Alexander S. Antonov

Bio: Alexander S. Antonov is an academic researcher from Moscow State University. The author has contributed to research in topics: Supercomputer & Parallel algorithm. The author has an hindex of 9, co-authored 27 publications receiving 275 citations.

Supercomputer Lomonosov-2: Large Scale, Deep Monitoring and Fine Analytics for the User Community

Abstract: The huge number of hardware and software components, together with a large number of parameters affecting the performance of each parallel application, makes ensuring the efficiency of a large scale supercomputer extremely difficult. In this situation, all basic parameters of the supercomputer should be constantly monitored, as well as many decisions about its functioning should be made by special software automatically. In this paper we describe the tight connection between complexity of modern large high performance computing systems and special techniques and tools required to ensure their efficiency in practice. The main subsystems of the developed complex (Octoshell, DiMMoN, Octotron, JobDigest, and an expert software system to bring fine analytics on parallel applications and the entire supercomputer to users and sysadmins) are actively operated on the large supercomputer systems at Lomonosov Moscow State University. A brief description of the architecture of Lomonosov-2 supercomputer is presented, and questions showing both a wide variety of emerging complex issues and the need for an integrated approach to solving the problem of effectively supporting large supercomputer systems are discussed.

AlgoWiki: an Open Encyclopedia of Parallel Algorithmic Features

Abstract: The main goal of this project is to formalize the mapping of algorithms onto the architecture of parallel computing systems. The basic idea is that features of algorithms are independent of any computing system. A detailed description of a given algorithm with a special emphasis on its parallel properties is made once, and after that it can be used repeatedly for various implementations of the algorithm on different computing platforms. Machine-dependent, part of this work is devoted to describing features of algorithms implementation for different parallel architectures. The proposed description of algorithms includes many non-trivial features such as: parallel algorithm complexity, resource of parallelism and its properties, features of the informational graph, computational cost of algorithms, data locality analysis as well as analysis of scalability potential, and many others. Descriptions of algorithms form the basis of AlgoWiki, which allows for collaboration with the computing community in order to produce different implementations and achieve improvement. Project website: http://algowiki-project.org/en/

A Study of the Dynamic Characteristics of Software Implementation as an Essential Part for a Universal Description of Algorithm Properties

Abstract: The AlgoWiki open encyclopedia of parallel algorithmic features enables the entire computing community to work together to describe the properties of a multitude of mathematical algorithms and their implementation for various software and hardware platforms. As part of the AlgoWiki project, a structure has been suggested for providing universal descriptions of algorithm properties. Along with the first part of the description, dedicated to machine-independent properties of the algorithms, it is extremely important to study and describe the dynamic characteristics of their software implementation. By studying fundamental algorithm properties such as execution time, performance, data locality, efficiency and scalability, we can give some estimate of the potential implementation quality for a given algorithm on a specific computer and lay the foundation for comparative analysis of various computing platforms with regards to the algorithms presented in AlgoWiki.

JobDigest – Detailed System Monitoring-Based Supercomputer Application Behavior Analysis

Abstract: The efficiency of computing resources utilization by user applications can be analyzed in various ways. The JobDigest approach based on system monitoring was developed in Moscow State University and is currently used in everyday practice of the largest Russian supercomputing center of Moscow State University. The approach features application behavior analysis for every job run on HPC system providing: the set of dynamic application characteristics - time series of values representing utilization of CPU, memory, network, storage, etc. with diagrams and heat maps; the integral characteristics representing average utilization rates; job tagging and categorization with means of informing system administrators and managers on suspicious or abnormal applications. The paper describes the approach principles and workflow, it also demonstrates JobDigest use cases and positioning of the proposed techniques in the set of tools and methods that are used in the MSU HPC Center to ensure its 24/7 efficient and productive functioning.

An Approach for Ensuring Reliable Functioning of a Supercomputer Based on a Formal Model

Abstract: In this article we describe the Octotron project intended to ensure reliability and sustainability of a supercomputer. Octotron is based on a formal model of computing system that describes system components and their interconnections in graph form. The model determines relations between data describing current supercomputer state (monitoring data) under which all components are functioning properly. Relations are given in form of rules, with the input of real monitoring data. If these relations are violated, Octotron registers the presence of abnormal situation and performs one of the predefined actions: notification of system administrators, logging, disabling or restarting faulty hardware or software components, etc. This paper describes the general structure of the model, augmented with details of its realization and evaluation at supercomputing center in Moscow State University.

Supercomputer Lomonosov-2: Large Scale, Deep Monitoring and Fine Analytics for the User Community

Abstract: The huge number of hardware and software components, together with a large number of parameters affecting the performance of each parallel application, makes ensuring the efficiency of a large scale supercomputer extremely difficult. In this situation, all basic parameters of the supercomputer should be constantly monitored, as well as many decisions about its functioning should be made by special software automatically. In this paper we describe the tight connection between complexity of modern large high performance computing systems and special techniques and tools required to ensure their efficiency in practice. The main subsystems of the developed complex (Octoshell, DiMMoN, Octotron, JobDigest, and an expert software system to bring fine analytics on parallel applications and the entire supercomputer to users and sysadmins) are actively operated on the large supercomputer systems at Lomonosov Moscow State University. A brief description of the architecture of Lomonosov-2 supercomputer is presented, and questions showing both a wide variety of emerging complex issues and the need for an integrated approach to solving the problem of effectively supporting large supercomputer systems are discussed.

The Popper Convention: Making Reproducible Systems Evaluation Practical

Abstract: Independent validation of experimental results in the field of systems research is a challenging task, mainly due to differences in software and hardware in computational environments. Recreating an environment that resembles the original is difficult and time-consuming. In this paper we introduce _Popper_, a convention based on a set of modern open source software (OSS) development principles for generating reproducible scientific publications. Concretely, we make the case for treating an article as an OSS project following a DevOps approach and applying software engineering best-practices to manage its associated artifacts and maintain the reproducibility of its findings. Popper leverages existing cloud-computing infrastructure and DevOps tools to produce academic articles that are easy to validate and extend. We present a use case that illustrates the usefulness of this approach. We show how, by following the _Popper_ convention, reviewers and researchers can quickly get to the point of getting results without relying on the original author's intervention.

Histone dynamics mediate DNA unwrapping and sliding in nucleosomes

Abstract: Nucleosomes are elementary building blocks of chromatin in eukaryotes. They tightly wrap ∼147 DNA base pairs around an octamer of histone proteins. How nucleosome structural dynamics affect genome functioning is not completely clear. Here we report all-atom molecular dynamics simulations of nucleosome core particles at a timescale of 15 microseconds. At this timescale, functional modes of nucleosome dynamics such as spontaneous nucleosomal DNA breathing, unwrapping, twisting, and sliding were observed. We identified atomistic mechanisms of these processes by analyzing the accompanying structural rearrangements of the histone octamer and histone-DNA contacts. Octamer dynamics and plasticity were found to enable DNA unwrapping and sliding. Through multi-scale modeling, we showed that nucleosomal DNA dynamics contribute to significant conformational variability of the chromatin fiber at the supranucleosomal level. Our study further supports mechanistic coupling between fine details of histone dynamics and chromatin functioning, provides a framework for understanding the effects of various chromatin modifications.

Parallel Efficient Sparse Matrix-Matrix Multiplication on Multicore Platforms.

Abstract: Sparse matrix-matrix multiplication (SpGEMM) is a key kernel in many applications in High Performance Computing such as algebraic multigrid solvers and graph analytics. Optimizing SpGEMM on modern processors is challenging due to random data accesses, poor data locality and load imbalance during computation. In this work, we investigate different partitioning techniques, cache optimizations (using dense arrays instead of hash tables), and dynamic load balancing on SpGEMM using a diverse set of real-world and synthetic datasets. We demonstrate that our implementation outperforms the state-of-the-art using Intel\(^{{\textregistered }}\) Xeon\(^{{\textregistered }}\) processors. We are up to 3.8X faster than Intel\(^{{\textregistered }}\) Math Kernel Library (MKL) and up to 257X faster than CombBLAS. We also outperform the best published GPU implementation of SpGEMM on nVidia GTX Titan and on AMD Radeon HD 7970 by up to 7.3X and 4.5X, respectively on their published datasets. We demonstrate good multi-core scalability (geomean speedup of 18.2X using 28 threads) as compared to MKL which gets 7.5X scaling on 28 threads.