Shift Workload

Execute tasks when the grid is powered by renewable energy

Carbon Aware Computing (2022) All sources

category: Sustainability class: Task Rules

Description

The basic idea behind the time shifting approach is moving the computing load into a point in time when the power grid has a maximum of renewable energy. This will result in a lower emission of CO2 of a computing task.

In addition to energy and hardware efficiency, carbon efficiency is one of the most important principles of green software development, like the Energiewende. Energy is used when it is available from renewables.

Applied to business processes, energy-intensive activities are not executed as soon as they become enabled, but at the point within their allowed window when the carbon intensity of the grid is lowest. Typical candidates are automated activities such as batch jobs, calculations, data processing or machine runs. The pattern therefore needs temporal slack: the service-level agreement or deadline defines how far execution may be shifted.

Implementation

We shift the energy demand to a point in time when the carbon intensity of the grid is lower. To do this we need forecast data and software libraries to calculate the appropriate execution window. Open tools exist for this, such as the Carbon Aware SDK of the Green Software Foundation. The Camunda Carbon Reductor case study shows how a process engine places such a check in front of an energy-intensive task, and Reducing carbon emissions in a core system describes time shifting for the workloads of a bank’s risk platform.

The pattern complements Delay Execution, which also postpones work but for flexibility reasons, and Use eco-friendly resources, whose time-slot variation applies the same idea to energy as a resource.

Key Takeaways

  1. Use temporal slack — Shift energy-intensive activities to times of low grid carbon intensity within their allowed execution window.
  2. Respect the SLA — The deadline or service level defines how far an activity may be postponed.
  3. Rely on forecasts — Carbon intensity forecasts and libraries such as the Carbon Aware SDK compute the best execution window.
  4. Report the savings — Track avoided emissions so that the effect of the shift becomes visible.

Performance considerations

QualityTimeFlexibilityCost080

■ This pattern   ■ Baseline

The ratings show a clear sustainability gain (70), with all other dimensions neutral. Shifting does not change what an activity does or how much energy it uses; it changes when the energy is drawn, so emissions per execution fall when more renewable power is in the grid. Time is rated neutral because the shift stays within the window the SLA allows: individual cases may finish later than before, but not later than promised. Cost, quality and flexibility stay neutral because the activity itself is unchanged. The forecast integration adds some implementation effort, and a shifted activity keeps less buffer for unexpected delays within its window.

Example

The Camunda Carbon Reductor case study applies the pattern in a process engine: an element placed before an energy-intensive task checks a forecast of marginal carbon intensity and postpones the remaining tasks if this saves carbon, while still meeting the maximum runtime of the process instance. In Reducing carbon emissions in a core system, UBS and Microsoft applied time shifting to risk-platform workloads. As an illustration at process level, an insurer could run its nightly batch for document generation in the hours with the lowest forecast carbon intensity instead of at a fixed start time, as long as results are ready by the start of business.

Spotting the opportunity in event data

Framed along the FLAC rule types — fitness, location, attribute, constraint.

Fitness: Processes with energy-intensive automated activities and loose timing, e.g. batch processing, reporting, data transfers, model training or machine runs in production. Location: Activities with slack between enablement and deadline, visible as the gap between the preceding event and the SLA due date, and activities scheduled at fixed times regardless of grid conditions. Attributes: Activity duration and energy use, execution timestamps, deadline or SLA per case, and the region of execution, joined with historical grid carbon intensity. Constraints: Don’t shift customer-facing or time-critical activities; treat the SLA as a hard constraint; consider that shifting many cases into the same window can create load peaks.

Sources & references

  1. Carbon Aware Computing

Share this pattern

Enjoy these

Foundational free Patterns

Apply Integral technology

Elevate physical constraints by applying new technology

Apply triage

Consider the division of a general activity into two or more alternative activities

Appoint case managers

Assign a responsible individual for handling each case type

Assign cases

Let workers perform as many steps as possible for single cases

Automate activities

Consider automating activities

Automate for environmental impact

Implement automation in a sustainable way

Buffer information

Buffer external information and subscribe to updates

Combine activities

Combine small activities into composite activities

Consolidate Work

Collect similar work items and work in batches

Constraint-based task assignment

Allocate tasks considering business process execution constraints

← Back

Connected to

Use eco-friendly resources

Replace underlying resources with eco-friendly alternatives

What's Happening?

Dec 2023
New Pattern

Consolidate Work

Nov 2023
Oct 2023
Oct 2023
New Cheat sheet

Redesign Essentials

Sep 2023
Sep 2023
Aug 2023
New Pattern

Task delegation

Aug 2023
Aug 2023
New Pattern

Green Feature

Aug 2023