Parallel Large-Scale Graph Processing Workshop
The workshop is intended to promote an open discussion of a large-scale graph processing using parallel computing. The main task of the workshop is to bring attention to the convergence of High Performance Computing, Big Data and Artificial Intelligence technologies and to encourage a discussion of the problems arising in parallel graph processing using these technologies. The workshop is a complementary to the GraphHPC conference.
The workshop will be held on September, 25th from 14:10 till 15:50.
Translation of Green-Marl Domain-specific Language for Graphs to Charm++ Computation Model
A. Frolov, JSC NICEVT
In the talk the results of the work on adaptation of the Green-Marl compiler to Charm++ computation model will be presented. Green-Marl is used for a parallel static graph analysis and adopts imperative shared memory programming model, while Charm++ implements a message-driven execution model. As an example for demonstration purposes the strongly-connected components problem will be used.
The comparison of large-scale graph processing algorithms implementation methods for Intel KNL and NVIDIA GPU
I. Afanasyev, V. Voevodin, Lomonosov Moscow State University
Current paper describes implementation approach to large-graph processing on two modern high-performance computational platforms: NVIDIA GPU and Intel KNL. The described approach is based on a deep a priori analysis of algorithm properties and potential, what helps to choose implementation method correctly. To demonstrate the proposed approach, shortest paths and strongly connected components computation problems have been solved for sparse graphs. The results include detailed description of the whole algorithm’s development cycle: from algorithms informational structure research and selection of optimal implementation methods, suitable for the particular platforms, to specific optimizations for each of the architectures. Based on the joint analysis of algorithm properties and architecture features, performance tuning, including graph storage format optimizations, efficient usage of memory hierarchy and vectorization is performed. Developed implementations demonstrate high performance and good scalability of the proposed solutions. In addition, a lot of attention was paid to profiling implemented algorithms with NVIDIA Visual Profiler and Intel® VTune ™ Amplifier utilities, which allow evaluating how optimally the developed algorithms are. This allows current paper to present a fair comparison, demonstrating advantages and disadvantages of each platform for large-scale graph processing.
Approaches to graph anomaly detection problem
A. Mazeev, A. Semenov, JSC NICEVT
Detecting anomalies in graphs is a vital task with numerous high-impact applications in the modern Big Data world. The approaches to the problem are categorized under various settings: unsupervised vs. (semi-)supervised, for static vs. dynamic graphs, for attributed vs. plain graphs. In the talk we will present a brief survey of graph based anomaly detection approaches and examine a typical supervised anomaly detection problem.
Benchmarking Apache Spark on cluster with Angara interconnect
A. Agarkov, A. Semenov, JSC NICEVT
Apache Spark is one of the most popular Big Data frameworks. Performance evaluation of Big Data frameworks is a topic of interest due to the increasing number and importance of data analytics applications within the context of HPC and Big Data convergence. In the talk we present performance evaluation of the algorithm for table association problem implemented in Apache Spark on the cluster with Angara interconnect.
Topics of interest
Topics of interest include, but are not limited to:
- Parallel graph algorithms
- Graph applications and High Performance Computing, Big Data, Artificial Intelligence
- Parallel programming models and runtime systems for graph processing
- Graph databases and its applications
- Performance evaluation of graph algorithms
- Graph applications and exascale
- Graph visualization
- May 15 – paper submission
- June 5 – author notification
- June 26 – camera-ready paper
Please see the RusSCDays-2017 submission terms.
- Voevodin V. V., Corresponding member of the Russian Academy of Sciences, MSU RCC (co-chair)
- Simonov A. S., PhD, JSC NICEVT (co-chair)
- Frolov A. S., DISLab (JSC NICEVT)
- Semenov A. S., PhD, DISLab (JSC NICEVT)
- Pozdneev A.V., PhD, IBM
- Daryin A. N., PhD, Yandex
- Korzh A. A., PhD, Micron
- Chernoskutov M. A., IMM Ural Dep. of RAS
Alexander Frolov, DISLab (JSC NICEVT), e-mail: alexndr.frolov at gmail com
Alexander Semenov, DISLab (JSC NICEVT), e-mail: alxdr.semenov at gmail com