Power Consumption
- 1970 - 1990: “IPAT formula” by Ehrlich & Ehrlich considered interactions between technology and ecology:
Ecological Impact = Population x Affluence x Technology
- 1990 - today: J. Huber extended the formula. Consensus today:
Ecological Impact = Population x Efficiency x Consistency x Sufficiency
Efficiency:
- Good/ Resource consumption
- High output (more computation), small input (say less electrical energy)
- How to optimize resource consumption of technology?
- today’s economical model
Consistency:
- Nature pollution/ Resource consumption
- How much of our 1 unit of consumption, polute the environment?
- How can technology be non-pollutive resp. climate positive?
- circular economy: existing goods are shared
Sufficiency:
- Good/ capita
- How much technology do we need?
- degrowth/ post-growth economy
| Efficiency | Consistency (circular economy) | Sufficiency (degrowth) | |
|---|---|---|---|
| Solutions | More computing performance, virtualisation | Reuse software, maintain open source software; Share resources; Use renewable energy sources in data centers | Use less; don’t use if its not necessary |
| Barriers | Rebound effect & Wirth’s law | Proprietary business models | Low acceptance rate |
- Energy = Power x Time
- Strong correlation between execution time and energy efficiency, but it is not the same
- Reasons: I/O, memory access, CPU idle,..
Technology Choice
- Programming language, frameworks, libraries
- Compiled languages tend to be faster than interpreted ones
- Common trade-off: efficiency vs. developer productivity
- Choose according your demand
- Consider:
- Use case
- Team skills
- Non-functional requirements, e.g. expected system load
- Project size
Energy Effciency
-
For data ceners, e.g.: PUE = Total Facility Energy /IT Equipment Energy
- Energy[J] = Power[W] x Time[s]
- W = J/s
- Energy Eff. = # computations/ Joule
| Hardware | Software | Analytically |
|---|---|---|
| Using hardware power meters and specialized development boards | Using software-defined power meters, meterics collection hooks inside programs and different heuristics | For instance through breaking down programs into individual CPU instructions |