The NIIF Cloud infrastructure currently operates on 31 dual-processor Fujitsu “Primergy RX100S6” machines. The manufacturer claims a working current draw of 300W for this machine, and NIIF’s own measurements reflect this claim. This low power is further distributed by using 37 Virtual Machines, machines which previously ran on physical machines.
As a baseline, the NIIF Cloud infrastructure team got to work calculating the annual power consumption of the previous machines:
37 machines x 0.3 kW per machine x 24 hrs/day x 365 days/year x 0.17 EUR/kWh = approximately €16,565.- per year
The true cost is actually more than this, as the 37 previous machines were not as energy efficient as the new ones, but no measurements were made to verify that, so the new machines’ rating was used.
Then the NIIF Cloud Infrastructure Team set up the VM Scheduler to load the physical machines with Virtual Machines which were already in use as much as possible. Of course, this was only allowed if was not to the detriment of the quality of the services, for example, making sure that any CPU intensive jobs had sufficient resources to run comfortably.
During the course of this rescheduling exercise, and after a period of careful monitoring, it became apparent that the 37 virtual machines could easily complete all their tasks on just 10 physical machines, which enabled the remaining 21 physical machines to be switched off, and indeed were unplugged.
The running cost of these 10 machines was then calculated:
10 machines x 0.3 kW per machine x 24 hrs/day x 365 days/year x 0.17 EUR/kWh = approximately €4,477.- per year
So, when they calculated the difference, this would bring a saving of approximately €12,088 annually, on the electricity costs alone, which represents a 73% cost reduction!
This was an experiment to test the conversion of just 37 physical machines in one location.
The potential cost savings, not only in electricity consumption and the resultant environmental impact, but potential for reduction in capital expenditure on physical machines, is huge if all of the thousands of cloud machines in the Universities, Colleges and Research Centers across Europe could be pooled in an IaaS cloud.
This blog post was written by Jim Buddin of TERENA, on behalf of Stefán Péter (NIIF, Hungary)
NIIF (The Hungarian NREN) : http://www.niif.hu/en