Merge pull request #26 from ScottLogic/SF-268

Sf 268

_data/information.yml

@@ -13,3 +13,5 @@
       link: /information/lifecycle/example/employee
     - name: Server Hardware
       link: /information/lifecycle/example/server
+- name: Networks
+  link: /information/networks

pages/categories/downstream.md

@@ -18,5 +18,12 @@ The [embodied carbon of end-user devices](/information/lifecycle/embodied) is no
 
 {% include linkedHeading.html heading="Network Data Transfer" level=2 %}
 
-These emissions are associated with the infrastructure enabling data transmission, enabling end-users to access the products and services. Network emissions can be estimated based on the size of the data transferred, the distance transmitted, and the carbon intensity of the regions through which the data is transferred.
+These emissions are associated with the infrastructure enabling data transmission, enabling end-users to access the products and services. Data that is consumed over the Internet is hard to measure as the specifications of the equipment used over the public internet are not available. Even if that data was available, the system has no control over what route the data takes. Therefore it is appropriate to use a proxy such as:
+- number of calls 
+- payload size
+- the distance transmitted
+- the carbon intensity of the regions through which the data is transferred
+
+All four categories of network listed in [the information on networks](/information/networks/networks) should be considered. Tools such as [co2.js](https://www.thegreenwebfoundation.org/co2-js/) and [Green Coding's Green Metrics Tool](https://www.green-coding.io/projects/green-metrics-tool/) can be used to gain insight and attributional metrics on the carbon emissions of data transfer.
+
 

pages/categories/operational.md

@@ -31,7 +31,21 @@ The energy consumed by on-premise servers and data centres.
 
 {% include linkedHeading.html heading="Networking Devices" level=4 %}
 
-The energy consumed by networking devices such as routers, switches, and Wi-Fi access points. 
+In [the information on networks](/information/networks/networks), networks are split into four categories - the operational network, the wired backbone internet, the wireless internet and your consumer's network. Of these, the first three need to be considered in Operational Emissions.
+
+Within the operational network, consider the running cost of the connected devices. These include, but are not limited to:
+- routers
+- switches
+- bridges
+- Wi-Fi access points
+- firewalls
+- modems
+- hubs
+- repeaters
+
+Research[^1] has shown that not all networking equipment will use power in line with the data it is transmitted over it. These devices will have a baseline load (i.e. the energy they consume for simply being switched on, regardless of the processing they do), a maximum load and a typical load. Most wired networking equipment, such as routers and switches, will consume at least 60% of their maximum power as a baseline going up to 90% of their maximum power under a heavy load[^2]. For accurate numbers, power monitors on the power socket are an excellent way of gathering data. As the power usage of networking equipment has very little variance while it is running, attributing the power consumption to specific applications or processes is less important.
+
+Data consumed from a supplier's API, that is external to your organisation, is accounted for in the indirect category. Public wired and wireless internet usage does not need to be measured in the Direct Emission category. Data sent from the service to end users is accounted for in [the downstream emissions](./downstream) category. 
 
 {% include linkedHeading.html heading="Employee Devices" level=4 %}
 
@@ -62,4 +76,9 @@ The emissions associated with cloud platform services like compute, storage, and
 Software-as-a-Service (SaaS) applications.
 
 {% include linkedHeading.html heading="Managed Services" level=3 %}
-Managed Security Operations, IT support and data backup services.
+Managed Security Operations, IT support and data backup services.
+
+## References
+
+[^1]: [Beyond network standby: A policy framework and action plan for low energy networks](https://nachhaltigwirtschaften.at/resources/iea_pdf/reports/iea_4e_network_standby_final_report.pdf) 
+[^2]: [Equipment power consumption in optical multilayer networks – source data](https://www.researchgate.net/publication/272819245_Equipment_power_consumption_in_optical_multilayer_networks_-_source_data)

pages/categories/upstream.md

@@ -34,7 +34,18 @@ Laptops, desktops, mobiles, printers, and peripherals used by employees.
 
 {% include linkedHeading.html heading="Networking Hardware" level=3 %} 
 
-Network routers, switches, Wi-Fi access points installed.
+When considering the upstream emissions of a network, consider the embodied carbon of any networking devices that are owned by the organisation. These include, but are not limited to:
+- routers
+- switches
+- bridges
+- Wi-Fi access points
+- firewalls
+- modems
+- hubs
+- repeaters
+- cables
+
+Ideally, data sheets from the manufacturer should be used to gather this data.
 
 {% include linkedHeading.html heading="Data Centre and Server Hardware" level=3 %} 
 

pages/information/networks/networks.md

@@ -0,0 +1,25 @@
+---
+layout: category
+title: Networks Emissions
+permalink: /information/networks
+---
+# Network Emissions
+Networks can be challenging when considering carbon emissions for two reasons:-
+- Networks are often shared.
+- Network equipment can draw a constant amount of baseline power, regardless of its usage.
+
+There are several carbon estimators that calculate the carbon emissions of software my multiplying the payload size in megabytes by a number (such as 0.06) to obtain a value in KWh. This technique is based on previous research that looked at the wired backbone internet. It is a fine proxy to use initially when evaluating an IT estate to see where effort is best applied to reduce carbon emissions. It also works well as an attributional estimate, allowing the organisation to allocate responsibility of emissions to different parties.
+
+However, this kind of estimation should not be used in a consequential context. This means that it should not use a multiplier to determine that carbon emissions have been lowered in a system by simply reducing its page size. This is because routers are high availability devices that consume most of their power, regardless of their load. Under high load, a wired router will still consume more power, but not a great deal more than when it is idle. Therefore, halving the size of its payload will not halve the power it consumes or the carbon emissions of the system.
+
+This does not mean that reducing the payload of a system transmitted by a network is without merit. By reducing page sizes and the number of network calls, the pressure the system contributes to both the public internet infrastructure and the clients networking hardware is reduced. This may allow upstream dependencies and downstream dependents to run less networking infrastructure or renew their hardware with less frequency.
+
+{% include linkedHeading.html heading="Network Categories" level=2 %}
+
+When considering measuring network hardware, it is useful to split it into four categories:
+- _The Operational Network_: This is the network of the organisation that owns the system. It includes routers, switches, WiFi access points and the like. This hardware can be catalogued and data sheets or power supplies can be inspected to determine the maximum power draw of each device. From this data, you can measure the baseline power draw and power draw under load of each device to calculate the carbon emissions of the network.
+- _Backbone Internet_: This includes the main infrastructure of the wired internet from an internet supplier, across all intermediate wired networks, to the closest it can get to the destination client's network or device. As this network carries traffic for all users of the internet, and cannot be measured,  an appropriate proxy can be used, such as multiplying page size in MB by 0.06 to gain an estimate in kWh.
+- _Public Wireless Networks_: This includes all public wireless networks such as 4G and 5G that carries data from its source (likely the wired backbone internet), to the destination device or network. Again, measurements here are impossible, so estimation techniques and proxy measurements must be used.
+- _The Client Network_: Finally, the destination client's network should be considered, if they are likely to have one. While client networks cannot be measured either, analytics and marketing data can provide insight into the kinds of customer who are using your system. This could be used to extrapolate an assumed network configuration for each segment of customers. For example, a B2B system aimed at small to medium sized enterprises can assume a network topology of a certain size, and multiply that up by the number of customers. Meanwhile a system that is used in homes could assume the presence of a home WiFi router and perhaps home mesh network.
+
+By considering each part of the data journey from its origin system to the final device in these categories, a better estimation of the carbon emissions created by the IT system can be put together. Each category of network should have appropriate measurements and proxies used to best estimate the total carbon cost of data transmission.