Green IT Best Practices

In this section, we will discuss various practices to leverage Hardware Sentry and its Green IT extension to take your organization to the path of greener IT. It is assumed that the solution has been correctly deployed on PATROL Agents and setup on TrueSight Presentation Server.

Use Groups to Represent Datacenters and Server Rooms

The goal is to report the energy consumption of your organization’s IT infrastructure and the associated cost and carbon footprint.

Technically, the solution analyzes all physical devices’ power consumption in Watts and calculates the corresponding amount of energy for a given period of time, typically in kWh (kilowatt-hour). To calculate the associated costs and carbon footprint, the Green IT extension requires you to specify the electricity price and carbon weight per kWh.

The price and carbon weight of a kWh largely depends on your electricity supplier. Your organization may have a different electricity supplier for each site that hosts servers and other IT equipment (or even just a different contract).

The Green IT extension can be configured with different electricity prices and carbon weight for different Groups in TrueSight. Therefore, we strongly recommend representing each datacenter or each server room with a separate Group in TrueSight.

In most organizations, a server’s DNS name reflects its physical location (example: server01.ladefense-paris.sentrysoftware.net). In this case, you can simply create Rule-based Groups in TrueSight, with a Device Selection Criteria matching on the Device DNS Name as in the example below:

Create a Rule-based Group in TrueSight for each Datacenter or Server Room

The electricity costs of each Group in TrueSight is monitored separately and displayed in the Green IT reports. But there are many Groups in TrueSight that are irrelevant to the electricity consumption and carbon footprint reports, notably groups created dynamically to represent VMware clusters and hosts. A typical TrueSight setup may have hundreds if not thousands of Groups defined.

To avoid clutter in the Green IT reports, we strongly recommend that you choose a naming convention to easily filter groups that represent datacenters or server rooms. For example, you can prefix the name of such groups with “DC” (for Data Center) or “Zone”.

You can then configure the Green IT extension to consider the electricity consumption only for the Groups that match a specified regular expression (DC or Zone as in the example below).

Filter TrueSight Groups for Green IT reports

Note: Devices may belong to several Groups and you may want to configure Groups of Groups (for example, a large "Everything" group). It is important to make sure that electricity prices and carbon weights are indicated only once for a given Device. For example: specify the electricity price and carbon weight per kWh for the "DC Paris-La Defense" Group, but not for the "DC Everything European" one.

Electricity Pricing

The Green IT extension will calculate the electricity costs of your servers based on a simple flat rate expressed in dollars per kWh. However, your organization probably gets invoiced with much more sophisticated pricing structures by its electricity suppliers.

Because of the technical challenges and usability issues, it is impossible to specify advanced electricity billing with tiering or even time of use (e.g. calculations need to happen once every billing period, for example, to establish peak consumption).

Our recommendation is to simply calculate the equivalent flat rate based on:

  • the total amount of the invoice for the considered period
  • divided by the total number of kWh for the same period

The calculation of the electricity consumed by server rooms and datacenters, servers, applications and services is aggregated over days and months. Applying the same flat rate will still provide an excellent approximation of the electricity costs because applications and even servers do not adapt their workload based on the electricity supplier billing rules, but on based on actual user demand.

If your organization is being invoiced for its electricity consumption globally, you can configure the Green IT extension with a global Electricity Rate that will be applied to all Groups (unless a specific value is configured for a group).

Electricity prices typically range from $0/kWh to $0.3/kWh. See Electricity prices per country (select the "for businesses" option), as reported by GlobalPetrolPrices.com.

Note About Currencies

In the Green IT reports, all prices are expressed in a virtual dollar ($), which has nothing to do with the actual US Dollar. The dollar symbol in the Green IT reports represent a monetary unit. You can decide arbitrarily that the dollar symbol in the reports is actually a Yen, a Pesos, a Canadian Dollar or any currency your organization is working with. However, it is critical to make sure the same currency is being used for all Groups, even if you have datacenters in different regions of the world using different currencies.

The Green IT extension calculates the electricity costs of applications and services, which may pan across several datacenters in different countries. No currency exchange rates will be taken into account by the algorithm. That’s why you need to convert (and refresh as often as necessary) the different electricity rates to one single currency that will be used across all reports.

Assess CO₂ Emission per kWh

Most electricity suppliers advertise their carbon footprint and expressly show the carbon weight of a consumed kWh of electricity on the invoice itself (if they do not promote it, it’s probably not good news). This is notably the case when your organization has a contract with the electricity supplier that stipulates a particular mix of energy sources.

As for the electricity rate, the Green IT extension can be configured only with a flat rate, expressed in kilograms of CO₂ per kWh. You will need to consider the average CO₂ emissions advertised by the energy supplier of your organization.

In general, this number does not vary much as electricity suppliers struggle to transition from energy sources with high Greenhouse Gas emissions to more renewable energy sources (it takes years to build new power plants). You probably won’t need to update the values you configure in the Green IT extension very often.

If your energy supplier does not provide you with the average CO₂ emissions in kg/kWh, you can get this information from various external sources. For state-owned suppliers, the information if probably published by the government in the applicable region. For private companies, the information is probably shared on the company’s Web site.

If this information still cannot be found, you can rely on energy mix data published by each country. Many energy suppliers in the private sector are “virtual” suppliers, as they simply distribute the energy produced by power plants owned by “real” suppliers, and each country or region only has a handful of “real” suppliers (and often only a single one, state-owned or formerly state-owned).

An excellent source of information is Climate Transparency, with details about the energy mix in various countries.

The average carbon emission per kWh in the G20 countries is 0.458 kg CO₂/kWh.

Notes About Units

Often the CO₂ emission is expressed in grams per kWh (gCO₂/kWh). Do not forget to convert to kilograms (divide by 1,000). The value will be much less than 1 in most cases.

Sometimes, CO₂ emission is expressed in tCO₂/TJ (tons of CO₂ per terajoule). A TJ equals 277,777 kWh. To convert tCO₂/TJ to kg/kWh, simply divide by 278.

Green Applications and Green Services

Similarly to how TrueSight organizes metrics and events, the Green IT extension calculates the electricity consumption, its associated costs and carbon emissions for each Application and Service defined in TrueSight.

This will allow the IT department to present a carbon emission report to the application owners.

Green IT Report for Services, with electricity costs and CO₂ emissions for each Service defined in TrueSight

For the calculations to work properly, each Application and Service must include all Devices (as defined in TrueSight) it relies on. If the Application or Service configuration includes only Attributes, the Green IT reports will not show any energy consumption.

Virtual Machines

If an Application relies on virtual machines (VMs), the corresponding Devices can be added to the Application definition. The Green IT extension will find the physical machine this VM runs on and estimates the fraction of electrical power the VM uses on the physical machine.

For the Green IT extension to retrieve the physical machine a VM runs on, the physical host needs to be monitored with PATROL for VMware vSphere, which will create the necessary metadata and groups in TrueSight.

Public Clouds

If your Applications and Services rely on an external public cloud infrastructure, the Green IT extension will not be able to calculate the carbon footprint of the part that is hosted outside of your physical infrastructure. You will need to contact your public cloud provider to get this information.

Cloud Provider Sustainability Information
Microsoft Azure The Sustainability Calculator lets you calculate the carbon footprint based on your consumption of cloud services
Amazon AWS General information about Amazon’s cloud infrastructure sustainability
Google Cloud Platform General information about Google Datacenters Efficiency
Machine Learning (generic) Generic Carbon footprint calculator for Machine Learning workloads

Datacenter Heating Margin and P.U.E.

P.U.E.

The P.U.E. (Power Usage Effectiveness) represents the energy efficiency of the datacenter itself: cooling, power distribution, etc. The P.U.E. does not represent the efficiency of the servers, switches, and storage systems.

A non-optimized datacenter may have a P.U.E. around 2, while a most efficient datacenter achieves a P.U.E. of 1.1. The P.U.E. is mostly driven up by a datacenter’s air conditioning systems, which typically represent 40% of the overall electricity consumption of a datacenter.

The Green IT reports do not compute the P.U.E. (because the extension lacks some important metrics to do so), but they will help you improve it significantly.

In most datacenters, the air conditioning system ensures the entire room’s ambient temperature is stable at 18 degrees Celsius, which is really unnecessarily low. Datacenter operators like their server rooms as cool as possible because they fear overheating problems (and rightly so). It is common knowledge that computers work better at lower temperatures.

However, it has been proven that computer systems can operate without problems with an ambient temperature significantly higher (see Google’s example, where they raised the temperature of their datacenters to 80°F, i.e. 26.7°C). This is the fastest and cheapest method to reduce the energy consumed by a datacenter and improve its P.U.E.

Each degree Celsius increment represent a 5% saving on the cooling systems.

Heating Margin

Most organizations cannot risk optimizing their datacenters’ temperature because they do not know how much they can raise the temperature without triggering various overheating issues (hardware failures, crashes, data corruption, etc.).

Now let’s remember that Hardware Sentry tracks every single temperature sensor inside servers, switches and storage systems, and check their values against alert thresholds recommended by the manufacturers. With this information, we know how much each device’s temperature can be raised before triggering a warning alert. This quantity is called the Heating Margin and is displayed for each physical Device in TrueSight with the Hardware Sentry Component.

Even better, once you have deployed and configured Hardware Sentry to monitor all of your servers, switches and storage systems, the Green IT extension aggregates the heating margin of all physical Devices belonging to a Group (representing a datacenter or a server room), over a period of 24 hours to cover temperature variations during the day.

Small server room represented as a Group in the Green IT reports, with its Heating Margin and Ambient Temperature

Example: If the Heating Margin for a Group is 23°C, it means you can potentially raise the temperature of the corresponding datacenter or server room by 23 degrees before triggering a warning on any of the temperature sensors in the servers, switches and storage systems.

Heating Margin Coverage

Before you start increasing the temperature of your datacenters, you need to make sure the calculated heating margin takes into account all physical Devices in a given server room.

First of all, you need to deploy and configure Hardware Sentry to monitor 100% of your IT equipment. With more than 250 platforms supported, this KM is the best hardware monitoring solution on the market, by far.

Unfortunately, even the best solution cannot monitor equipment that do not report the value of their temperature sensors, or do not have appropriate alert thresholds. For such Devices, it is not possible to assess their Heating Margin.

Having a few poorly designed switches or servers in your datacenter should not prevent you from trying to improve its energy efficiency. The Green IT reports display for each Group (i.e. each datacenter or server room) the percentage of equipment that exposes enough information to calculate their heating margin. This percentage is called Heating Margin Coverage.

The Heating Margin Coverage report lists the physical Devices for each the heating margin can be calculated or not

Higher coverage percentage means more confidence in the Heating Margin value and, therefore, less risk of overheating a piece of equipment while increasing the room’s temperature.

Recommendations

The table below summarizes our recommendations in terms of temperature increase in your datacenters.

Heating Margin Coverage Max. Temperature Increase Saving
< 10°C < 50% 1°C or 2°C 5% to 10%
< 10°C 50% to 90% 3°C 15%
< 10°C > 90% 5°C 25%
> 10°C < 50% 3°C to 5°C 15% to 25%
> 10°C 50% to 90% 5°C to 10°C 25% to 50%
> 10°C > 90% reduce margin to 10°C typically 25% to 70%
> 10°C > 95% reduce margin to 5°C typically 50% to 90%
Keywords:
co2 carbon emission ghg energy power consumption green hardware km patrol