Continuous performance monitoring for large-scale parallel applications

Dooley, Isaac; Lee, Chee Wai; Kalé, Laxmikant V.

doi:10.1109/hipc.2009.5433181

Cited by 3 publications

(4 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SC-based schemes are widely accepted because they are compatible with standard protocols. However, they require the interaction between the foreign server and the home server, which may lead to the single point of failure [34], and induce large authentication transmission overhead because of the long distance between the foreign server and the home server. Moreover, recent studies [35], [36] have shown that SC-based schemes cannot provide strong user anonymity and non-traceability, and most of them cannot provide session key security and resistance to sophisticated attacks.…”

Section: Related Workmentioning

confidence: 99%

CPAL: A Conditional Privacy-Preserving Authentication With Access Linkability for Roaming Service

Lai

Liang

et al. 2014

IEEE Internet Things J.

View full text Add to dashboard Cite

Abstract-The roaming service enables mobile subscribers to access the Internet service anytime and anywhere, which can fulfill the requirement of ubiquitous access for the emerging paradigm of networking, e.g., the Internet of Things (IoT). In this paper, we propose a Conditional Privacy-preserving authentication with Access Linkability (CPAL) for roaming service, to provide universal secure roaming service and multi-level privacy preservation. CPAL provides an anonymous user linking function by utilizing a novel group signature technique, which can not only efficiently hide users' identities, but also enable the authorized entities to link all the access information of the same user without knowing the user's real identity. Specifically, by using the master linking key possessed by the trust linking server, the authorized foreign network operators or service providers can link the access information from the user to improve its service, while preserving user anonymity, e.g., using individual access information to analyze user preferences without revealing user's identity. Furthermore, the subscribers can also use this functionality to anonymously query their usage of service. In addition, CPAL has an efficient revocation function which revokes a group of users at the same time. Through extensive analysis, we demonstrate that CPAL resists various security threats, and provides more flexible privacy preservation compared to the existing schemes.Meanwhile, performance evaluations demonstrate its efficiency in terms of communication and computation overhead.

show abstract

Section: Related Workmentioning

confidence: 99%

CPAL: A Conditional Privacy-Preserving Authentication With Access Linkability for Roaming Service

Lai

Liang

et al. 2014

IEEE Internet Things J.

View full text Add to dashboard Cite

show abstract

“…The Charm++ developers have utilized LiveViz to build an in situ performance monitoring and visualization tool. [98] This tool gathers utilization statistics about the running program on every processor, efficiently compresses the utilization data, and merges the compressed data in a reduction from all the processors. During execution, the performance monitoring tool is just another module which runs alongside the application (like other Charm++ functionalities do), without affecting the application significantly.…”

Section: Charm ++mentioning

confidence: 99%

“…An AMT RTS provides a specific implementation of an AMT model. 11,12,16,17,22,26,97,98,101,[107][108][109]111 anti-dependency See Write-After-Read. 25, 26, Glossary: Write-After-Read API An application programmer interface (API) is set of functions and tools provided by a library developer to allow an application programmer to interact with a specific piece of software or allow a developer to utilize prebuilt functionality.…”

Section: Glossarymentioning

confidence: 99%

ASC ATDM Level 2 Milestone #5325: Asynchronous Many-Task Runtime System Analysis and Assessment for Next Generation Platforms

Baker

Bettencourt

Bova

et al. 2015

View full text Add to dashboard Cite

This report provides in-depth information and analysis to help create a technical road map for developing nextgeneration programming models and runtime systems that support Advanced Simulation and Computing (ASC) workload requirements. The focus herein is on asynchronous many-task (AMT) model and runtime systems, which are of great interest in the context of "exascale" computing, as they hold the promise to address key issues associated with future extreme-scale computer architectures. This report includes a thorough qualitative and quantitative examination of three best-of-class AMT runtime systems-Charm++, Legion, and Uintah, all of which are in use as part of the ASC Predictive Science Academic Alliance Program II (PSAAP-II) Centers. The studies focus on each of the runtimes' programmability, performance, and mutability. Through the experiments and analysis presented, several overarching findings emerge. From a performance perspective, AMT runtimes show tremendous potential for addressing extremescale challenges. Empirical studies show an AMT runtime can mitigate performance heterogeneity inherent to the machine itself and that Message Passing Interface (MPI) and AMT runtimes perform comparably under balanced conditions. From a programmability and mutability perspective however, none of the runtimes in this study are currently ready for use in developing production-ready Sandia ASC applications. The report concludes by recommending a codesign path forward, wherein application, programming model, and runtime system developers work together to define requirements and solutions. Such a requirements-driven co-design approach benefits the high-performance computing (HPC) community as a whole, with widespread community engagement mitigating risk for both application developers and runtime system developers.

show abstract

“…Even the most efficient implementation directly in hardware causes some overhead compared to regular floating-point arithmetic. Nevertheless, as shown in [14] , most CPUs do not contain a hardware implementation of subnormal numbers. In this case, a kernel trap is raised to allow the operating system kernel to handle a floating-point operation with a subnormal number.…”

Section: The Influence Of Subnormal Numbersmentioning

confidence: 99%

Comparison of eigensolvers for symmetric band matrices

Moldaschl

Gansterer

2014

Science of Computer Programming

View full text Add to dashboard Cite

We compare different algorithms for computing eigenvalues and eigenvectors of a symmetric band matrix across a wide range of synthetic test problems. Of particular interest is a comparison of state-of-the-art tridiagonalization-based methods as implemented in Lapack or Plasma on the one hand, and the block divide-and-conquer (BD&C) algorithm as well as the block twisted factorization (BTF) method on the other hand. The BD&C algorithm does not require tridiagonalization of the original band matrix at all, and the current version of the BTF method tridiagonalizes the original band matrix only for computing the eigenvalues.Avoiding the tridiagonalization process sidesteps the cost of backtransformation of the eigenvectors. Beyond that, we discovered another disadvantage of the backtransformation process for band matrices: In several scenarios, a lot of gradual underflow is observed in the (optional) accumulation of the transformation matrix and in the (obligatory) backtransformation step. According to the IEEE 754 standard for floating-point arithmetic, this implies many operations with subnormal (denormalized) numbers, which causes severe slowdowns compared to the other algorithms without backtransformation of the eigenvectors. We illustrate that in these cases the performance of existing methods from Lapack and Plasma reaches a competitive level only if subnormal numbers are disabled (and thus the IEEE standard is violated).Overall, our performance studies illustrate that if the problem size is large enough relative to the bandwidth, BD&C tends to achieve the highest performance of all methods if the spectrum to be computed is clustered. For test problems with well separated eigenvalues, the BTF method tends to become the fastest algorithm with growing problem size.

show abstract

Continuous performance monitoring for large-scale parallel applications

Cited by 3 publications

References 19 publications

CPAL: A Conditional Privacy-Preserving Authentication With Access Linkability for Roaming Service

CPAL: A Conditional Privacy-Preserving Authentication With Access Linkability for Roaming Service

ASC ATDM Level 2 Milestone #5325: Asynchronous Many-Task Runtime System Analysis and Assessment for Next Generation Platforms

Comparison of eigensolvers for symmetric band matrices

Contact Info

Product

Resources

About