Assignment 3

Problem Statements:

• Mass use of AI and advanced computer technology consumes extreme amounts of water. 

• LLM computing requires large amounts of power that results in a need to be cooled, causing mass water consumption.

Restatement: 

Real Problem: Mass power consumption from LLM and advanced computing causes overheating in their devices. This causes a need to be cooled that results in an extreme use of water. 

 Stated Boundaries: In some solutions, current technology may not be able to produce effective solutions within a reasonable price range. However, future technology may be more effective. 

Actual Boundaries: Money, Time

Initial Goals: Increase efficiency of power usage to reduce water consumption to a certain degree.

Meaningful Goals: Set a new standard for power efficiency. 

Inputs: Efficient algorithm, computer hardware. 

Outputs: Efficiency of the system.

Present State-Desired State (PS-DS) Strategy:

• Present State: Current LLM and advanced computing systems consume extreme amounts of water

• Desired State: A reduction of power consumption so the water output is reduced.

Kepner-Tregoe (KT) situation analysis: 

Timing: 

• Urgency: AI usage continues to grow exponentially meaning the consumption of freshwater can only be expected to worsen

• Deadlines: Many AI companies have been forced to implement carbon-neutral policies and have set deadlines for when these rules shall be in place (5 years, 10 years, etc.)

                Trend:

• Increasing: Demand for AI usage in data centers is on a rapid rise 

• View On Sustainability: Public scrutiny surrounding resource usage is intensifying 

• Technology: Advances in alternative and more efficient cooling systems are emerging but at a relatively slow rate

Impact:

• Resource Scarcity: Excessive water usage might lead to conflicts with local communities and worsen drought conditions in certain areas

• Operational Cost: Decreased water availability combined with increasing demand and need for wastewater treatment is driving up operational cost

• Reputation: Companies run the risk of ruining their reputation in the eyes of the public if they continue to be associated with unsustainable water consumption

Order of Actions

1. Gathering Data: 

• Map current water usage at data centers

• Identify future levels of water availability at these locations

2. Evaluate Options:

• Conduct studies on alternative cooling

• Asses ROI and compatibility with current systems

         (C) Prototyping:

• Implement possible solutions at a small scale to gather data 

• Measure reduction in water consumption and monitor reliability 

         (D) Rollout:

• Begin implementing the best-performing solutions

• Develop standards for these new water-efficient cooling systems 

          (E) Monitoring and compliance:

• Continue to monitor water and energy consumption 

• Adjust the systems based on performance feedback and possible future regulations

What? :

• Isn’t Known: Efficient ways to cool computer systems without consuming large amounts of water via  water cooling

• Is Observed:  Extreme water usage in data centers due to water cooling in PC’s.

• Constraints: CPU/GPU heat in PC’s, Storage Space in Data centers, Cooling System Setups.

• Is Important: Immediate lowering of water consumption while staying within a certain margin of efficiency. Believe that environmental costs are top priority even if some amount of cooling efficiency is lowered.

• Goals/Expectation: To find an efficient cooling system method with a similar margin of efficiency while drastically reducing water consumption.

When?

• Problem Occurrence- Within the last decade due to LLM’s rising in technology prevalence and AI advancement. 

• Solution Implementation - As soon as possible, while global warming concerns related to fossil fuels and carbon emissions may be more urgent, water consumption is very dangerous for the environment and should not be put to the side. As the demand for LLM’s begins to grow, water will be exponentially consumed to follow these demands.

• Change Occurrence - There is research regarding alternate ways to cool computing systems, but most fall short in efficiency.

Who?

• Providing More Information - Power, Electrical, and Thermal Engineers will be analyzing and providing information regarding thermal and water consumption.

• Customers - Data Companies, Computer Part Companies, AI Companies

• Providing Current Information - Environmental Specialists

• Affected by a Problem? - Water consumption affects everyone on the planet that uses water, efficiency reduction due to environmental concerns will affect the data companies mainly.

Where?

• Problem Occurrence - Data Centers, specifically the PC”s cooling systems.

• Equipment Location - Large Data centers that are housed for training LLM”s at an efficient rate.

Why?

• Importance - Global warming and environmental concerns are constantly being overlooked in lieu of advancing data. It has been shown in the past that ignoring these concerns causes major problems. With the quickly advancing rate of AI - technology and the training of LLM’s, water will be an extreme concern.

• Solution Works? - Developing a new cooling system that reduces water consumption while staying efficient will be revolutionary for environmental concerns and keeping our planet’s resources in check.