a new peer-to-peer updating mechanism that helps to direct the driver to another route when road changes occur. Moreover, the peer-to-peer updating mechanism can enable the independent monitoring of road conditions and the updating of maps for newly-constructed roads, as well as the analysis of road congestions, traffic density, and people movements at certain times. Note that this work focuses on the conceptual ideas and approaches intended to improve independent maps, and the detailed algorithms have been left for future work.


2021 ◽  
Author(s):  
Benjamin Butler ◽  
Justin Roberts ◽  
Matthew Kelsey ◽  
Steffen Van Der Veen

Abstract Multilateral wells have been proven over decades and have developed into a reliable and cost effective approach to mature field rejuvenation and extended commercial viability. This paper will discuss case studies demonstrating a number of techniques used to create infill multilateral wells in existing fields with a high level of reliability and repeatability. Techniques reviewed will cover cutting and pulling production casing to drill and case a new mainbore versus sidetracking and adding laterals to an existing mainbore. Discussion will also cover completion designs that tie new laterals into existing production casing providing significantly greater reservoir contact. Temporary isolation of high water-cut laterals brought into production later in the well's life through bespoke completion designs will also be discussed. Case studies will include discussion of workover operations, isolation methods, and lateral creation systems. Where available, resulting field performance improvements will also be discussed. In Norway, slot recoveries are commonly performed by cutting and pulling the 10-3/4" casing, redrilling a new mainbore, and running new casing. This enables junction placement closer to unswept zones and easier lateral drilling to targets. It does have drawbacks, however, related to the additional time required to pull the subsea xmas tree and challenges associated with pulling casing. In 2019, Norway successfully completed a 10-3/4" retrofit installation, whereas a sidetrack was made from the 10-3/4" and an 8-5/8" expandable liner was run down into the reservoir pay zone where two new laterals were created. The 8-5/8" liner saved time otherwise spent having to drill the section down to the payzone from the laterals. These wells have a TAML Level 5 isolated junction, Autonomous Inflow Control Devices (AICDs) in each lateral, and an intelligent completion interface across the junction, enabling active flow management and monitoring of both branches. In Asia, infill laterals were added to existing wellbores by sidetracking 9-5/8" casing and tying production back to the original mainbore. These dual laterals were completed with intelligent completions to enable lateral flow management and monitoring of both laterals. In Australia, dual laterals were created in a similar fashion; laterals are added to existing wells; however, a novel approach was used to manage water from existing lower mainbore laterals whereby they are shut in at completion and opened later when the new lateral is watered out. The older lateral now produces at lower water cut given the time allowed for water coning in the lateral to relax. Using this practice, production is alternated back and forth between the two laterals. In the Middle East, an older well has been converted from TAML Level 4 to Level 5 in order to prevent detected gas migrating into the mainbore at the junction. This conversion of a cemented junction well has enabled production to resume on this well. The well was converted to incorporate an intelligent completion to enable flow control of each lateral. This paper intends to provide insights into the various mature field re-entry methods for multilateral well construction, and a review of the current technology capabilities and well designs through the review of multiple case histories.


Author(s):  
Matthias Schneider

IntroductionUsers of linked data require access to an increasing number of heterogeneous datasets from diverse domains, often held in different secure research data environments, especially for multi-jurisdictional projects. Under the traditional model of data access, projects are required to transfer and harmonise the necessary datasets in one central location before analysis can be undertaken, increasing the time required for data acquisition and preparation. Objectives and ApproachIn a federated data environment, analysts query distributed datasets held in a network of multiple secure data environments via a central virtual database, without requiring the data to move. Instead, the data is analysed as close as possible to its storage location, minimising the amount of data transfers and giving data custodians more control over their data. This symposium explores the challenges and opportunities of establishing and operating a distributed network of federated secure research data environments. Leading organisations operating data platforms in various jurisdictions present for 15 minutes each the current capabilities of their platforms, the landscape of data environments in their jurisdictions and potential approaches to key questions such as: Harmonising/federating data sources Data security Data governance Discoverability/metadata Performance The audience is the then invited to participate in discussing the topic for the remaining 30 minutes. The following individuals have been approached to represent their organisations in this symposium: Professor David Ford, Swansea University: UK Secure eResearch Platform (UK SErP) Charles Victor, Institute for Clinical Evaluative Sciences (ICES): ICES Data & Analytic Virtual Environment (IDAVE) Professor Louisa Jorm, Centre for Big Data Research in Health, University of New South Wales: E-Research Institutional Cloud Architecture (ERICA) Professor Kimberlyn McGrail, Population Data BC: Secure Research Environment (SRE) Results / Conclusion / ImplicationsThis symposium will help formulate requirements for and barriers to distributed networks of federated secure research data environments, and create a foundation for data analytics across multiple platforms.


2009 ◽  
Vol 180 ◽  
pp. 012046 ◽  
Author(s):  
Michael Wilde ◽  
Ioan Raicu ◽  
Allan Espinosa ◽  
Zhao Zhang ◽  
Ben Clifford ◽  
...  

Sign in / Sign up

Export Citation Format

Share Document