Benchmarking board SCOPE at Freiburg location

The (experienced) performance of an own VM in a cloud always depends on others´ activities. Therefore, not only the surveys of the general capacitiy parameter on the system level, which are mainly done by the administrators, should be considered, but also the performance from the user´s perspective. Looking at these measurements also delivers information regarding potential current problems at one of the Cloud locations.

In order to give users insight into the bwCloud SCOPE performance, there is now a new board that delivers a detailed overview of bwCloud workload spikes , from the perspective of a benchmarking test. This board is available at https://mm01.ka.bw-cloud.org/d/DFvnCbWZk/benchmarking?orgId and includes various overviews that are generated with the help of different benchmarking tests.

Alternatively, there is also the Kiosk dashboard https://mm01.ka.bw-cloud.org/d/v0kxl5rmz/kiosk?orgId. This one offers an overview of all regions. For the new benchmarking board, graphs are generated mainly from the results of two tools: sysbench and fio. That works as follows: The OpenSource Agent for metrics from system and sensor data, Telegraf, runs a script every hour. This script creates a VM based on an image designed for benchmarking. This VM then runs tests and then delivers the results backs to Telegraf. From there, the data is saved in a database, from where Grafana takes them for the display of the graphs. The benchmarking image was created with Ansible Scripts and can therefore create reproducible and comparable images/results at any time. It is therefore possible to dedeuct how "fast" the VMs currently are from the graphs that the benchmarking script creates. The measured data are: - Duration of Measurement. - fio (file operations) latency: here, the maximal, minimal and their average and total IOPs are saved. - sysbench Memory/RAM: here also max, min, avg and sum can be called up, additionally the amount of events as well as the total test time. - sysbench CPU: here also max, min, avg and sum can be called up, additionally events per sec, as well as the 95 percentile. - sysbench thread latency: here also max, min, avg and sum can be called up, additionally the total test time. This allows graphs to show a short and simplified overview of the system workload from the perspective of a VM. They also offer insight into the effects of local and cloud-wide events.