Virtu Systems conducted load testing of a new segment of CommCloud metrocluster

CommCloud metrocluster is designed to ensure the smooth operation of the IT infrastructure for large and medium-sized businesses. The purpose of the testing was to identify potential problems in the operation of the server hardware, to find out the weak points in the work of the CPU and RAM and to eliminate them.

The CommCloud team was testing the LINPACK library for 72 hours. During the test, server performance was close to the maximum. Processors and RAM were used by at least 90%. All servers and running virtual machines worked stably without any crashes and hangs, and the tested hardware elements, processors and memory showed no faults.

The VMware ESXi hypervisor is installed and the servers are clustered. We created 20 virtual machines to provide cluster servers with the full load. We used multiple virtual machines to accurately distribute the load to server elements and conduct testing in according to productive environments. Virtual machines were created with the following parameters: OS Windows Server 2016 Standard, CPU – 4 cores for 3GHz, RAM – 37 GB, HDD – 150 GB. The LINPACK test, which library is designed to solve a random system of linear equations using the Gauss method on multiprocessor systems, is used to form the TOP500 list (the rating of the five hundred most powerful computing systems in the world). We ran this test on each virtual machine in two-three-way mode with 32GB of RAM. The size of the matrix of the system of linear equations in 65486 was chosen based on the amount of RAM of the allocated virtual machine. The calculations were made using the formula 2n3 / 3 + 2n2 (n is the size of the matrix).

The enterprise solution is built on Lenovo components. “The results of the test confirmed the reliability of Lenovo equipment, as well as the quality of assembly and transportation,” says Ilya Popov, director of IT infrastructure at Virtu Systems.

The next step is to conduct DR-testing and performance tests on typical configurations of information systems.