d4c-service-main-site/content/docs/dissemination/_index.md

title, weight, prev, next

title	weight	prev	next
🌐 Data Dissemination Strategy	3	/docs/processes/field_imagery/	/contact/

{{< callout type="important" icon="sparkles" >}} We prioritize interoperability, long-term preservation, and decentralized resilience. {{< /callout >}}

High-Level Overview

flowchart TD
    classDef linkNode stroke:#0000EE,color:#0000EE,stroke-width:2px;
    subgraph mirrors [Mirrors & Preservation]
        SourceCoop[Source Cooperative]
        Zenodo[Zenodo]
        InternetArchive[Internet Archive]
    end

    Sources[Open Data Sources]
    Processes[Transformation Pipelines]
    Artifacts[Systems-Ready Data]
    
    subgraph CoreInfra [Data for Canada Infrastructure]
        Portal[Object Storage]
        Metadata[FAIR Data Catalog]
    end
    
    Distribution[Decentralized Distribution]
    Torrent[BitTorrent Swarm]
    
    subgraph Consumers [Consumption]
        Users[Data People & Developers]
        Systems[Automated Systems]
    end

    %% Flow with Animations
    Sources a1@--> Processes
    a1@{animate: true, animation: slow}
    
    Processes a2@--> Artifacts
    a2@{animate: true, animation: slow}
    
    Artifacts a3@--> Portal
    a3@{animate: true, animation: slow}
    
    Portal a4@--> Metadata
    a4@{animate: true, animation: fast}
    
    Metadata a5@--> Distribution
    a5@{animate: true, animation: fast}

    %% Distribution Logic
    Distribution a7@-.->|Primary| mirrors
    a7@{animate: true, animation: slow}
    
    Distribution a8@-.->|Resiliency| Torrent
    a8@{animate: true, animation: slow}

    %% Mirror Connections
    mirrors a12@--> Users
    a12@{animate: true, animation: slow}
    
    mirrors a13@--> Systems
    a13@{animate: true, animation: slow}
    
    mirrors a9@-.->|Seeding| Torrent
    a9@{animate: true, animation: fast}

    %% Torrent Connections
    Torrent a10@--> Users
    a10@{animate: true, animation: fast}
    
    Torrent a11@--> Systems
    a11@{animate: true, animation: fast}

    %% Click Actions
    click Sources "https://www.dataforcanada.org/#high-level-overview/" _blank
    click Processes "https://www.dataforcanada.org/docs/processes/" _blank
    click Artifacts "https://www.dataforcanada.org/docs/getting_started/" _blank
    click Metadata "https://stac-utils.github.io/stac-geoparquet/latest/spec/stac-geoparquet-spec/" _blank
    click Zenodo "https://zenodo.org/communities/dataforcanada/" _blank
    click SourceCoop "https://source.coop/dataforcanada/" _blank
    click InternetArchive "https://archive.org/details/@diegoripley/uploads/" _blank

    %% APPLY STYLES TO LINKED NODES
    class Sources,Processes,Artifacts,Metadata,SourceCoop,Zenodo,InternetArchive linkNode

🏗️ Open Processing Architecture

We believe that true open data requires open production. To ensure the longevity and resilience of Canada's data infrastructure, we treat our data pipelines as open source software artifacts. We provide the "blueprints" alongside the data, allowing any user to verify our work or rebuild the dataset from scratch on their own infrastructure.

The Blueprint Model

Our processing strategy relies on three immutable components to guarantee transparency:

1. Build Manifests: Every dataset version is accompanied by a strict manifest. This locks the exact "ingredients" used: the cryptographic hashes of the raw source files, the specific Git commit of the processing code, and the configuration parameters.
2. Environment Definitions: Rather than opaque binaries, we publish the exact Infrastructure as Code (IaC) definitions (e.g., Dockerfiles). This allows users to inspect the system context (GDAL versions, libraries, and dependencies) and build the environment themselves.
3. Deterministic Builds: By combining a Build Manifest with our Environment Definitions, any user can execute a deterministic build. This process guarantees a bit-for-bit identical copy of the official Data for Canada artifact, ensuring that the pipeline is independent of our specific servers.

Mirrored Source Artifacts: Crucially, we do not rely solely on external version control systems like GitHub, which may change or disappear. A complete snapshot of the processing code, environment definitions, and manifests is bundled with every data release. These source artifacts are replicated across Source Cooperative, Zenodo, the Internet Archive, Data for Canada infrastructure, and the community, ensuring that the method of creation is preserved with the same redundancy as the result.

Dissemination Process

Once data products reach a production-ready state, they enter a dissemination flow designed for permanence and performance:

• Cloud-Native First: Priority is given to performant, system-to-system file formats (e.g., Parquet) to enable high-throughput applications without the need for local parsing.
• Persistent Identification: Every dataset version is assigned a DOI for citation and immutability.
• The FAIR Data Catalog: Global metadata is aggregated into a single, queryable FAIR Data Catalog. This catalog acts as the "brain" of the system, tracking all versions and DOIs, and directing users to the optimal source within our multi-tier storage network:
  • Source Cooperative serves as our primary mirror for all datasets, including large-scale products like orthoimagery.
  • Zenodo serves as our repository for long-term academic preservation and provides a high-speed mirror for European users.
  • The Internet Archive is utilized strategically for specific datasets to ensure historical redundancy.
  • Data for Canada Infrastructure is utilized strategically for specific datasets of high-value.

Decentralized Distribution (BitTorrent)

We are piloting BitTorrent to maximize infrastructure resilience. By leveraging HTTP Web Seeding (BEP 19), torrents will be seeded simultaneously by Source Cooperative, Zenodo, Data for Canada infrastructure, and community peers. This ensures high availability without a single point of failure. Current laboratory work is available in the Decentralized Distribution Labs.

Work in the Lab: Smart Nodes

To further democratize access and ensure the persistence of Canada’s open data, we are experimenting with the features defined in previous work done by Academic Torrents.

A Smart Node functions as a "set-it-and-forget-it" volunteer server, an automated library branch for our data infrastructure.

• Automated Mirroring: Unlike a standard download, a Smart Node automatically synchronizes with our central FAIR Data Catalog. It intelligently fetches new or "at-risk" datasets to ensure they remain available even if the central portal experiences downtime.
• Volunteer-Powered Resilience: This model allows partner institutions (ex. universities, research labs) and public volunteers to donate bandwidth and storage. By running a Smart Node, contributors actively protect vital Canadian datasets from being lost or gated.
• Dynamic Storage Management: The node software monitors network health to optimize resource usage. Leveraging BitTorrent's capability for selective piece mapping, the node does not need to store the entire catalog. Instead, it identifies specific file indices or "rare" pieces within the metadata and sends granular REQUEST messages for only those blocks. This allows a node with limited storage (ex. 500GB) to provide critical redundancy for a much larger archive (ex. 50TB) by surgically targeting only the data that is currently under shared.

We are currently refining the concepts from smart-node-transmission to work seamlessly with our catalog, enabling a fully decentralized data mesh for Canadian geospatial information.

graph TD
    %% Node Definitions with custom labels
    Catalog[("FAIR Data Catalog")]
    SmartNode["Volunteer Smart Node<br/>(Limited Storage Optimization)"]
    BTNetwork(["BitTorrent Peer Network<br/>(Massive Data Pool)"])

    %% The Process Flow
    Catalog -->|"1. Syncs metadata & identifies 'at-risk' data"| SmartNode
    
    note["Note: The Node does NOT<br/>download the whole file."]
    SmartNode -.- note

    SmartNode -->|"2. Sends granular REQUESTs for specific pieces only<br/>(e.g., 'Send Piece #804 of Dataset B')"| BTNetwork
    
    BTNetwork -.->|"3. Transfers ONLY the requested blocks"| SmartNode

    %% Optional Styling for visual clarity
    classDef central fill:#e1f5fe,stroke:#0277bd,stroke-width:2px;
    classDef node fill:#fff9c4,stroke:#fbc02d,stroke-width:2px,stroke-dasharray: 5 5;
    classDef network fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px;
    
    class Catalog central;
    class SmartNode node;
    class BTNetwork network;