INFORMATION ON UTILIZATION OF DATA CENTER RESOURCES WITH MESSAGE BROKER IMPLEMENTATION

• Main Authors: Samokhin N.Yu., Khoruzhnikov S.E., Trubnikova V.M
• Format: Article
• Language: English
• Published: Saint Petersburg National Research University of Information Technologies, Mechanics and Optics (ITMO University) 2018-05-01
• Series: Naučno-tehničeskij Vestnik Informacionnyh Tehnologij, Mehaniki i Optiki
• Subjects: data center; virtualization; monitoring; billing; message broker; Zabbix; RabbitMQ
• Online Access: https://ntv.ifmo.ru/file/article/18150.pdf

A cloud of geographically distributed data centers developed by ITMO University (ifmo.ru) is presented.To increase reliability, as well as to collect statistical data, a special monitoring system is created that monitors the state of systems and the utilization of hardware resources. In addition to building a map of statistical data, information on resource utilization is suggested for the provider to perform a billing. An approach to monitoring system with asynchronous communication is proposed. An option with message broker implementation is shown. Out of two basic billing models, billing by quotas and billing by actual usage, we propose a model for organizing a dialogue between the administrator and the monitoring system in order to obtain data on resource utilization by actual usage. The use of a database for storing requests and responses is proposed, as well as the asynchronous dialogue using queues technology and message broker. Approbation was carried out on testing equipment with an actual infrastructure simulation. Unique asynchronous multi-agent system was obtained for statistics gathering on data center resources utilization. RabbitMQ message broker, PostgreSQL database and Zabbix monitoring system implementations are shown. The dialogue system was optimized with the help of Python code. The designed model can be recommended for implementation aimed at utilization of data obtained in different data centers, including distributed ones. Suggested model can be used in high-loaded systems providing virtual resources as a service.