AWS Deadline Cloud Shared Job Bundles: Simple, Scalable, & Powerful

The Modern Rendering Paradigm on AWS Deadline Cloud

Render management has undergone a massive generational shift. For decades, visual effects facilities maintained bulky centralized render dispatch servers, custom submission wrappers with thousands of lines of fragile legacy code, and disparate template systems across departments.

With the advent of AWS Deadline Cloud, rendering natively embraces elastic cloud primitives. One of its most powerful yet underutilized features is Shared Job Bundles.

Shared job bundles allow studio pipeline TDs and technical supervisors to package, validate, and publish standardized, reusable job templates directly onto AWS Deadline Cloud queues. Teammates can then browse, inspect, parameterize, and submit complex render tasks with zero extra server infrastructure.

What Exactly Is a Job Bundle?

Under the hood, AWS Deadline Cloud operates on the open OpenJobDescription (OJD) specification. An OpenJobDescription bundle consists of:

1. template.yaml / template.json: Declares the parameters (e.g. frame range, resolution, camera, output path), required worker capabilities (GPUs, CPU count, RAM, operating system), and task definition scripts.
2. parameter_values.yaml: Default parameter assignments.
3. Asset Manifests & Helper Scripts: Any custom pre-flight or post-flight scripts, Python helper hooks, or submission manifests required to prepare the run.

# OpenJobDescription Job Template Example
specificationVersion: 'jobtemplate-2023-09'
name: Nuke Daily Comp Render
parameterDefinitions:
  - name: ScriptPath
    type: PATH
    objectType: FILE
    description: Absolute path to the .nk script
  - name: FrameRange
    type: STRING
    default: 1001-1150
    description: Frames to execute
steps:
  - name: RenderFrames
    parameterSpace:
      taskParameterDefinitions:
        - name: Frame
          type: INT
          range: '{{Param.FrameRange}}'
    script:
      actions:
        onRun:
          command: 'nuke'
          args:
            - '-x'
            - '-F'
            - '{{Task.Param.Frame}}'
            - '{{Param.ScriptPath}}'

How Shared Job Bundles Work

When you share a job bundle to a queue, Deadline Cloud packages the entire directory into a compressed archive (.ojd) and uploads it to your queue's associated Amazon S3 bucket under the job-bundles/ prefix.

Because the bundle lives in the queue's S3 storage:

• Instant Discoverability: Any artist with IAM access to that queue can see the shared bundle in their Deadline Cloud GUI or CLI.
• Strict Versioning: Updates are atomic. Multiple iterations can be tested without breaking artist daily renders.
• Zero Local Client Dependencies: Artists do not need complex local toolchain installations; the queue workers pull the exact bundle definition and execution steps from S3.

# Packaging and sharing a bundle using the AWS Deadline Cloud CLI
deadline bundle share \
  --queue-id queue-0123456789abcdef0 \
  --bundle-dir ./pipeline/bundles/nuke_comp_v2 \
  --name "Nuke Comp 4K Standard" \
  --description "Standard 4K OpenEXR composite render with OCIO ACEScg conversion"

Managing Shared Bundles via the CLI

The AWS Deadline Cloud CLI makes management straightforward for Pipeline TDs:

# List all active shared bundles on a queue
deadline bundle list --queue-id queue-0123456789abcdef0

# Download a shared bundle to inspect or modify
deadline bundle download \
  --queue-id queue-0123456789abcdef0 \
  --bundle-id bundle-abcdef123456 \
  --output-dir ./inspected_bundle

# Submit a job directly using a shared bundle template with custom parameters
deadline job submit \
  --queue-id queue-0123456789abcdef0 \
  --shared-bundle-id bundle-abcdef123456 \
  --parameter-values 'ScriptPath=/shows/proj_alpha/shots/sq01/sh020/comp/sh020_comp_v004.nk' \
  --parameter-values 'FrameRange=1001-1050'

Real-World Pipeline Integration: Maya, Nuke & Flow Production Tracking

At DownZero, we integrate Shared Job Bundles directly into artist DCC menus:

1. Artist One-Click Export: In Nuke or Maya, the artist clicks DownZero > Dispatch to Deadline Cloud.
2. Parameter Extraction: A lightweight PySide UI auto-populates the current project, shot, and write nodes queried live from Autodesk Flow Production Tracking (ShotGrid).
3. Queue Assignment: Based on the show's quota and current deadline, the job is dispatched to a queue backed by AWS Spot instances for non-critical dailies, or On-Demand GPU workers for delivery masters.
4. Automated Slating & Review: Once tasks finish, an integrated post-flight task renders an H.264 review quicktime with shot slating and uploads it automatically to Flow Tracking.

Key Benefits Observed in Production

• 70% Reduction in TD Dispatch Debugging: Because all workers run the verified bundle script from S3, "it works on my machine" errors are completely eliminated.
• Instant Cost Control: Budgets are enforced at the queue level; queue policies prevent runaway frame rendering.
• Effortless Multi-Region Scaling: Bundles can be synchronized across global queues in US-West (Oregon) and EU-West (London) within seconds.