When talking about IIOT, the common use-case to consider is getting data out of devices. There are use-cases where machines or people are controlling other machines with commands, which I believe is equally IIOT. How does sparkplug/mqtt handle giving a machine inputs or calling methods on a machine? Also, I think there is some confusion on polling vs subscribing to nodes vs Part14 Pub/Sub. I don't know anyone that would use polling in OPC-UA for all the good reasons you listed. The typical way to get data over OPC-UA is to subscribe to a node (including specifying (1) how often the server should internally sample the internal data source, (2) how often you want to be notified of changes, (3) how many queued changed values to receive, and (4) filters on the data (only send changes to 1 decimal place instead of changes in there 5th decimal place, for example)) Then Pub/Sub is following the mqtt model (broker in the middle of two devices). Thanks for the video.

Thanks for the response to the comments!

I think it is a mistake to claim that OPC UA is not report by exception. Even the classic OPC DA supported report by exception, meaning only when the data was changed in the server it was sent to the client that had subscribed the data. No polling required by the client.

I am trying to understand OPC UA for a masters assignment about industry 4.0. The tittle of the video states: "What is OPC-UA Pub Sub?" This video is more like why MQTT is better than OPC UE according to Walker.

Hallo walker, i so love your teaching.

I just try to understand OPC UA Pub/Sub and what I have learned is that you can use the MQTT retain function to send ONCE the opc ua structure by change and afterwards only the pure data in the payload of MQTT... Possibly I am wrong with this conclusion or why Walker do not mention this? I don't know Sparkplug B standard but I guess using the MQTT retrain functionality dramatically lower the payload of OPC UA pub/sub (using MQTT as transportation layer). Could somebody validate this?

Caution with Pub Sub: it still isn't a real event mechanism. What it actually does is it moves the polling mechanism from the client to the server eliminating the redundant communication required to poll the server (at least with Beckhoff PLCs it does). If the monitored value in the server changes more than once within the polling interval, all intermediate values will be missed. Worst case scenario: suppose you monitor a boolean, lets say a one-shot spike generated by a rising edge of some signal. You will simply miss it if the polling frequency is lower than the frequency of the task that generates the one shot.

Great video as always, but i have something to add, i think that it's not always mqtt protocol that we use for messages in opc ua standards, i myself use AMQ ! could you please share a video about it so others may benefits from it :) thanks a lot

I really like the way you talk :D What is the purpose of a OPC/UA Server then, at all? I guess decyphering opc/ua messages from clients that were transported from using mqtt (or amqp which i never got in touch with), saving them (database-style) and making them accessible (readable, filterable, something like that)? Edit: If OPC/UA is just a format standardization, why do there exist opc/ua server (/client) implementations? I can't quite wrap my head around it, seems bogus to me, going to need to investigate quite a bit further.

I came here searching Pub/Sub (Title says What is OPC UA Pub/Sub). But 90% of the video was not about it and it seems more of a propaganda video.