Data centers (DC) are digital warehouses for information technology (IT) equipment such as computer servers, network switches, and storage devices. To maintain their computational reliability, waste heat from the DC facility is removed by dedicated cooling hardware such as computer room air conditioning units and rear door heat exchangers. Because these cooling hardware systems demand 40-50% of the data center electricity consumption (often in the order of few million kWh), their energy-efficient operation is imperative for the optimization of total cost of ownership for a DC. This team has developed a real-time software application that will recommend the optimal cooling set-points for different cooling hardware in a data center.
The proposed software (DataCOOL) will deliver real-time end-to-end cooling resource optimization software for data centers. The critical value proposition of DataCOOL lies in its data center cooling cost optimization, derived from elastic resource allocation and risk mitigation from stochastic data center traffic. These values are derived from a model-order reduction algorithm. The proposed technology could potentially save 15-25% of a data center's annual operational expenditure. An extensive literature survey indicates that the proposed methodology is substantially different from the existing approaches. The primary target customers would include large IT companies with data centers such as: Cisco, Microsoft, Google, Amazon, and Facebook. On the other hand, the team also expects to work closely with mid-size data center companies like Equinix, Internap, and QTS to validate that the proposed prototype is delivering claimed cost-saving benefits.
Data centers are the epicenter of the contemporary digital economy. However, ever-increasing energy consumption in data centers--as high as 2% of the world's electricity usage--is posing a serious threat to the sustainability of data center operations. In July 2015, under I-CORPS support, Team DataCOOL Solutions started with a cooling energy optimization technology for data centers. When we started the I-Corps program, we were uncertain about the customer segments that would value our technology. Figure 1 shows the initial version of the business model canvas, indicating the untested hypotheses. We conducted 112 interviews in total over a six week period and tried to understand our value propositions, customer segments, and the product-market fit. Figure 2 shows the final business model canvass based on the lessons learned in the program. As marked in the blue callouts in Figure 2, the key lessons learned in the program are: Additional value propositions: It was validated that cooling cost saving, without increasing the risk of service interruption is indeed valuable. Furthermore, we discovered an additional value proposition in terms of the capacity improvement for data center cooling systems. Our proposed technology also improves the data center cooling capacity. However, we did not realize that it could be a significant value. This is a major insight gained in the I-CORPS program. Clarity on customer segments and channel: As we started the program, we thought we would be directly selling to the data centers. However, soon we realized that the data center industry has a far more complex ecosystem (shown in Figure 3), including multiple players such as data center design firms. We also identified the multi-sided nature of the potential corporate sales process. While a typical data center chief information officer will be the payer for our technology, data center facility operators will be the users. The IT managers of the data center will act as the influencer. We need to be especially careful about the data center facility managers who could be the potential saboteurs. Revenue Model: We explored the standard purchase practice in data center cooling industry. The potential revenue streams are identified to be License Fee, Maintenance Fee, Hardware, and Implementation. Potential Partners: We identified and initiated conversations with the potential channel partners: Lawrence Berkeley National Laboratory for testing, Intel for distribution, and Georgia Tech Research Corp. for IP Licensing. Big Cloud Computing Companies as Customers: We initially thought that our largest customers would be large cloud service providers such as Google. However, after interviewing with them, we realized that most of them design data center as a system, and may not be very open to working with startups. At the end of the program, we decide to form a startup. Last October, a corporation—AdeptDC co. (www.adeptdc.com)—was set up. We have been awarded Georgia Research Alliance (GRA) Phase I grant to develop a minimal viable product (MVP). We have also applied for NSF SBIR award. The immediate future plan includes following activities: MVP development Customer validation Early-stage customer creation Utility patent filing
