Wednesday, December 16, 2020
Online seminar
Nechaevskiy A.V.

Methods and tools for simulating distributed data storage and processing systems based on the monitoring results

(Materials of a PhD thesis)
Modern data storage and processing systems represent composite distributed software and hardware complexes built using grid and cloud technologies. Simulation modeling allows one to detect bottlenecks in the architecture of data centers, to conduct experiments with changing the topology and replacing resources to check the proposed system solutions without directly interfering with the functioning of working setups and to test algorithms for task management and resource allocation among user groups. At present, the processes of simulation and monitoring are considered as independent tasks, not related to each other. To enhance the accuracy of the results, it is required to use the statistics accumulated during work using monitoring systems as input data for simulation. This entails the development of software tools that combine the processes of simulation and monitoring. The presented software complex SyMSim for simulating data storage and processing systems implements the synthesis of modeling and monitoring and allows one to carry out preliminary studies of different options for organizing the IT infrastructure, to evaluate the capacities of the existing architecture in solving data storage and processing tasks and to give recommendations for its optimization. SyMSim was used to simulate the computing centers of the BM@N and MPD experiments at the NICA accelerator complex, MPI calculations and the IHEP computing center in Beijing.
The information about seminar and the link to connect via Webex are available at Indico.