R718.com: Could you please tell us a little about the iDataCool system installed in the Regensburg University computing center, and the basic function of the adsorption chiller?
Tilo Wettig: iDataCool is a large computer cluster jointly developed by the University of Regensburg and the IBM Research and Development Lab Böblingen. It is based on IBM's iDataPlex platform, whose air-cooling solution was replaced by a custom water-cooling solution that allows for computer cooling with hot water (70°C/158°F), thanks to an InvenSor adsorption chiller that operates efficiently at these temperatures. A significant portion of the heat energy generated by the computer can be recovered in the form of chilled water, which can then be used to cool other machines in the computing center.
The basic function of the adsorption chiller is to convert the waste heat of the computer (i.e., hot water) into chilled water that can be used to cool other computers.
R718.com: What is the principal innovation in this project?
Tilo Wettig: The main innovation is the use of the waste heat of a large computer to drive an adsorption chiller. This leads to significant energy and cost savings.
R718.com: How long has the system been in operation and how do you evaluate its performance?
Tilo Wettig: The system has been in operation for over 1.5 years now. It performs as expected, and we have not observed any major problems. The above-mentioned paper provides details on performance benchmarks.
R718.com: Compared to the previous air cooling system for the computer centre in what are the major benefits of the water cooling solution in your project? What problems were encountered in the iDataCool system?
Tilo Wettig: We save on energy (and therefore on costs) that would otherwise have been spent on air cooling. In addition, we reuse some of the energy that was fed into the computer, which leads to further energy and cost savings. Another advantage is that the compute density of a water-cooled system can be much higher than that of an air-cooled system.
Our main problem is the imperfect thermal insulation of the (pre-existing) racks. This means that not all heat generated by the computer is transferred to the water, but that we lose some heat to the air of the computing center. This effect becomes worse at higher temperatures. However, this problem can be fixed in future rack designs by better thermal insulation.
R718.com: According to your information, are there other data centre projects using or that plan to use adsorption chillers in Europe?
Tilo Wettig: The only other project I know of is a smaller-scale project called CoolMUC at the LRZ Garching.
R718.com: What do you think are the key markets for, and technological barriers to, the widespread adoption of adsorption chillers in data centre cooling?
Tilo Wettig: The obvious market for adsorption chillers in data center cooling is in hot climates where free cooling is difficult most of the year and where chilled water is always needed. As for the technology, it would be helpful if the future development of adsorption chillers could target two issues:
a) higher COP already at lower driving temperatures
b) smaller units with higher cooling capacity
R718.com: How do you evaluate the market potential of adsorption chillers in data centre cooling in the future?
Tilo Wettig: This is hard for me since I'm not an expert in economics. Certainly the market potential would greatly improve if the two issues mentioned above could be improved upon.