CHAPTER 4: ACCOUNTING AND METRICS

Green IT needs baseline data to be established so progress can be appropriately tracked. At this point, there are two topics that need to be reviewed. The first involves accounting for energy costs and the second is process-related metrics.

Accounting for energy costs

The first step is to understand how much is being spent on electricity for the data centers. Many times this charge is buried in other electric utility bills and needs to be broken out, tracked and trended over time.

In terms of electricity, the more granular the understanding the better. Studies show that only around 30% of the electricity actually goes to the IT equipment⁶. The rest goes to cooling, power systems, and so on. By understanding electrical consumption and thus costs down to the device level, informed management decisions can be made.

Forecasting

For estimating purposes, to compute the approximate annual cost of a device, use the following logic:

Kilowatt Hours (kWh) = (Watts x Hours Used)/1000

Cost = kWh x Cost/kWh

The hours used can vary but if you have something running 24 hours a day x 365 days/year then that is 8,760 hours/year.

Let’s do the math for a 3,000 watt rack:

kWh = 3,000 x 8,760 / 1,000 = 26,280

If the cost is $0.0953 / kWh then:

Annual cost = 26,280 x $0.0953 = $2,504.48

As you can see, electrical costs can add up fast. Furthermore, if you assume that for each watt used by an IT device that is 30% of the energy that came into the data center, then suddenly that $2,504.48 becomes $8,347.28 worth of electricity just to run that one rack.

Variance analysis

IT service, rack and device level power consumption data can be used to conduct variance analysis to compare what was expected power consumption against what was actually consumed. This analysis may identify opportunities for improvement.

Process metrics

To begin with, there aren’t standard metrics today for tracking how IT is performing relative to power consumption. With that said, it is important to develop metrics that drive the desired behavior in the organization and provide the needed management information.

To begin we will look at metrics from The Green Grid, an industry association⁷.

Power usage effectiveness (PUE)

Identifies the ratio of total power to IT equipment. It can be used to compute potential demand. It is important to understand that the numerator is comprised of the total power feed going into the data center – cooling, power distribution, IT equipment, lights, etc.

PUE = Total data center power / IT equipment power

If PUE = 2.5 then for each watt of IT equipment added the total electricity demanded from the power grid will 2.5 watts. If a device requires 200 watts, then 200 x 2.5 = 500 watts will actually be needed from the grid.

Data center infrastructure efficiency (DCiE)

This is the reciprocal of PUE and can be used to report how much power is being used just by IT equipment.

DCiE = 1 / PUE x 100%

Green IT process metrics

Again, metric selection must be based on objectives and the culture of the organization. The following are potential options for consideration:

Monthly electrical demand and costs

If the main objective is to drive down electrical demand and costs then understanding consumption and costs from month to month is one of the most important elements to comprehend.

The more detailed it can be then the more opportunities there will be to identify areas of improvement and also to drive accountability in terms of what IT services provide to what functional areas are consuming large amounts of power.

Number of Green IT assessments performed

This is to track how many assessments were performed in a given period.

Number and percentage of Green IT assessments performed on schedule

Assessments should be scheduled to ensure that they happen and that all areas in scope are covered in the allocated time (for example, if all areas should be looked at twice per year).

Number and percentage of IT services reviewed

By reviewing IT services for electrical consumption, cooling needs and opportunities for improvement, then demand can be better managed.

Number and percentage of ghost systems detected

This will identify how many abandoned servers were detected. If there is a defined process for decommissioning then this would identify how well groups are complying with the process and/or notifying IT that the servers are no longer needed.

Number and percentage of servers consolidated

This will identify how many servers were consolidated in a given period. As servers are consolidated and unnecessary ones decommissioned then electrical demand should go down all things being equal.

Number and percentage of servers virtualized

This metric will identify how many physical traditional servers have been decommissioned and combined into a virtual server environment.

Some organizations may wish to combine consolidation and virtualization depending on the strategic direction.

Note: If a 3,000 watt rack is consolidated into a large virtual blade server environment, then the net effect may be to see electrical demand increase in the short term until utilization in the blade system is such that there are actual reductions in overall electrical demand.

Number and percentage of process integrations

Given the population of processes in the organization, this is the number and percentage of processes that have been formally integrated with.

Number of standards non-conformances identified

This is the number of configuration items that when inspected, did not conform to the organization’s standards for Green IT. For example, the number of servers with nonconforming power supplies. Groups may want to tier the severities as well to differentiate between small and material exceptions.

Number and percentage of standards nonconformances mitigated

This is the number and percentage of detected non-conformances that have been addressed to a state deemed acceptable by management.

Number of management exceptions

This is the number of formally approved exceptions where management said a given CI did not need to conform to a Green IT standard or policy.

General metrics considerations

There are a number of points to factor in when planning how to leverage metrics:

•  Meaningful and manageable – there are many different metrics that can be selected from. Rather than do all, select metrics that have meaning to the organization and how the Green IT process is intended to operate and also select metrics that have accurate data that can be readily collected.

•  Metrics drive behavior – bear in mind that the metrics selected will affect how stakeholders behave. Early on following the implementation of Green IT, metrics relating to effectiveness and compliance to the process will be very important. As the process matures, economy and efficiency will become increasingly important.

•  Absolute numbers and percentages – when designing reports, it can be critical to decide when to use absolute numbers, percentages, or a combination. For example, to report that 100% of assessments were conducted on time sounds good but was that 100% of one scheduled assessment or 100% of 50 assessments? The use of absolute numbers can communicate volume to the reader.

•  Trending and baselines – metrics reported at a point in time only tell part of the story. By trending and setting baselines, organizations can gain further insight. The use of statistical process control (SPC) principles can provide even further analytical abilities to understand whether the process is in control.

⁶ “Guidelines for Energy-Efficient Datacenters”,
The Green Grid, 16 February 2007,
http://www.thegreengrid.org.

⁷  “The Green Grid Data Center Power Efficiency Metrics: PUE and DCiE”, The Green Grid, 2007, http://www.thegreengrid.org.