vac:dst:ift:2026q1-dst-lab
Description
DST lab is a set of machines that are mainly used in a composed Kubernetes cluster. In this environment, we can do a set of experiments to test regressions, new functionalities and features, or experimental changes on any decentralized system. In these experiments, we look for abnormal behaviors, we do measurements to study the performance and robustness of the system. These results can be also used to compare with theoretical results and analyze the scaling behaviors of the systems. At the same time, we will work on improving the lab’s resources and time allocation, while allow provisioning to other teams dedicated machines if they are needed.
Task list
Analyze current stack
- fully qualified name:
vac:dst:ift:2026q1-dst-lab:analyze-current-stack - owner: Mamoutou
- status: not started
- start-date: 2026/01/01
- end-date: 2026/03/31
Description
Compare the current stack with new stack proposals. Study if changes are helpful and improve overall performance of the lab. Report findings with analysis and benchmarks of the results.
Deliverables
- Code:
- Reports:
Optimize data scrapping
- fully qualified name:
vac:dst:ift:2026q1-dst-lab:optimize-data-scrapping - owner: Mamoutou
- status: not started
- start-date: 2026/01/01
- end-date: 2026/03/31
Description
Improve the monitoring system of the lab. Adjust scraping frequency and selected metrics. Design a storage solution to store the data so it can be retrieved if necessary. Optimize the storage to reduce used space if data is older than a certain period.
Deliverables
- Code:
- Reports:
Lab health monitoring
- fully qualified name:
vac:dst:ift:2026q1-dst-lab:lab-health-monitoring - owner: Mamoutou
- status: not started
- start-date: 2026/01/01
- end-date: 2026/03/31
Description
Design a set of metrics/dashboards that can be used to monitor the health of the lab. The metrics should be useful to detect abnormal behaviors and to detect potential issues. What we want to achieve with this is to be sure that we can compare experiments from one week to a different week being confident that the results can be trusted, without the need of repeat the same experiment again.
Deliverables
- Code:
- Reports: