Optimizing Data Centre Energy and Environmental Costs
Bibliographic record
Abstract
Data centres use an estimated 2% of US electrical power which accounts for much of their total cost of ownership. This consumption continues to grow, further straining power grids attempting to integrate more renewable energy. This dissertation focuses on assessing and reducing data centre environmental and financial costs. Emissions of projects undertaken to lower the data centre environmental footprints can be assessed and the emission reduction projects compared using an ISO-14064-2-compliant greenhouse gas reduction protocol outlined herein. I was closely involved with the development of the protocol. Full lifecycle analysis and verifying that projects exceed business-as-usual expectations are addressed, and a test project is described. Consuming power when it is low cost or when renewable energy is available can be used to reduce the financial and environmental costs of computing. Adaptation based on the power price showed 10-50% potential savings in typical cases, and local renewable energy use could be increased by 10-80%. Allowing a fraction of high-priority tasks to proceed unimpeded still allows significant savings. Power grid operators use mechanisms called ancillary services to address variation and system failures, paying organizations to alter power consumption on request. By bidding to offer these services, data centres may be able to lower their energy costs while reducing their environmental impact. If providing contingency reserves which require only infrequent action, savings of up to 12% were seen in simulations. Greater power cost savings are possible for those ceding more control to the power grid operator. Coordinating multiple data centres adds overhead, and altering at which data centre requests are processed based on changes in the financial or environmental costs of power is likely to increase this overhead. Tests of virtual machine migrations showed that in some cases there was no visible increase in power use while in others power use rose by 20-30W. Estimates of how migration was likely to impact other services used in current cloud environments were derived.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".