What is the UNIFIED NAMESPACE?
@Kelly Watt shared this awesome video describing the UNIFIED NAMESPACE I thought @LMtx @dougsandy_92230 and others working on Industrial IoT architectures would be interested in.
https://www.youtube.com/watch?v=PB_9HIgSCWc
I envision this being hosted in a type of edge-cloud, no? Love to hear your perspectives..
4.0 Solutions
YouTube
What is the UNIFIED NAMESPACE?
In this new video series, Walker Reynolds explains what is unified Namespace looks like and why it matters in the world of automation.
Watch this video next! Unified Namespace Q&A
https://youtu.be/IiUZTSGjCQI
Thanks for watching!
Subscribe!
👉 https://bit.ly/SubTo40Solutions
LinkedIn 👉 https://bit.ly/40SolutionsLinkedIn
Twitter 👉 https://bit.l...
14 Replies
This is important stuff to anyone working in the Industrial IoT. Challenges there are very different than "enterprise" IoT. In the industrial world, every system is unique with massive numbers of endpoints from many different mfg's, from many different decades, and near zero replication. This is very different than an enterprise IoT solution (let's say, smart retail in a Convenience store. In the C-store, you may have 10-100 endpoints. Once you do your first integration, you "simply" replicate it across thousands of identical C-stores. Both may result in similar number of endpoints and data scale, but two vastly different engineering challenges.
Thanks Mike. I'm trying to understand how and where this would be deployed.... like an edge server?
Not that it's that important I'm just trying to conceptualize it in my current mental "IIoT hierarchy".
like, an edge cloud...?
Likely, "depends." In a single site mfg facility, it could be an on prem server that all the applications could subscribe to for their various data needs. In multisite scenarios, would likely be up another layer (cloud, or corporate data center for example). The real key is breaking away from the confines of the traditional "ISA 95 pyramid" where data can only flow in a point to point/linear fashion (from sensor into a PLC modbus register, from PLC up to SCADA, etc) with zero semantics/context and instead getting that data along with its context into a data broker that feeds all the application layers.
So the question here becomes ---- if you have to go to the cloud, are some of the IoT/industry 4.0 use cases restricted by latency?
"Don't worry. TSN will save us!"
- Former Employee 6866
Here get up to speed here https://litmus.io/
You can full a functional UNS with this tool out of the box with loads of existing drivers.
Absolutely! Shifting away from the traditional ISA 95 pyramid and embracing a more flexible, context-aware data flow allows for a more efficient and integrated approach. A centralized data broker supporting multiple sites can enhance collaboration and streamline processes across the entire manufacturing ecosystem. It's a step towards a more interconnected and data-driven industrial landscape.
Yes, a fair question. But, it can also be a hybrid such that data consumed "in-motion" with latency and determinism requirements stays local, while supervisory type data is moved further up the stack. Much of that "in-motion" data becomes obsolete within seconds so no need to move it further upstream.
Hybrid as in two UNS (one for in motion and another for at rest data) or hybrid deployment architecture (as in different levels of the hierarchy)
?
Probably no single answer to that one. In my mind, it turns into a bit of a hierarchy pretty quickly based on who needs the data, what levels of latency and determinism is required and the useful lifespan of the data is. A supervisory layer could subscribe to a subset of the data from edge layers. And I would envision quite different needs for different environments (a process plant vs. electrical utility vs. a discrete mfg. All that said, the picture is a bit clearer if we keep a clean demarcation point of the UNS sitting above the control layer such that it's not dealing with "real time." That's how I typically see it depicted. I can quickly blur that line, however, when I think of processes where the definition of low latency and determinism may be single-digit seconds instead of the more typical 100mS or less.
@Kelly Watt do you have any perspectives on the architecture? Looking at the video you posted above it seems like Litmus is approaching a UNS-style deployment with "middleware", which would allow for somewhat fluid deployments as @InsightMike suggests.
@LMtx appears to be an AWS Greengrass analog... agree?
It looks like low/no code AWS Greengrass. For sure, that approach addresses a need in the market.
I don’t agree as there is value for much of this data upstream
Litmus is great at the edge and has very strong connectors and can build a UNS of sorts but you would need a broker like HiveMQ for cloud communication and other tools.
Depends on your source data which include business systems method telemetry in the digital twin world
Certainly, Litmus excels at edge computing, showcasing robust connectors and the capability to construct a Unified Namespace (UNS). However, for cloud communication and additional functionalities, integrating a broker like HiveMQ becomes essential. The effectiveness of these tools depends significantly on the nature of your source data, encompassing telemetry from business systems and methods crucial in the context of the digital twin paradigm. The choice of tools is intricately tied to the specific requirements and intricacies of your data landscape. How do you envision leveraging these tools within your specific use case or project?