Greening of Streaming has an objective to improve the energy efficiency of streaming ecosystems. It adopts a WATTs first target, in the belief that by focusing on reducing the power used by the complete system there will be benefits both in a reduced TCO and in a raised ESG position, ultimately benefiting stakeholders, shareholders, subscribers and consumers.
On reviewing the supply chain for common frames of reference around which to orient systemic change with the goal of improved energy efficiency, WATTs have proven to be the one consistent measure that ALL streaming service providers can potentially measure. From energy supply through the telecoms and internet switching and routing, right up into the service provisioning and out to the consumers, every participating digital system can potentially provide WATT measurement.
Some, we have found, will only provide Ampere and Voltage, but WATTage can be derived from these KPIs. Other infrastructures have enterprise management systems that can monitor power demand from individual processes, chipsets, PSUs, PDUs and top-of-rack systems. It is complex to measure streaming services that span multiple systems. In many cases these are ‘nested’ systems such as a process on a CPU on a server with PSUs, in racks with PDUs which in turn are monitored at a ‘room’ or ‘availability zone’. The wrong measurement will count the same WATT multiple times, or miss evaluating the demand for energy-hungry ‘parent systems’. It is therefore critical that a complete systemic view be taken before any claims about the power requirement are made by individual contributors to the ecosystem, and particularly before any claims of reduction in power are made. Not least, it may be that the measurement is focused on the wrong power load and a reduction in one area may cause unintended increases, and even significant increases in power consumption elsewhere in the complete system view. Without establishing a consistent way of measuring, we can’t hope to ‘actually’ reduce energy consumption.
Despite the challenges of measuring and relating streaming services to power demand, Watts do at least provide a single constant reference throughout all disciplines. Energy providers bill in kWh at the point of consumption by the infrastructure providers. kWh are volumetric counts that are metered. For larger enterprises this billing can be resolved to 15-min increments. There is also typically a ‘peak’ supply load measured in Watts. All systems must run within that peak loading. Actual load is not often monitored or reported outside of the energy providers’ own facilities.
A high-level schema of a streaming service and its relationships to measuring WATTs throughout the system:
This diagram is a high-level schema of a streaming service and its relationships to measuring WATTs throughout the system.
In the diagram above a demarcation of ‘scopes’ is laid out (in blue) as per the GHG protocol, Scope 1 and 2 (‘those we can directly influence’) are in the centre box. The balance of supply (billed in kWh from energy providers) can be directly controlled. The kWh is in turn consumed by a combination of Network Facilities and Application Hosting facilities. These are also under our control, and a strong engineering focus should be, and is being fostered by Greening of Streaming, to focus on measuring this power load, and on reducing that demand – from ‘any’ energy supplier.
The supply to these systems is measured in kWh. The Network Facilities and Application Hosting systems do not by default measure kWh consumption, but can typically readily provide current power consumption levels in Watts directly (or indirectly via a calculation of Voltage and Current draw).
The balance between network facilities and application hosting energy demand will vary depending on the role the streaming service provider plays in the ecosystem. This means that at the supply stage we have to be able to measure the power demand for the provisioned infrastructure required to make the streaming service available. While the energy consumption will most closely correlate to the kWh billing from the energy providers, rather than with the traffic, the infrastructure will scale up as the service models scale up, and increase the energy demand. Planning for energy as that infrastructure scales is of essential importance.
Some model conversion metrics have been included below to help create some idea of the relationship between these measures of Watts from the infrastructure and kWh from the supply in some model scenarios.
| Total Wh | Total Ah | Voltage | Interval (mins) | Equals to kWh | Traffic GB/hour | Bitrate Mbps (with overhead) | Viewers | Viewers/kWh/interval | Wh/viewer/interval |
| 250.00 | 15.00 | 1.00 | 140,000.00 | 5.00 | 62,222 | 62,222.22 | 0.02 | ||
| 1.50 | 230.00 | 15.00 | 1.38 | 180,000.00 | 5.00 | 80,000 | 57,971.01 | 0.02 | |
| 26,400.00 | 60.00 | 26.40 | 2,970,000.00 | 5.00 | 1,320,00 | 50,000.00 | 0.02 | ||
| 50.00 | 15.00 | 0.20 | 22,500.00 | 5.00 | 10,000 | 50,000.00 | 0.02 | ||
Data attribution models then have a role in helping to translate actual energy consumption and power demand to make the infrastructure correlate to an identifiable ‘Watts per service’ model, so we can understand how power demand will evolve with the scaling of the business.