Skip to main content
Version: v3.9.0

Core Concepts

The enterprise edge has become a business-critical frontier. Today, up to 75% of enterprise data is created outside the data centers. It is generated across factories, stores, remote sites, vehicles, and infrastructure. The scale is massive.

At the same time, edge environments remain resource-constrained by design. They operate in real-world conditions, with limited connectivity and tight operational margins. Yet the hardware landscape is changing fast. New generations of edge CPUs and GPUs are now powerful, affordable, and widely available. This makes real-time analytics, Edge AI, and autonomous decision-making practical at the edge.

Traditional centralized cloud models cannot keep up. Latency is dictated by physics. Costs grow with every round trip. When connectivity is degraded, remote sites lose autonomy. That creates risk, complexity, and operational friction for distributed deployments.

Datasance PoT is an enterprise-grade unified edge orchestration platform designed for large-scale distributed edge environments. It enables organizations to manage, secure, and operate thousands of applications and software workloads across heterogeneous Edgelet nodes, from a single, consistent operating model.

Datasance PoT brings cloud-native principles to the edge without ignoring real-world constraints. It is secure, vendor-agnostic, and open by design. By reducing operational complexity and restoring local autonomy, it helps enterprises regain control of their edge strategy.

Enterprise edge challenges​

Edge computing runs some or all processing and storage at the wide-area edge instead of only in a central data center or cloud.

Running IT at the enterprise edge, in standalone or hybrid mode, is fundamentally different from running it in the cloud. Edge solutions must embrace the constraints of real sites, real hardware, and unreliable networks.

Scale and diversity​

Enterprises rarely operate at the edge in only a few locations. They operate in tens, hundreds, or thousands of sites: factories, retail stores, energy assets, vehicles, and remote infrastructure. Edge environments combine different CPU architectures, GPUs, operating systems, and hardware generations. Point solutions and custom scripts do not scale and become fragile quickly.

Increasing applications and changing requirements​

Enterprise applications need more real-time or near-real-time processing. Workloads are more latency-sensitive. Unstable upstream bandwidth, high egress cost, and lack of redundancy push critical processing closer to the edge and require more autonomous operation when the WAN is down.

Security and management​

Edge devices are often physically exposed and untrusted. Securing them at scale is hard. Traffic between headquarters and edge sites, and between edge sites, must be encrypted. Distributed edge platforms need built-in trust, automation, and verifiable control. Lifecycle management (onboarding, updates, patching, monitoring) is costly without a unified control plane.

Cost​

Digitalization increases infrastructure spend. Siloed edge stacks duplicate capex. Manual operations and parallel toolchains inflate opex. A single orchestration layer reduces redundant hardware and operational toil.

Edge AI and fleet assets​

AI is moving to the edge, but models and retrieval corpora are large, versioned, and often subject to air-gap or strict egress policies. Teams need fleet-wide control over which model revision and which knowledge bundle runs where, without copying artifacts by hand on every device. Governance matters: RBAC should govern who can deploy or attach AI assets to production Edgelet nodes. Binding models and knowledge in microservice specs keeps application manifests portable while the platform handles distribution.

Introducing Datasance PoT​

Datasance PoT onboards edge sites as Edgelet nodes. Edgelet is the node agent, the application runtime, and, on Linux, the container engine in one binary. The Linux OTA download is about 30 MB and already contains the engine, so a host does not need Docker or Podman. Linux builds cover amd64, arm64, arm (32-bit), and riscv64. The same binary can run standalone from the edgelet CLI, then join a cluster with potctl.

Installation and upgrades go through potctl and the EdgeOps Console. The platform picture is on Architecture. The engine itself is Get to know Edgelet.

Key capabilities​

Datasance PoT provides the capabilities required to run and operate software workloads across distributed edge environments. It focuses on the full lifecycle of applications, services, and edge operations with unified control at scale.

Edge environment and Edgelet nodes​

Datasance PoT simplifies how edge environments are onboarded. Edgelet nodes use the Edgelet runtime to establish secure identity and connectivity to the Control Plane. Installation and upgrades are automated through potctl and the EdgeOps Console, without deep per-site customization.

The platform provides an open operational layer above hardware and operating systems. You manage diverse environments as one cluster while preserving local isolation.

Application and workload lifecycle orchestration​

Datasance PoT orchestrates the complete lifecycle of containerized workloads at the edge. Applications are deployed, updated, versioned, and retired consistently using declarative YAML manifests.

Environmental differences are abstracted so teams focus on what should run and where, not on per-site wiring. That enables repeatable deployments, controlled rollouts, and safe updates across edge clusters.

Configuration, secrets, and policy management​

Configuration values and secrets are defined centrally and distributed securely to edge workloads at runtime. Role-based access control (RBAC) ensures users, automation, and workloads stay within intended scope.

AI models and knowledge assets​

Platform train v3.9.0 adds fleet catalogs for Edge AI assets using the same declarative model as applications and registries.

  • Model catalog (kind: Model): LLM and inference weights. Sources include Hugging Face Hub models and OCI artifacts (model-spec, ModelPack, ModelKit, and ORAS-compatible blobs) through registry type hf or oci.
  • Knowledge catalog (kind: Knowledge): RAG corpora, datasets, and document bundles. Hugging Face datasets and OCI bundles use the same registry types. Weights stay in Model; knowledge uses a separate namespace.

Typical lifecycle:

  1. Define Model or Knowledge YAML and apply it to the Controller catalog.
  2. Attach assets to one or more Edgelet nodes so the platform pulls and tracks the right revision.
  3. Bind catalog names in microservice spec.models and spec.knowledge so running workloads use fleet-managed content, not host paths.

EdgeOps Console exposes Config → AI Model Catalog and Config → AI Knowledge Catalog. Operators can also use potctl deploy, attach model, attach knowledge, and matching describe commands.

Learn more: AI Model Catalog, AI Knowledge Catalog, Microservices - models and knowledge, and the Edge AI wafer defect tutorial.

Secure connectivity​

Datasance PoT establishes trust between the Control Plane and edge devices. Each Edgelet node and workload uses verified identity with built-in certificate management.

A secure service mesh (Router, Skupper-based) and distributed message bus (NATS) are provided by default. Services communicate with encryption across nodes and sites while edge networks can remain isolated from a routing perspective.

Observability and operations​

Teams observe workload status, connectivity, and health across sites from one operational view. potctl supports automation-first workflows. The EdgeOps Console supports day-two operations, including v3.9 Configure flows for AI catalogs and cluster config.

Open-source foundation with enterprise support​

Datasance PoT is built on an open-source foundation shared with Eclipse ioFog, with transparency and extensibility. Datasance complements that foundation with enterprise-grade support for mission-critical deployments.

Ready for more?
Group 3See anything wrong with the document? Help us improve it!

Edit this page on GitHub!