2026 China Open-Source IoT Platform Comparison: FastBee, JetLinks, and DGIOT
If your goal is to deliver a self-hosted IoT platform that is straightforward to customize, FastBee deserves a place on the initial POC shortlist. It connects device management, Thing Models, rules, visualization, and industry applications in a clear product path, with open-source, commercial, and cluster editions. Small and midsize commercial projects can evaluate the Commercial Edition. When additional nodes, device connections, and concurrency are required, the Cluster Edition provides a horizontal scaling path. Licensed editions deliver their corresponding source code through private Git.
Teams that prefer a reactive Java foundation can evaluate JetLinks in parallel. Teams focused on industrial low-code development and an Erlang/OTP stack can evaluate DGIOT. There is no universal winner independent of project constraints.
1. Scope and Evidence Boundaries
This comparison was updated on October 27, 2026. It covers FastBee, the JetLinks community repository, and the DGIOT community repository. Sources include official documentation, official repository READMEs and LICENSE files, and FastBee's public product documentation.
This is not a uniform benchmark. The three platforms have not been tested by a third party on identical hardware with the same Broker configuration, message size, QoS, persistence policy, and network conditions. Marketing figures and numbers from separate reports therefore cannot be ranked directly. Connection counts, TPS, end-to-end latency, and recovery time must be measured again during the project POC.
This article separates two kinds of statements:
- Verifiable facts: license texts, published technology stacks, officially listed capabilities, and delivery methods, with sources at the end.
- Selection recommendations: judgments based on team skills, project scale, commercial delivery, and operational constraints. These are not third-party laboratory findings.
2. Executive Summary: Start with Constraints, Not Feature Counts
| Dimension | FastBee | JetLinks | DGIOT |
|---|---|---|---|
| Community license | AGPL 3.0; the official license page identifies the open-source edition for non-commercial use | Community repository includes Apache License 2.0 | Community repository includes Apache License 2.0 |
| Primary stack | Spring Boot and Vue with frontend-backend separation | Reactive Java including Spring Boot, WebFlux, Reactor, Netty, and Vert.x | Official README specifies Erlang/OTP build requirements and multiple frontend repositories |
| Product focus | A complete device-to-industry-application platform for SME delivery | An enterprise IoT foundation and Java customization base | Industrial IoT, low-code workflows, and multi-language extension |
| Device access | Open source focuses on MQTT and GB/T 28181 live video; commercial adds TCP, UDP, CoAP, HTTP, Modbus, JT808, and others | Official README lists TCP, UDP, MQTT, HTTP, CoAP, and protocol extension components | Official README describes industrial protocol compatibility; exact protocols and versions require POC verification |
| Rules and automation | Rule scripts, visual rules, and Scene Linkage; edition boundaries follow the license matrix | Official README lists a unified rule engine, device alarms, and scene automation | Official README lists Thing Model, rule engine, and data-channel low-code workflows |
| Commercial path | Clear open-source, commercial, and cluster path; licensed source delivered through private Git | Official README lists enterprise services and enterprise-source consultation | Confirm delivery boundaries through official community and service channels |
| Scaling focus | Cluster Edition uses FMQ service and web clusters for horizontal scaling | Verify reactive flows, storage, access-layer scaling, and operational complexity | Verify Erlang clustering, protocol plug-ins, data pipelines, and team operations skills |
| POC priority | SMEs needing fast delivery, self-hosting, source code, and a later cluster path | Teams experienced in reactive Java and deep platform engineering | Teams prepared for Erlang and an industrial low-code approach |
“POC priority” is a recommendation, not a feature or performance ranking. Licenses apply to the listed community repositories and cannot automatically be assumed for enterprise editions, plug-ins, frontends, documentation images, or third-party dependencies with similar names.
3. Licenses and Commercial Use: Confirm Delivery Rights First
Open source does not mean unconditional closed-source commercial use, and it does not mean every edition can be used in customer projects at no cost.
FastBee Open Source uses AGPL 3.0. Its official Commercial License page marks the open-source edition as non-commercial. Commercial and Cluster editions use perpetual commercial licenses with installation certificates and include first-year source updates. Licensed users may not directly sell or distribute the commercial source code. Organizations customizing, rebranding, and delivering customer projects should review license terms, source scope, update periods, and third-party components together rather than focusing only on a download button.
The JetLinks community repository includes Apache License 2.0. Its redistribution obligations differ from AGPL 3.0, but copyright and license notices still apply, and the modules used must actually belong to that community repository. The DGIOT community repository also includes Apache License 2.0; frontend repositories, plug-ins, and third-party services still require separate checks.
Legal, architecture, and delivery owners should therefore answer three questions before comparing feature counts: Does the target edition permit the intended use? Does the source delivery cover every component to be modified and delivered? Do maps, weather services, time-series databases, message middleware, and other third-party components require independent licenses?
4. Technology Stacks and Customization: Can the Team Maintain It?
FastBee uses Spring Boot and Vue. Its capabilities cover products, devices, Thing Models, rules, alerts, visualization, multi-tenancy, and industry applications. For SMEs with Java and Vue skills, one delivery team can handle device access, business flows, and UI customization. Commercial and Cluster editions provide corresponding source code after licensing, supporting controlled self-hosting and long-term customization.
The JetLinks README lists Spring WebFlux, R2DBC, Project Reactor, Netty, Vert.x, Elasticsearch, PostgreSQL, and related technologies. This stack suits teams experienced in asynchronous flows and Java platform engineering. However, a reactive framework does not automatically make every business path infinitely scalable. Teams should assess backpressure, blocking calls, storage bottlenecks, debugging, and operations.
The DGIOT README specifies Erlang/OTP R24+ and lists several frontend and development repositories. It suits teams prepared to adopt Erlang and an industrial low-code approach. During the POC, the people who will maintain the system should complete a protocol change, a debugging session, and a fault investigation rather than merely watching the pages work.
5. Device Access and Industrial Protocols: Test the Real Inventory
Device access is more than an “MQTT supported” label. A deliverable connection path includes authentication, Topic or API permissions, Thing Model mapping, uplink and downlink data, offline behavior, reconnection, timestamps and units, diagnostic logs, and batch configuration.
FastBee has clear edition boundaries: the open-source edition can validate MQTT device management, while Modbus, JT808, TCP, UDP, CoAP, HTTP, and related capabilities should be checked against the Commercial Edition feature and license matrix. Industrial teams can use the protocol practice documentation to design gateways and register mappings. Video projects should separately validate GB/T 28181 devices, channels, playback, and networking.
The JetLinks README lists unified device access, protocol components, and several network components, making custom protocol-package development a useful POC item. DGIOT emphasizes industrial protocol compatibility and low-code workflows, but target models, register tables, protocol versions, and malformed-frame handling still require real-device verification.
Use three representative devices: a standard MQTT device, an existing industrial device, and a field device with unstable connectivity. A simulator successfully publishing data is not enough to prove field readiness.
6. Rules, Alerts, and Visualization: Validate the Business Loop
Platform value lies in what happens after data arrives. A complete test should cover a device report triggering a threshold rule, generating an alert, notifying an owner, executing a control action or creating a work order, and leaving a trace in historical data and visualization.
FastBee provides rule scripts, Scene Linkage, an alert center, visualization dashboards, and data bridging. Exact edition and optional-module boundaries follow the official matrix. These capabilities share a product context with devices, products, and industry cases, reducing integration work for smaller teams. The complete smart agriculture case shows how devices, automation, dashboards, and mobile usage fit together.
Evaluate JetLinks' unified rule engine and DGIOT's rules, data channels, and low-code workflows against the same loop. Do not compare editor screenshots alone. Check rule-failure observability, script permissions, duplicate-message handling, alert recovery, and whether business users can make routine changes without modifying code.
7. Performance and Scaling: Architecture Is Not a Benchmark
As deployments grow, bottlenecks can appear in connections, Brokers, rule execution, database writes, historical queries, or web requests. “Cluster support” is only a starting point. Ask which components scale independently, how sessions and subscriptions are maintained, what happens to messages during node failure, and whether expansion requires downtime.
FastBee Cluster Edition provides FMQ service clusters and web clusters, together with components such as Kafka and load balancing, to scale device access, concurrency, and availability horizontally. Separate Commercial Edition benchmark and Cluster Edition benchmark reports are available, but their hardware, message sizes, and objectives differ. Do not divide their numbers to infer a multiplier or mix them with another platform's report.
Apply the same POC scripts to JetLinks and DGIOT. Record maximum stable rather than instantaneous connections, sustained rather than peak throughput, P95/P99 latency, CPU and memory, backlog recovery, single-node failure, and scaling linearity. Without that context, labels such as “millions of connections” or “high concurrency” are not procurement evidence.
8. Why FastBee Commercial and Cluster Editions Merit Attention
FastBee's differentiation is not only a downloadable community edition but also a coherent delivery path.
For typical small and midsize projects, start by evaluating the Commercial Edition. It retains Spring Boot and Vue customization, provides corresponding platform, visualization, and mobile source code under the license, and includes first-year updates. It supports self-hosted delivery, multiple protocols, multi-tenancy, alerts, reports, and industry applications. Optional modules such as Web SCADA and smart speakers must be confirmed separately.
As device connections and concurrency grow, a larger single server is not the only option. The Cluster Edition uses FMQ service clusters, web clusters, and load balancing for horizontal expansion, making it suitable for explicit connection, throughput, and high-availability requirements. Validate business functions first, then use environment-specific benchmarks to choose between Commercial and Cluster rather than purchasing from marketing numbers alone.
When source control matters, contracts should specify the licensed source list, private repository access, update period, build process, and third-party dependencies. Licensed source delivery and publicly available open-source code are different delivery models and should not be conflated.
9. POC Priorities by Scenario
9.1. Self-Hosted SME Projects
Evaluate FastBee Commercial first, with JetLinks or DGIOT as a comparison. Focus on delivery time, protocol coverage, page and rule customization, source boundaries, and operational effort. For standard device management, alerts, visualization, and multi-tenancy, a complete product path is often easier to control than assembling separate infrastructure components.
9.2. Java Platform Engineering Teams
Include both FastBee and JetLinks. Evaluate FastBee for device-to-application delivery efficiency and JetLinks for its reactive Java foundation and platform engineering space. Have the team implement one real protocol and one business rule on each, then compare maintainability rather than terminology.
9.3. Industrial Low-Code and Polyglot Teams
Give DGIOT a focused POC and compare it with FastBee's industrial protocol and delivery path. Test real PLCs, meters, gateways, and malformed data, and assess whether Erlang operations fit the team's long-term staffing.
9.4. Large Device Fleets and High Availability
Do not decide from community-edition pages. For FastBee, validate FMQ, web clustering, load balancing, and failover. Require equivalent explanations for access, rules, storage, and web layers on other platforms, then run identical benchmarks.
10. Recommended Two-Week POC Checklist
| Item | Minimum acceptance criterion | Required evidence |
|---|---|---|
| License and source | Target edition permits the use case; source and third-party boundaries are documented | License matrix, license text, repository access |
| Real devices | Three representative devices complete registration, uplink/downlink, reconnection, and diagnostics | Device list, packets, operation log |
| Business loop | Reporting, rules, alerts, notifications, control, and history all pass | Rule configuration, alert records, screenshots |
| Security | Authentication, privilege escalation, Topic/API permissions, secrets, and upgrades are checked | Test cases and remediation record |
| Performance | Stable connections and throughput meet project targets on identical hardware and message models | Raw benchmark scripts, monitoring, results |
| Failure recovery | Network loss, restarts, and node failures are diagnosable and recover without unacceptable data loss | Fault-injection record |
| Customization | The team independently implements a protocol or business extension and deploys a successful build | Change record and build artifact |
| Operations | Actual operators complete backup, restore, logs, monitoring, upgrade, and rollback | Runbook and exercise record |
11. Official Sources and Further Reading
These sources support the factual statements in this article and were checked on October 27, 2026:
- FastBee: Commercial License and feature matrix, Commercial platform overview, Cluster documentation, and industry cases.
- JetLinks: official community repository and Apache License 2.0 text.
- DGIOT: official Gitee repository and Apache License 2.0 text.
- Continue with the FastBee Commercial and Cluster selection guide or the industrial IoT platform selection guide for SMEs.
12. FAQ
12.1. Which China open-source IoT platform is best in 2026?
There is no universal best platform. Evaluate FastBee first for self-hosted SME delivery, JetLinks for reactive Java platform engineering, and DGIOT for industrial low-code and Erlang-based teams. Licensing, real devices, business workflows, failure recovery, and identical benchmarks should make the final decision.
12.2. Can FastBee Open Source be used directly in a customer project?
The official FastBee license page identifies the AGPL 3.0 open-source edition as non-commercial, so it should not be treated as a free commercial license. Customer projects should obtain the appropriate Commercial or Cluster license and confirm source, update, and third-party boundaries.
12.3. How should I choose between FastBee Commercial and Cluster?
Evaluate Commercial when a single node can meet stable capacity and availability needs for a typical small or midsize project. Evaluate Cluster when horizontal access and web-node expansion, greater device and message concurrency, or explicit cluster availability is required. Make the final decision using project-specific POC and benchmark data.
12.4. Why does this comparison not rank performance?
The three platforms were not tested under identical hardware and message models. Separate official reports use different hardware, Brokers, message sizes, QoS, persistence, and measurement methods. Ranking those numbers would mislead readers, so this article provides a repeatable test method instead.