CCAC Computer Power Management Saves Significant Energy Use and Costs
Community College of Allegheny County (CCAC) saw a lot of potential for energy savings in their campus computer network. The annual cost of operating almost 4,000 personal computers plus a data center was about $240,000. The college implemented a network-wide power management system, with a focus on computer laboratories, automatically implementing sleep settings for all computers sitting unused for a given length of time.
The energy savings were immediate, dramatic, and persistent. Total kWh consumption by all campus computers plummeted 74% following implementation, translating to $60,000-$75,000 saved annually.
In the past, CCAC’s IT department disabled sleep mode and hibernation on campus desktops at the BIOS, or hardware level, so that neither Windows nor any other software could tell the computers to sleep. This only allowed them to be on or off, and the display was able to turn off automatically. One reason for this approach was to avoid “blue screen of death” (BSOD) crashes that affected some older computers when woken from sleep mode. Another reason was that it was believed that leaving computers on was the best way to patch systems more reliably. However, BSOD are not very common with modern computers, and CCAC IT revised their patching strategy to queue patches when machines are turned on, and only install during maintenance periods to avoid peak use times by users. Given the removal of these barriers, CCAC IT decided there was great opportunity to pursue a network-wide power management system.
To begin changing sleep settings, the first step was to revert the disabled sleep and hibernation settings without requiring on-the-ground staff to physically do so at each computer. Since most of the computers are Dell branded, CCAC IT were able to target those with a BIOS configuration package created from a utility called Dell Command Configure. IT used a reference computer to take a snapshot of its BIOS settings using Command Configure, and then applied it to all other computers remotely. Another built-in report from Configuration Manager was able to identify computers that were missed by remote changes. On-the-ground technicians were then able to physically make changes to this much smaller group of computers. The decision to remotely deploy settings to the majority of campus computers greatly increased savings.
After the hardware was made to be able to sleep, CCAC IT utilized Group Policy to deploy Windows power plans rather than purchase additional software to manage power options. Group Policy is a feature for managing user and computer settings that any business with a Microsoft domain already has access. The power plan options are the same options one can adjust on their own Windows computer – Group Policy allowed them to create and force power plans for all computers. With the existing Active Directory structure, CCAC IT already had the different types of computers and use cases separated into organizational units that made sense for the different power plans they would need to deploy in different areas.
It was important for the college community to be aware of the coming changes and the goals in mind. CCAC communicated the change to employees via email and an article in the IT departmental newsletter. For a few months after settings were implemented, there was a noted increase in related help requests from users, but eventually the settings gained acceptance as it became the norm for everyone. Feedback was also used to adjust settings for different use cases.
Energy consumption across all computers dropped from 120,908 kWh/month to 31,708 kWh/month, a 74% savings. On average, each computer is now consuming 9 kWh/month from 33 kWh/month.
CCAC was already using Microsoft’s Configuration Manager tool to gather power metrics directly from computers as a function of its built-in reporting capabilities. CCAC IT were able to use this tool to continue to monitor power consumption as power settings were changed.
The graphs below come from Configuration Manager reports. Figures 1 and 2 represent the daily kWh consumption of all computers before and after power management system implementation. Previously, there was a small dip in consumption on weekends, a high of about 4,500 kWh and a low of about 3,800 kWh. After all new settings were implemented, daily changes were much more dramatic, with a new high of about 1,250 kWh and a low of about 450 kWh (note the change in y-axis scale).
Figures 3 and 4 represent computer activity on one model of computer monitored particularly closely. The blue line in both figures is the number of computers that are powered on at a particular time of day. The red line is the number of computers actively being used. Previously, at any given time of day, there were always at least 900 computers of this model turned on, but not being used. Ideally, the gap between the blue and red line should be tighter together and nearly parallel – a computer should not be on if it is not being used. This gap was narrowed substantially by the power management system.