Holger Brunst

The Vampir Performance Analysis Tool-Set

This paper presents the Vampir tool-set for performance analysis of parallel applications. It consists of the run-time measurement system VampirTrace and the visualization tools Vampir and VampirServer. It describes the major features and outlines the underlying implementation that is necessary to provide low overhead and good scalability. Furthermore, it gives a short overview about the development history and future work as well as related work.

Introducing the open trace format (OTF)

This paper introduces the new Open Trace Format. The first part provides a small overview about Trace Format Libraries in general and existing Formats/Libraries and their features. After that the important requirements are discussed. In particular it concerns efficient parallel and selective access to trace data. The following part presents design decisions and features of OTF comprehensively. Finally, there is some early evaluation of OTF. It features comparison of storage size for several examples as well as sequential and parallel I/O benchmarks. At the end, a conclusion will summarize the results and give some outlook.

Performance Optimization for Large Scale Computing: The Scalable VAMPIR Approach

Performance optimization remains one of the key issues in parallel computing. Many parallel applications do not benefit from the continually increasing peak performance of todays massively parallel computers, mainly because they have not been designed to operate efficiently on the 1000s of processors of todays top of the range systems. Conventional performance analysis is typically restricted to accumulated data on such large systems, severely limiting its use when dealing with real-world performance bottlenecks. Event based performance analysis can give the detailed insight required, but has to deal with extreme amounts of data, severely limiting its scalability. In this paper, we present an approach for scalable event-driven performance analysis that combines proven tool technology with novel concepts for hierarchical data layout and visualization. This evolutionary approach is being validated by implementing extensions to the performance analysis tool Vampir.

Tools for scalable parallel program analysis: Vampir NG, MARMOT, and DeWiz

Large-scale high-performance computing systems pose a tough obstacle for todays program analysis tools. Their demands in computational performance and memory capacity for processing program analysis data exceed the capabilities of standard workstations and traditional analysis tools. A comparison of the sophisticated approaches of Vampir NG (VNG), the Debugging Wizard DeWiz and the correctness-checking tool MARMOT provides novel ideas for scalable parallel program analysis. While VNG exploits the power of cluster architectures for near real-time performance analysis, DeWiz utilises distributed computing infrastructures for distinct analysis activities. MARMOT combines automatic runtime and partially distributed analysis.

Comprehensive Performance Tracking with Vampir 7

Vampir 7 is a performance visualization tool that provides a comprehensive view on the runtime behavior of parallel programs. It is a new member of the Vampir tool family. This new generation of performance visualizer combines state-of-the-art parallel data processing techniques with an all-new graphical user interface experience. This includes fast local and remote event data browsing, searching, filtering, clustering, and summarization. The software is ported to Unix, Windows, and Apple platforms. This article gives an overview of the novel techniques and features of Vampir 7.

Performance analysis of large-scale OpenMP and hybrid MPI/OpenMP applications with VampirNG

This paper presents a tool setup for comprehensive event-based performance analysis of large-scale openMP and hybrid openmp/ MPI applications. The KOJAK framework is used for portable code instrumentation and automatic analysis while the new VAMIIR NG infrastructure serves as generic visualization engine for both openMP and MPI performance properties. The tools share the same data base which enables a smooth transition from bottleneck auto-detection to manual in-depth visualization and analysis. With VAMPIR NG being a distributed data-parallel architecture, large problems on very large scale systems can be addressed.

High performance event trace visualization

This paper presents a joint effort to make huge event traces accessible for interactive program analysis. It combines a distributed software architecture with compressible data structures and customized query algorithms. The advanced technologies are discussed both theoretically and practically. Based on a proof-of-concept implementation the response times for typical queries are presented to show the practical relevance of this combined approach.

Scalable performance analysis of parallel systems: Concepts and experiences

Publisher Summary This chapter describes extended concepts and experiences to support performance analysis on systems with a couple of thousand processors. It presents a new tool architecture that has recently been developed in the ASCI/Earth simulator scope to support applications on very large machines and to ease the performance optimization process significantly. The major goals of the distributed parallel approach can be formulated as follows: (1) keep performance data close to the location where they were created, (2) perform event data analysis in parallel to achieve increased scalability, where speedups are on the order of 10 to 100, (3) limit the network bandwidth and latency requirements to a minimum to allow quick performance data browsing and analysis from remote working environments. VNG consists of two major components, an analysis server (vngd) and a visualization client (vng). The local visualization clients use a special purpose low latency protocol to communicate with the parallel analysis server, which allows for scalable remote data visualization in real-time. Distributed memory programming on the server side furthermore allows the application and the user to benefit from the huge memory and computation capabilities of todays cluster architectures without a loss of generality.

Online Remote Trace Analysis of Parallel Applications on High-Performance Clusters

The paper presents the design and development of an online remote trace measurement and analysis system. The work combines the strengths of the TAU performance system with that of the VNG distributed parallel trace analyzer. Issues associated with online tracing are discussed and the problems encountered in system implementation are analyzed in detail. Our approach should port well to parallel platforms. Future work includes testing the performance of the system on large-scale machines.

Event Tracing and Visualization for Cell Broadband Engine Systems

Event-based software tracing is a common technique for developing and optimizing parallel applications. It provides valuable information to application designers. This paper discusses software tracing on the Cell Broadband Engine, a heterogeneous multi-core processor, which is widely used in video game consoles, blade servers, and even supercomputer studies. However, the complex design of the Cell architecture poses challenging problems to developers. Our new monitoring approach improves this situation significantly as it visualizes Cell specific events on the SIMD cores that are usually hidden to programmers. We use the Vampir tool suite for visualization. Our design seamlessly integrates with the respective MPI monitor which additionally enables the tracking of large hybrid Cell/MPI applications.