One of the unexpected benefits of server refresh is potential energy rebates from local organizations or utilities looking to encourage IT Sustainability. Understanding the financial benefits of a four-year server refresh from a ROI/TCO perspective is relatively easy, however balancing tight schedules and deliverables in a data center can be a challenge.
Several of my colleagues, Tom, Dave and Mary, from our Computing Solutions and Services Team wanted to find a way to keep their internal customers (finance and management inside Intel IT) focused to stay on course to refresh servers on a four year cycle. They were fortunate to find a local non-profit organization, chartered with helping utility ratepayers use less energy. It was a match!
So..how did the relationship develop? I chatted with Tom, Dave and Mary to find out the ins and outs. Like any good relationship, it took time and effort.
Part I of 2: I interviewed Tom, Manager of the Computing Solutions and Services team within IT’s Engineering Computing organization and David, System Administrator and Data Center Planner within the computing Solutions and Services Team. Tom and his team manage the Linux servers used by Intel’s design engineers in northwest America and David, owns server landings and data center space allocation.
In Part 1 we discuss how to get started…..
Tom/David – Describe the work you did with the Energy Trust of Oregon (ETO).
• David, Mary, and I worked with the Energy Trust of Oregon to create a custom rebate that would help motivate our internal customers to save power in our data centers by refreshing 4 year old computers with new multi-core computers. To do this we had to understand the power use, and the CPU performance of the old and new computers. We also needed the estimated costs of the new computers, and the power draw. With this information the Energy Trust of Oregon created a rebate based on the reduction in the power used when old computers are replaced with a smaller number of computers that have an equal amount of compute power. The end formula looked something like this:
#old * old_CPU / new_CPU = #new; #old * old_power - #new * new_power = power saved; rebate = a fraction of the power saved, or a fraction of the cost of the project
Tom - Wasn’t it difficult gathering the documentation and data for the agreement?
• The first time it was a little difficult to find the best sources for the data. We looked at industry standard benchmarks for calculating CPU performance and at vendor specs for the power measurements. In the end we decided it was more accurate to measure the CPU performance and power usage based on our internal application usage.
Tom - How did you know what the refresh ratios should be?
• We used benchmarks that are based on internal applications. By running these benchmarks on the old and new computers we could calculate the throughput of each system type. Due to the single threaded nature of the benchmarks an instance of the benchmark was run for each core in the CPU. In the end the total throughput was compared.
David - How did you get power ratings?
• The power readings were established through a voltage meter. I might note that after the system boots, the Amps, Watts, and Volts normalize to lower values. It is recommended to launch a CPU stress test easily downloaded from the Internet. There are several available to select from. We used the same benchmark stress test application to simulate loads against the system(s) starting at 50% CPU utilization to 95%. The readings stayed fairly consistent from 60-80% CPU utilization with little to no increases. Capturing the average readings helped reinforce normalized production runs and actual simulation. We validated this by launching an actual job to the system(s) watching the increase and decrease of CPU utilization during the job start time and completion.
Tom - So, this was a customized agreement with the ETO; are rebates available that happen automatically (like you get when you purchase a new refrigerator)?
• Yes. This agreement was customized in order to get the most power savings for the high number of computers that are online. The Energy Trust of Oregon also has a standard Virtualization rebate that doesn’t require working directly with an ETO engineer.
David - What type of return did the ETO expect in the way of power saved?
• First and foremost, reduced power consumption.
• Although the ETO is based on green power and reducing power consumption, ETO has provided the catalyst for those customers willing to reduce kWh/yr or Kilowatt hours per year. This saves the local power company from building out new electrical grids, thus saving power companies millions of dollars for new infrastructure. So in truth, the small amount of refresh each company can do benefits the power company which in turn will reflect savings on your monthly/quarterly power bill.
Stay tuned for Part 2 where we will share the results and benefits