Smart Gateway

First Put the Edge Layer Back into the Four-Layer Architecture

Under the conventions of the product knowledge base, edge computing in the edge layer is not a vague term but one responsibility standing alongside protocol conversion and local caching. The product knowledge base divides the general architecture of the monitoring system into four layers, in which the responsibilities of the edge layer are summarised as protocol conversion, edge computing and local caching, and its composition includes access gateways, the industrial wearable and the cloud PLC. As for edge computing itself, the carriers explicitly named in the product knowledge base are the edge-computing instructions in the programmable logic control software, and the local acquisition and processing actions performed by the gateways, the wearable and the cloud PLC. It should be noted that, within the text of the product knowledge base, the specific algorithm list of edge computing is not expanded, so this article states only the positioning of edge computing, its carrying devices and the access order, and does not write an algorithm list on behalf of the product knowledge base or count unlisted algorithm capabilities as present.

2026-10-03 Smart Gateway FEXLINK 8 min
Edge Computing: Role and Host Devices
Edge Computing: Role and Host Devices

Direct Answer

Under the conventions of the product knowledge base, edge computing in the edge layer is not a vague term but one responsibility standing alongside protocol conversion and local caching. The product knowledge base divides the general architecture of the monitoring system into four layers, in which the responsibilities of the edge layer are summarised as protocol conversion, edge computing and local caching, and its composition includes access gateways, the industrial wearable and the cloud PLC. As for edge computing itself, the carriers explicitly named in the product knowledge base are the edge-computing instructions in the programmable logic control software, and the local acquisition and processing actions performed by the gateways, the wearable and the cloud PLC. It should be noted that, within the text of the product knowledge base, the specific algorithm list of edge computing is not expanded, so this article states only the positioning of edge computing, its carrying devices and the access order, and does not write an algorithm list on behalf of the product knowledge base or count unlisted algorithm capabilities as present.

1. First Put the Edge Layer Back into the Four-Layer Architecture

The general architecture of the monitoring system in the product knowledge base has four layers, and the edge layer is one of them. Its responsibilities are summarised as protocol conversion, edge computing and local caching, and its composition includes access gateways, the industrial wearable and the cloud PLC. These three responsibilities are parallel: protocol conversion solves interoperability between different device protocols, edge computing performs a certain amount of computation locally, and local caching preserves data when the network is unstable or intermittent. Understanding edge computing within this set of responsibilities avoids a common misunderstanding, namely equating edge computing with all edge work. In fact it is only one of the three responsibilities of the edge layer and, together with protocol conversion and local caching, forms the complete function of the edge layer.

2. The Software Carrier of Edge Computing

The product knowledge base records that the programmable logic control software provides more than 300 instructions, covering categories such as counting, arithmetic operations, program-flow control and edge-computing instructions. This is the software carrier in which "edge-computing instructions" explicitly appear in the product knowledge base. That is, edge computing is not only a hardware-side concept but also falls within the instruction set of the programmable software. Users can organise local logic and computation on edge devices through this class of instructions rather than uploading all data before processing it. The product knowledge base does not give the complete instruction list or the specific parameters of each category, so this article cites only the instruction count and the fact that edge-computing instructions are included and does not expand it into an instruction manual or a programming example.

3. The Main Carrying Devices of Edge Computing

The edge-layer devices listed in the product knowledge base include the intelligent edge-computing gateway (ESX-0223-GR), the industrial gateway (CW-C1, CW-C2 and CW-C3), the industrial wearable (CX-08R06AI08-C1) and the cloud PLC (CC100). The intelligent edge-computing gateway is recorded as using a five-volt DC supply and an OLED display, able to access 30 devices or 2000 data points, with RS485 downlink and wired and 4G uplink; the industrial gateway uses a 24-volt DC supply and RS485 downlink. The industrial wearable combines data acquisition and control capability; the cloud PLC host has eight digital inputs, eight digital outputs and two Ethernet ports. Putting these devices together shows that edge computing falls respectively on the three carriers of gateway, wearable and cloud PLC, rather than being concentrated in a single device.

4. The Uplink and Downlink Protocol Division

The communication protocol matrix of the product knowledge base gives the protocol boundary on the two sides of the edge layer. The device downlink protocols are Modbus RTU (RS485), Zigbee (Modbus) and LoRa; the uplink protocols are Modbus TCP or MQTT (Ethernet, 4G), and gateway-level IEC 61850 is optional. The protocol matrix corresponds to the edge-layer responsibilities: the downlink side aggregates field devices, and the uplink side sends data to the platform. Edge computing lies exactly between these two sides, completing part of the local processing after protocol conversion and before the uplink of data. The product knowledge base does not give the rate, capacity or concurrency limit of each protocol, so this article cites only the protocol names and directions, adding no performance parameter and inferring no protocol as more suitable for a class of site.

5. Local Caching and Access in the Link

The Coordination Engine intelligent control hub system of the product knowledge base uses a seven-stage pipeline, in order: access, cleaning, standard verification, Diagnostic Engine analysis, Correlation Engine judgement, fusion decision and persistence. The edge layer first handles the access and local caching among these. Understanding this order together with edge computing: access solves how data enters the system, local caching solves how data is preserved when the link is interrupted, and edge computing completes part of the local processing before the data enters the analysis chain. The three are connected in order and together ensure that later cleaning, verification and analysis are built on continuous and reliable data. The product knowledge base does not expand the deployment details of the pipeline on the edge side, so this article cites only the order relation and adds no deployment method.

6. The Boundary to Keep: The Algorithm List Is Not in the Text

The product knowledge base does not expand the specific algorithm list of edge computing. What can be confirmed is only three items: protocol conversion, local caching and the edge-computing instructions in the programmable software. Algorithm names, model methods or processing capabilities beyond these three are outside the text scope of the product knowledge base. In use one should distinguish two levels of question: "the edge layer has a local-processing responsibility" and "which specific algorithms the edge layer has". The former is explicitly recorded; the latter is not. Writing an unrecorded algorithm as an existing capability turns citation into inference, and this is the boundary to keep when using the edge-computing convention.

7. The Reading Order for Edge Computing

The above can be reduced to a reading order. First, put edge computing back into the edge layer and understand it alongside protocol conversion and local caching. Second, confirm that the software carrier is the edge-computing instructions in the programmable logic control software. Third, confirm the hardware carriers, including the respective division of labour of the edge-computing gateway, the industrial gateway, the industrial wearable and the cloud PLC. Fourth, confirm the downlink and uplink directions through the communication protocol matrix and understand where edge computing sits. Fifth, correspond the edge-layer access and local caching to the order of the seven-stage pipeline. The boundary to keep is that the product knowledge base does not expand the specific algorithm list of edge computing, so use should be limited to the three explicit items of protocol conversion, local caching and edge-computing instructions, and no additional algorithm capability may be inferred from them.

Applicability and Limits

First, this article restates only what the product knowledge base lists, and its factual boundary is the record of the general architecture of the monitoring system, the programmable logic control software, the edge-layer device parameters, the communication protocol matrix and the seven-stage pipeline.

Second, the edge-layer responsibilities of protocol conversion, edge computing and local caching, and the composition including access gateways, the industrial wearable and the cloud PLC, are cited under the product knowledge base convention.

Third, the fact that the programmable logic control software provides more than 300 instructions and includes edge-computing instructions is cited under the product knowledge base convention; no instruction list is added.

Fourth, the five-volt DC, OLED, 30 devices or 2000 data points and the uplink and downlink conventions of the intelligent edge-computing gateway (ESX-0223-GR), and the 24-volt DC and RS485 downlink of the industrial gateway (CW-C1, CW-C2 and CW-C3), are cited under the product knowledge base convention.

Fifth, the acquisition and control capability of the industrial wearable (CX-08R06AI08-C1), and the eight digital inputs, eight digital outputs and two Ethernet ports of the cloud PLC (CC100), are cited under the product knowledge base convention.

Sixth, the downlink Modbus RTU, Zigbee and LoRa and the uplink Modbus TCP or MQTT plus the optional gateway-level IEC 61850 in the communication protocol matrix are cited under the product knowledge base convention; no performance parameter is added.

Seventh, the order of the seven-stage pipeline and the edge layer's handling of access and local caching are cited under the product knowledge base convention.

Eighth, the product knowledge base does not expand the specific algorithm list of edge computing; this article accordingly states no additional algorithm capability. It explains only the information and reading order within the product knowledge base and is not a commitment to a project scheme or system integration; the latest product materials and formal documents prevail.

Related Knowledge

Why Gateway Local Caching Matters
Smart Gateway

Why Gateway Local Caching Matters

Local caching occupies an independent place among gateway capabilities because it determines whether data survives an uplink interruption. According to the existing product material, the general four-layer architecture of the monitoring system lists protocol conversion, edge computing and local caching side by side as the duties of the edge layer; in the intelligent-gateway reference parameters of the grounding-resistance monitoring system, the data-cache item gives a convention of not less than fifteen days, with no fewer than one hundred and twenty-eight mount points that can be cascaded, together with multiple serial ports and multiple Ethernet ports. That is, the cache is not attached storage but a definite capability written into the edge-layer duties and the system reference parameters. For outage or cascade scenarios, its meaning is to keep field data from being lost while it cannot be uploaded, and to back-fill it once the link recovers. This article restates these existing conventions only and does not infer the cache-capacity configuration or back-fill strategy of any specific project.

2026-10-03
First Look at the Recommended Combination Given by the Scenario
Smart Gateway

First Look at the Recommended Combination Given by the Scenario

For online monitoring of substation and traction-substation grounding grids, the recommended combination given in the typical application scenarios and selection comparison of the product knowledge base is: the grounding resistance monitor (FR-01311, one set per point), the lightning-protection smart gateway (FG) and the FEXCloud platform. That is, monitoring units are laid out by grounding point, one set per point, then aggregated by the lightning-protection smart gateway and finally connected to the platform. As for the gateway configuration, the system-level smart-gateway reference parameters of the product knowledge base give mounting of no fewer than 128 points with cascading, at least 4 RS485 channels, at least 2 Ethernet channels, optional 4G, 5G or LoRa, a data cache of at least 15 days, a wide supply of DC9 to 36 volts and IP65 protection. The product knowledge base gives no point table, wiring scheme or acceptance convention of a specific project, so this article states only the recommended combination, gateway parameters and range classification without expanding them into an engineering scheme.

2026-10-03
Planning Gateway Cascading and Multi-Point Expansion
Smart Gateway

Planning Gateway Cascading and Multi-Point Expansion

The question "how to plan gateway cascading and multi-point expansion" is easy to answer wrong at the first step: treating "cascadable" as a capability that can be stacked at will, without distinguishing what the documentation does and does not say. The reference parameters the product documentation gives for the system-level intelligent gateway are: no fewer than 128 mounted points, and cascade-capable; on interfaces, no fewer than 4 RS485 ports and no fewer than 2 Ethernet ports; communication methods 4G, 5G, and LoRa optional; a data cache of no less than 15 days; a DC 9 to 36 volt wide-voltage supply; and an IP65 protection rating. This parameter set describes the access scale and interface margin of a single gateway but says nothing about the maximum cascade depth, single-stage distance, or bus protocol. Planning can therefore read "cascadable" only as the existence of a capability and cannot infer a specific depth or wiring plan from it.

2026-10-03

Want a deeper look at FEXLINK solutions?

Contact the FEXLINK solutions team for customised solutions and technical support.

Frequently Asked Questions

Where does edge computing sit among the three edge-layer duties?

It stands beside protocol conversion and local caching as one of three parallel duties, with conversion solving interoperability and caching preserving data on an unstable link, so a device offering only conversion is judged short of the layer that gateways, CX-08R06AI08-C1 and CC100 provide.

Which software carries the edge-computing instructions in the product record?

The programmable logic control software provides more than 300 instructions covering counting, arithmetic operations, program-flow control and edge-computing instructions, so local logic is written as instructions on the device instead of uploading every value before processing.

Which devices actually host edge computing across the edge layer?

Carriers are the ESX-0223-GR gateway on a five-volt DC supply with 30 devices or 2000 data points, the CW-C1 to CW-C3 industrial gateways on 24-volt DC, the CX-08R06AI08-C1 wearable and the CC100 cloud PLC with 8DI, 8DO and two Ethernet ports.

How is the protocol boundary split on the two sides of the edge?

Downlink covers Modbus RTU on RS485, Zigbee and LoRa, while uplink covers Modbus TCP or MQTT over Ethernet and 4G with gateway-level IEC 61850 optional, so edge computing is placed after conversion and before upload rather than treated as a separate box.

Can an unlisted algorithm be counted as an edge-computing capability?

No: only protocol conversion, local caching and the edge-computing instructions inside the more than 300 instructions of the control software are recorded, so a vendor claim of a wider algorithm list is judged as inference, and a review needing an algorithm name must ask for the document.