2.3 x86 Asset Efficiency

What is it?

Compute power is subdivided in two categories, doing this circumvents the discussions on whether RISC/EPIC or x86 processors are better. Both architectures have their merit and can be optimized independently because they are most of the times designed and managed independently from each other.

For servers, use the CPU load percentage as the leading number for characterizing usage. With virtualization and other consolidation techniques, memory/RAM often is the limiting factor. In either case, relate the installed capacity, its utilization rate and the reference capacity and decide what the outcome means:

  • Based on CPU load and the score is low? Next servers should use less processors/cores and have low-voltage processors. You have plenty of CPU capacity and you can do fine with less.
  • Based on memory/RAM and the score is low? Put less memory in the new servers and you could even move memory from the current servers to new servers that are procured with deliberately less memory.

For CPU load reference capacity you can use benchmarks like SPECint 2006, SPECpower and alike. There is even one that focuses specifically on virtualization: SPECvirt.

Finally, relate the result to the amount of power or energy used by the devices. The choice for low-energy devices or power savings options is rewarded in the final result.

Have a look at the parent page for this KPI for an example and advice on how to proceed with this kind of KPI.


Why should you care?

The KPI covers both the effective use of capacity and design of the x86 server architecture. The server is always on and it consumes a large amount of electrical energy (the device, cooling and UPS). When designing the server environment and deciding which devices are needed, you should take the impact on energy consumption into account. Make the server capacity scalable so the utilization and therefore its asset efficiency and energy consumption is optimal.


How does it relate to the other items in the model?

Increasing asset efficiency is one of the most effective ways of improving overall energy efficiency. High utilization ratios will be reflected in relatively lower costs for energy, hardware and support contracts.

What is the impact of increasing the utilization by decreasing the installed capacity?

  • KPI DCiE is impact negatively because IT draws less power. The efficiency of the cooling system and the UPS will drop too unless their capacity is adjusted to the new situation.
  • KPI Floor usage is impact negatively because less IT devices are in the data center. The solution could be to decrease the size of the data center.
  • KPI’s Bypass and Recirculation are impacted negatively if the airflows are not adjusted properly with blanking panels, perforated tiles, etc.
  • KPI Compute Management stays neutral or will improve. To be able to determine which and what capacity can be removed, historical trends are required. If this KPI is not at the right level yet, improve it first before deciding what installed capacity to remove.
  • KPI Change and Configuration Management is impacted positively when improving the process(es) to check for capacity in a change, to timely expand the capacity and to record the capacity and how it is used in the CMDB.
  • KPI Product Lifecycle Management is impacted positively if the process is improved. When procuring new devices, add the real energy consumption (or power draw) to the criteria. During the device’s operational lifetime, add the energy consumption to the TCO to determine when the device should be replaced.
  • KPI Capacity Management is impacted positively if the process is improved. When the amount of overcapacity is less than before, you should monitor the current and future/expected capacity utilization closely.
  • KPI Service Level Management is impacted positively if the process is improved. Because of the decision to remove over-capacity, you show a better awareness of energy consumption. It can be combined with charging back real energy costs to IT and/or defining energy reduction initiatives.