Fine-Grained Per-Outlet Hourly Carbon Emission Assessment in Data Centers via Intelligent PDUs Employing Data-Driven Clustering

Erdogan D., Erdogan A. T., Gokce V. B., Aksoy F. C., Aksoy B. F., ZEHİR M. A.

5th International Conference on Informatics and Software Engineering, IISEC 2026, Ankara, Turkey, 5 - 06 February 2026, pp.561-565, (Full Text) identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/iisec69317.2026.11418450
  • City: Ankara
  • Country: Turkey
  • Page Numbers: pp.561-565
  • Keywords: carbon emission, data center, energy consumption, grid mix, pdu
  • Marmara University Affiliated: Yes

Abstract

The share of data center electricity consumption in national energy demand, along with the associated carbon emissions, has been increasing significantly in several leading countries. Beyond monthly or annual emission calculations using yearly average emission rates at the data center facility scale, and even beyond rack-level calculations, fine-grained power distribution unit per-outlet level assessment with hourly dynamic grid mix can enable user-, server-, application- or virtual machine (VM)-specific detailed tracking of carbon emissions, supporting effective emission management. This study investigates a year-long PDU per-outlet level hourly emission assessment using real data center and grid mix data, employing unsupervised data-driven clustering. The results of the case study provide valuable insights into the enhanced visibility and evaluation capabilities enabled by the proposed methodology.