The Green and Virtual Data Center - Enabling Efficient, Effective, Productive Green IT and Computing

There are many different points of interest for collecting metrics in an IT data center for servers, storage, networking and facilities along with various points of interest or perspectives. Data center personal have varied interest from a facilities to a resource (server, storage, networking) usage and effectiveness perspective for normal use as well as planning purposes or comparison when evaluating new technology. Vendors have different uses for metrics during R&D, Q/A testing and marketing or sales campaigns as well as on-going service and support. Industry trade groups including 80 Plus, SNIA and the green grid along with goverentment groups including the EPA Energy Star are working to define and establish applicable metrics pertenet for Green and Virtual data centers.

To help put some things into context, stimulate thought and discussion, the above figure shows in the lower left a simple energy guide sticker label found on many common appliances based on some form of use. In the top left there is a different form of energy usage label in this case the EPA Fuel Economy Estimates reflecting amount of work or activity performed per unit of energy (miles per gallon city or highway).

Also shown in the middle of the slide are EPA categories or what are known in the IT server and storage environments as tiers and price bands of different solutions ranging from utility to small and low cost solutions being compared over city, highway, high-speed, with air-conditioning among other different applicable activity to show how to compare and contrast different vehicles. Also shown are best practices and driving tips. Food for thought!

Acronym	Description	Comment
DCiE	Data center Efficiency = (IT equipment / Total facility power) * 100	Shows a ratio of how well a data center is consuming power
DCPE	Data center Performance Efficiency = Effective IT workload / total facility power	Shows how effective data center is consuming power to produce a given level of service or work such as energy per transaction or energy per business function performed
PUE	Power usage effectiveness = Total facility power / IT equipment power	Inverse of DCE
Kilowatts (kw)	Watts / 1,000	One thousand watts
Annual kWh	kWh x 24 x 365	kWh used in on year
Megawatts (mw)	kW / 1,000	One thousand kW
BTU/hour	watts x 3.413	Heat generated in an hour from using energy in British Thermal Units. 12,000 BTU/hour can equate to 1 Ton of cooling.
kWh	1,000 watt hours	The number of watts used in one hour
Watts	Amps x Volts (e.g. 12 amps * 12 volts = 144 watts)	Unit of electrical energy power
Watts	BTU/hour x 0.293	Convert BTU/hr to watts
Volts	Watts / Amps (e.g. 144 watts / 12 amps = 12 volts)	The amount of force on electrons
Amps	Watts / Volts (e.g. 144 watts / 12 volts = 12 amps)	The flow rate of electricity
Volt-Amperes (VA)	Volts x Amps	Sometimes power expressed in Volt-Ampres
kVA	Volts x Amp / 1000	Number of kilovolt-ampres
kW	kVA x power-factor	Power factor is the efficiency of a piece of equipments use of power
kVA	kW / power-factor	Killovolt-Ampres
U	1U = 1.75”	EIA metric describing height of equipment in racks.
Activity / Watt	Amount of work accomplished per unit of energy consumed. This could be IOPS, Transactions or Bandwidth per watt.	Indicator how much work and how efficient energy is being used to accomplish useful work. This metric applies to active workloads or actively used and frequently accessed storage and data. Examples would be IOPS per watt, Bandwidth per watt, Transactions per watt, Users or streams per watt. Activity per watt should also be used in conjunction with another metric such as how much capacity is supported per watt and total watts consumed for a representative picture.
IOPS / Watt	Number of I/O operations (or transactions) / energy (watts)	Indicator of how effectively energy is being used to perform a given amount of work. The work could be I/Os, transactions, throughput or other indicator of application activity. For example SPC-1 / Watt, SPEC / Watt, TPC / Watt, transaction / watt, IOP / Watt.
Bandwidth / Watt	GBPS or TBPS or PBPS / Watt Amount of data transferred or moved per second and energy used. Often confused with Capacity per watt	This indicates how much data is moved or accessed per second or time interval per unit of energy consumed. This is often confused with capacity per watt given that both bandwidth and capacity reference GByte, TByte, PByte.
Capacity / Watt	GB or TB or PB (storage capacity space / watt	Indicator of how much capacity (space) or bandwidth supported in a given configuration or footprint per watt of energy. For inactive data or off-line and archive data, capacity per watt can be an effective measurement gauge however for active workloads and applications activity per watt also needs to be looked at to get a representative indicator of how energy is being used
Mhz / Watt	Processor performance / energy (watts)	Indicator of how effectively energy is being used by a CPU or processor.
Carbon Credit	Carbon offset credit	Offset credits that can be bought and sold to offset your CO2 emissions
CO2 Emission	Average 1.341 lbs per kWh of electricity generated	The amount of average carbon dioxide (CO2) emissions from generating an average kWh of electricity

Metrics include Data center Efficiency (DCiE) via the greengrid which is the indicator ratio of a IT data center energy efficiency defined as IT equipment (servers, disk and tape storage, networking switches, routers, printers, etc) / Total facility power x 100 (for percentage). For example, if the sum of all IT equipment energy usage resulted in 1,500 kilowatt hours (kWh) per month yet the total facility power including UPS, energy switching, power conversation and filtering, cooling and associated infrastructure costs as well as IT equipment resulting in 3,500 kWh, the DCiE would be (1,500 / 3,500) x 100 = 43%. DCiE can be used as a ratio for example to show in the above scenario that IT equipment accounts for about 43% of energy consumed by the data center with in this scenario 57% of electrical energy being consumed by cooling, conversion and conditioning or lighting.

Power usage effectiveness (PUE) is the indicator ratio of total energy being consumed by the data center to energy being used to operate IT equipment. PUE is defined as total facility power / IT equipment energy consumption. Using the above scenario PUE = 2.333 (3,500 / 1,500) which means that a server requiring 100 watts of power would actually require (2.333 * 100) 233.3 watts of energy that includes both direct power and cooling costs. Similarly a storage system that required 1,500 kWh of energy to power would require (1,500*2.333) 3,499.5 kWh of electrical power including cooling.

Another metric that has the potential to have meaning is Data center Performance Efficiency (DCPE) that takes into consideration how much useful and effective work is performed by the IT equipment and data center per energy consumed. DCPE is defined as useful work / total facility power with an example being some number of transactions processed using servers, networks and storage divided by energy for the data center to power and cool the equipment. An relatively easy and straightforward implementation of DCPE is an IOPs per watt measurement that looks at how many IOPs can be performed (regardless of size or type such as reads or writes) per unit of energy in this case watts.

DCPE = Useful work / Total facility power, for example IOPS per watt of energy used

IOPS per Watt = Number of IOPs (or bandwidth) / energy used by the storage system

The importance of these numbers and metrics is to focus on the larger impact of a piece of IT equipment that includes its cost and energy consumption that factors in cooling and other hosting or site environmental costs. Naturally energy costs and CO2 (carbon offsets) will vary by geography and region along with type of electrical power being used (Coal, Natural Gas, Nuclear, Wind, Thermo, Solar, etc) and other factors that should be kept in perspective as part of the big picture. Learn more in Chapter 5 "Measurement, Metrics, and Management of IT Resources" in the book "The Green and Virtual Data Center" (Auerbach).