Energy-conscious computing

Finding ways to save energy incrementally in established IT environments might seem less effective than a more radical re-engineering of the data center. The IBM optimization approach proposes to aim at the heavy use of virtualization, decommissioning of old hardware, migration and consolidation of servers and applications, enhancement of cooling and temperature management, and the reuse of waste heat, etc. to accomplish a holistic IT and facility management effect. But owing to soaring energy costs, bottom-line savings in the respective areas are becoming increasingly attractive. At current growth rates of energy consumption, a customer in 20 years will need 16 times the energy capacity of today's data centers. Furthermore, savings on the IT side go hand in hand with potential savings on the facility side (CRAC, UPS, PDU, etc.).

Energy costs amount to approximately 40% of the total operational cost of a data center. In any case, the sum of small energy savings realized by increased transparency and awareness of granular and rolled-up resource usage can have a significant impact by allowing financial investments to be delayed that would otherwise be needed earlier to scale up capacity. To this extent, the GECO project aims at supporting proactive data-center energy management in the context of IBM's Green IT initiative. Specifically, our goals are to enable detailed insight into existing data-center energy consumption by moving the scope of energy monitoring beyond already electricity-meter-enabled devices, to enable refined analysis and charge-back by mapping energy consumption to business entities and to move the focus from single energy consumers to the interplay of the interdependent business components involved in the IT service provisioning. The overall goal is to establish energy requirements as a central criterion in planning and decision processes.

1. Thorsten Hake, "Relating energy consumption of IT systems to their business usage," Diploma thesis at ZRL, January to August 2008.
2. Andreas Kohler, "Granular power consumption allocation to organizational entities," Student project, August to November 2008.