Subscribe
Share
Share
Embed
Low-code automation has grown up, and the competition is getting spicy. In this episode of Pop Goes the Stack, F5's Lori MacVittie and Joel Moses are joined by Aubrey King as they dig into the heavyweight duel between N8N and FlowFuse—two platforms promising to empower teams to automate anything without waiting for overworked developers. We cut through the marketing fluff and look at the real differences in architecture, deployment models, extensibility, security posture, and operational experience. How do they scale? Who controls your data? And what happens when the automation breaks at 2 a.m.? If you care about automation that doesn’t collapse under real-world pressure, you’ll want to hear this.
Read our F5 research for more on the status of automation in IT: https://www.f5.com/resources/reports/state-of-application-strategy-report
Read our F5 research for more on the status of automation in IT: https://www.f5.com/resources/reports/state-of-application-strategy-report
Creators and Guests
Host
Host
Lori MacVittie
Distinguished Engineer and Chief Evangelist at F5, Lori has more than 25 years of industry experience spanning application development, IT architecture, and network and systems' operation. She co-authored the CADD profile for ANSI NCITS 320-1998 and is a prolific author with books spanning security, cloud, and enterprise architecture.
Guest
What is Pop Goes the Stack?
Explore the evolving world of application delivery and security. Each episode will dive into technologies shaping the future of operations, analyze emerging trends, and discuss the impacts of innovations on the tech stack.
00:00:05:07 - 00:00:22:09
Lori MacVittie
Hi. Welcome back. This is Pop Goes the Stack, where emerging tech trends go to get unpacked, interrogated, and occasionally roasted. I'm your host, Lori MacVittie, who is still waiting for an AI that understands sarcasm.
00:00:22:12 - 00:00:23:26
Joel Moses
Oh, that would be great.
00:00:23:27 - 00:00:52:27
Lori MacVittie
Still waiting. Still waiting. Today on Pop Goes the Stack, we are talking about automation and the fact that it's got choices today. For years, automation meant workflows. Still kind of does, right. Still workflows, they just might be more dynamic, chaotic than we'd like. And there's some tools coming out that are really good at helping to build these kinds of automations.
00:00:52:27 - 00:01:19:27
Lori MacVittie
Things like FlowFuse built on Node-RED. Joel's a fan. And N8N, which is built on something Kubernetes obviously because there's an eight in the name. Right, that's what it tells me.
Joel Moses
It's a good assumption.
Lori MacVittie
It's some kind of thing. Anyway, on the other side is, you know, agents actually making the decisions, building the workflows in real time and doing things.
00:01:19:28 - 00:01:30:08
Lori MacVittie
So you've got two choices. Today we wanted to talk about the low code automation with teeth, I'm told. To do that, we've got Joel Moses as usual. Joel.
00:01:30:09 - 00:01:32:04
Joel Moses
Hello, Lori. Good to be here.
00:01:32:06 - 00:01:59:15
Lori MacVittie
Awesome. And we brought Aubrey King in because he's a fan of N8N and can
Aubrey King
Big time.
Lori MacVittie
cause he's used it. So we've got some experience here. Joel's looked into, you know, all of these. He's got good perspectives too. And we wanted to discuss what they are briefly. And then, you know, pros cons,
Joel Moses
Yeah.
Lori MacVittie
you know when they're appropriate, when you should keep humans in the loop, you know,
Aubrey King
Always.
Lori MacVittie
which fork to take. Always,
Joel Moses
Definitely.
Lori MacVittie
like okay, we're done.
00:01:59:15 - 00:02:01:13
Lori MacVittie
Thanks, everyone.
Joel Moses
Right.
00:02:01:15 - 00:02:27:00
Joel Moses
Well, let me point out really quickly that the reason that Aubrey and I are talking about FlowFuse, which is a commercialized version of Node-RED, which is an open source project, and N8N is that both are essentially open source projects with communities around them. And so, but I want to be very clear about this. They're actually just big, big packages of node.JS code that is glue logic that helps you craft workflows using a visual editor.
00:02:27:04 - 00:02:47:22
Joel Moses
So from that perspective, they're sort of kissin' cousins. Node-RED hews towards the IoT side of the fence and N8N goes towards kind of the IT automation side of the fence. But, they're effectively two sides of the same coin. They're low-code. You know, we used to call these things no-code
Aubrey King
Yeah.
00:02:47:23 - 00:02:50:26
Joel Moses
and then people realized, oh, there's always a little bit of code.
00:02:50:29 - 00:02:52:10
Lori MacVittie
There's actually code.
00:02:52:17 - 00:02:53:22
Joel Moses
A little bit.
00:02:53:24 - 00:02:56:24
Aubrey King
No. No such thing as no-code.
00:02:56:26 - 00:02:59:24
Joel Moses
Yeah. So what do you like about N8N, Aubrey?
00:02:59:27 - 00:03:03:00
Aubrey King
I'm not allowed to say "everything," am I? I can't, actually there
Lori MacVittie
No.
00:03:03:02 - 00:03:05:19
Joel Moses
You can say it. You'll be wrong, but you can say it.
00:03:05:21 - 00:03:27:11
Aubrey King
there are things actually that I don't like about it, but I will kind of get into it a little bit. It makes getting started easy. So for me, what I mean by that is, for me, I had a hard time understanding how maybe, AI applications especially, how they would actually talk to each other within the confines of an infrastructure.
00:03:27:13 - 00:03:51:04
Aubrey King
And, I found that N8N is very, very easy to kind of put together a flow, right, we talked about flows. So you can put together a flow in five minutes that's working in functional calls. Maybe if you have like a machine that's running Ollama somewhere else in your network or if you have a, let's say, an OpenAI subscription that you're going to plug in and use with your N8N also.
00:03:51:04 - 00:04:16:00
Aubrey King
You could very easily use some kind of networking tools like the three of us and anyone probably watching this are used to, to actually, you know, take SSL off and take a look at the messages that are being passed back and forth, what do we have in plain text once that happens, all that kind of good stuff that most people who have been living the F5 life for a long time are going to be thinking about. So,
Joel Moses
Right.
00:04:16:01 - 00:04:22:21
Aubrey King
for me, real easy to get going. That's probably the number one thing I like about N8N.
00:04:22:24 - 00:04:31:14
Lori MacVittie
Why do I smell hard coded credentials somewhere in there? You mentioned, "hey, it could take a a key," and I'm like, where does it store that? I mean,
00:04:31:16 - 00:05:00:08
Aubrey King
It's hashed.
Joel Moses
That's an excellent point, Lori,
Lori MacVittie
Oh, well it's hashed. It's okay then.
Joel Moses
and again part of the risk of these low code tools is that there are certain things that they do to make things easier. They give you convenience. And, you know, convenience is kind of an interesting thing. Convenience is, however, a loan. You pay back that loan using architecture and design and process. And that's where these tools actually fall short.
00:05:00:09 - 00:05:25:00
Joel Moses
We'll get to that in a second. So I need to cover the FlowFuse side of the coin here. So, you know, FlowFuse is a commercial offer that's built on the top of Node-RED and FlowFuse is essentially a package of tools that provides scalability for the Node-RED instances. Node-RED is essentially just kind of a workflow runner that really doesn't have any idea about scaling or enterprise readiness.
00:05:25:03 - 00:05:44:24
Joel Moses
And so this is a package that fits on top of that. Node-RED, of course, has a long history. It's a project that came out of IBM and specifically for IoT workloads. And it's got a tremendous number of plugins and add-ons and things that you can decorate with, which is kind of why I like it.
00:05:44:24 - 00:06:07:00
Joel Moses
It's an extra large grab bag of routines. So if I need to do something with a particular type of technology, or I need to speak a particular type of protocol, it's there. I can prototype with it. Now, my use of it is not necessarily to productionize things. My use of it is to experiment to see if certain things can be glued together in certain ways.
00:06:07:02 - 00:06:29:02
Joel Moses
And for that, it's awesome. But again, it's a package of Node.JS--just like N8N is--and it helps you visually construct and create a workflow that works and you can test things with. And then, once you've got all that down, you got to move it into production somehow.
00:06:29:02 - 00:06:35:23
Joel Moses
And that's where the rubber meets the road. And that's where these systems kind of cease utility.
00:06:35:25 - 00:07:05:12
Aubrey King
That production thing is really what kind of got me about N8N to be honest. It, as my massive scale brain started working, right, cause most of my operations career was all in service provider; most of my SE time was all in service provider as well. So, when I start thinking about service provider networks and trying to scale N8N to that, I have a really hard time understanding that. Especially since N8N doesn't have a, even though the name would indicate otherwise,
00:07:05:17 - 00:07:20:15
Aubrey King
it doesn't really play nice with Kubernetes just yet. Docker 100%.
Lori MacVittie
What?
Aubrey King
Docker all day, but I have not seen any sort of easy to Kubernetify, you know,
Joel Moses
Nice word, I like that.
00:07:20:18 - 00:07:21:16
Lori MacVittie
That's not a word. That's not a word.
00:07:21:18 - 00:07:23:05
Aubrey King
It's got to, it's got to be somewhere.
00:07:23:05 - 00:07:23:18
Joel Moses
It is now.
00:07:23:18 - 00:07:25:09
Lori MacVittie
It's not.
00:07:25:12 - 00:07:43:27
Joel Moses
Yeah, I totally agree. You know, one of the risks here is that these things by default are meant for...so let's talk about its intersection with AI and AI assistants. So a lot of these tools--FlowFuse is one of them--they have an AI assistant that allows you to describe what routine you want built
00:07:43:27 - 00:08:10:16
Joel Moses
and it will build it for you. Now the output of N8N and the output of FlowFuse AI assistants is they show you a visual representation of the workflow. And the reason they show you a visual representation is these tools are primarily targeted to people who don't have any knowledge of actual computer languages.
Aubrey King
Yeah.
Joel Moses
So they need to see a visual representation and it produces what effectively is like a Visio graph of the routine.
00:08:10:18 - 00:08:36:17
Joel Moses
That's fantastic. The trouble is, it forgets a lot of the stuff that programing languages and language runtimes are good at, which is debugging, observability, logging. So, the routine is very visual. It shows you exactly what's going to happen, but it won't show you what happens if something fails.
Lori MacVittie
Well, and it's
Joel Moses
That is left to the person to develop that.
00:08:36:19 - 00:08:44:02
Joel Moses
And these are not targeted to developers. And that's where
Lori MacVittie
It's a disconnect.
Joel Moses
that's where things go wrong.
00:08:44:04 - 00:09:04:01
Lori MacVittie
It's still running code underneath
Joel Moses
It is.
Lori MacVittie
no matter how it's represented. I mean, we've had these kind of tools forever. If you go back to things like business process automation, they had these beautiful Java frameworks and servers and you can
Joel Moses
Oh yeah.
Lori MacVittie
Hey, it was real. I used them.
Joel Moses
Look,
Lori MacVittie
They were horrible. But you could do the mapping thing and it was, right
00:09:04:01 - 00:09:23:06
Lori MacVittie
you could map the flow. You could visually do it, you just had to drag and drop it yourself. Now the AI will do it for you, but there's still code underneath that needs, as you pointed out, it needs to be debugged. It needs to be secured. It needs to be, you know, have logging, all sorts of things that don't sound like they're included.
00:09:23:08 - 00:09:45:02
Joel Moses
That's absolutely true. You know, it's, I need to be very clear here, FlowFuse and N8N are open source variants with commercial offers. There are actual commercial offers for low-code systems like Zapier and IFTTT and Airtable and Zoho, and there's a ton of things in this particular space that are intended for production use in production environments.
00:09:45:02 - 00:10:04:19
Joel Moses
Now, I once saw a Low-code platform demo, and I'm not going to name names as to what the platform was, where someone built an order to invoice pipeline in six minutes. You know, everybody was, you know, applause, standing ovation, there were tears, people were passing out in the aisles, and then someone asked a really key question:
00:10:04:22 - 00:10:29:07
Joel Moses
"well, our ERP systems times out sometimes, what happens when that happens?" And it was like the presenter was a deer in headlights, did not know how to answer that one. And that's when I realized that low-code demos are essentially cooking shows where everything is pre chopped, everything is preheated, the mess is already cleaned up, but in practical sense, lots of other stuff can happen in the kitchen.
00:10:29:07 - 00:10:38:23
Joel Moses
And that's not represented. And that's the problem with these systems. These systems don't necessarily incorporate what happens if things go wrong.
00:10:38:25 - 00:11:01:28
Aubrey King
It's funny
Lori MacVittie
But the concept is good. I mean it's necessary because there are those two approaches, right. One view of automation in enterprise settings is going to be you're going to build a workflow and it may use AI in different ways, or even make calls out to AI to do things, but it is going to drive the automation so that you keep control.
00:11:02:01 - 00:11:22:10
Lori MacVittie
And the other way is 'here, agent, just here's some tools, do this thing' and now it's going to go try and do it on its own. And so these tools are leading toward that first option where some control, you know, benefits a AI, but you still have the control. And the other one is let's see what happens.
00:11:22:16 - 00:11:48:16
Joel Moses
Yeah.
Aubrey King
It's interesting, the timeliness of this conversation or this piece of it anyways, because right before coming to record with you guys, I was working on a flow that is broken and
Joel Moses
Hmmmmm.
Aubrey King
Yeah. And here's the problem, like, where do you go?
Joel Moses
Yeah.
Aubrey King
So, you know, and that's a funny thing for me to say because I work for our community resource,
00:11:48:16 - 00:12:07:07
Aubrey King
right, which is where these problems should be sort of nipped. But my experience with N8N--and I don't know how this would work with FlowFuse, I suspect it would be a support call--but with N8N I put in a "Get Issue" and that Get Issue they took a look at it and triaged it to configuration.
00:12:07:15 - 00:12:25:10
Aubrey King
This is not an error, it's a configuration problem. Go to the support, the community support.
Joel Moses
Right.
Aubrey King
So I go to community support I put exactly the same thing in. Right? Community comes back and they look at this and they go, "Yeah, no. This looks like a problem with the coding of the node. This does not look like a configuration issue."
00:12:25:10 - 00:12:42:27
Aubrey King
Okay, great. So I go back to Get Issues and I say, you know, "Hey, this is what community gave me: it's definitely a problem in the node." And they come back with, "No we used a different model and it's a model problem. So that's the problem. So go back to community and..."
00:12:42:27 - 00:12:43:24
Joel Moses
Ping pong, ping pong. Yeah.
00:12:43:24 - 00:12:48:16
Lori MacVittie
So nothing's changed.
Aubrey King
No.
Joel Moses
Yeah, well
Lori MacVittie
For the user it's the same thing, back and yeah.
00:12:48:22 - 00:13:15:27
Joel Moses
Well realize that these platforms are designed to empower non-technical users, right? And so the community is built up of a predominantly large set of non-technical users who actually don't delve too deeply into how the actual logic works. And so, you know, a platform will empower non-technical users for about three weeks, and then IT gets a ticket that says, "Workflow broken. Node says it doesn't run, don't know why. Please fix ASAP.
00:13:15:27 - 00:13:30:18
Joel Moses
It's mission critical." And then you have to go hunt in the community for exactly why that would be the case. And that's when you find out that your automation pipeline isn't a helpful coworker; it's more like a badly behaving emotional support animal.
00:13:30:20 - 00:13:42:03
Aubrey King
Well and the logs conversation is incredible too, because really the error that I got, ready for it right, "Unauthorized."
00:13:42:06 - 00:13:43:13
Joel Moses
Oh I love those errors. Yeah.
00:13:43:14 - 00:13:46:07
Aubrey King
That's it. Unauthorized. Nothing else.
00:13:46:09 - 00:13:49:00
Lori MacVittie
Okay. What's unauthorized?
00:13:49:03 - 00:13:50:10
Aubrey King
Yeah. Exactly.
00:13:50:12 - 00:13:55:14
Joel Moses
Yeah. Now these load
Lori MacVittie
Well the credentials weren't hard coded; it was the hash. The hash was wrong, Aubrey, see.
00:13:55:15 - 00:13:58:25
Aubrey King
I didn't even get that far. There's no credentials.
00:13:58:27 - 00:14:16:04
Joel Moses
Yeah. Now we're going to leave scaling aside for a second. But one of the problems with these systems is that, effectively, in order to really deepen their functionality, they are extensible in nature. Meaning you can actually put little segments of code in some of these nodes to modify the behavior of the nodes.
00:14:16:04 - 00:14:42:02
Joel Moses
Now, the problem is that these systems want to put you on rails, and every time you modify the default configuration of a node, you get off the rails. And, you know, being able to extend things with custom code sounds empowering until you realize what that actually means, which is that you've written this thing, you've deployed it through a visual editor, it fails, and all it says is "unauthorized" or "node failed."
00:14:42:04 - 00:14:49:20
Joel Moses
And congratulations, you just invented a new way to be on call.
00:14:49:23 - 00:15:17:07
Joel Moses
So that's a big problem with these systems. Again, one of the things that if a company is going to invest in things like Zapier or N8N or FlowFuse that you've got to have is you have to have some sort of oversight, and some form of standards to move workflows into production, ensuring that they actually have the right things to be run scalably and to have them be able to be trouble shot.
00:15:17:10 - 00:15:20:07
Joel Moses
And I don't see that a lot.
00:15:20:10 - 00:15:27:20
Aubrey King
Upgrades sound like a nightmare to me. Upgrade, rollback, upgrade
Joel Moses
Yeah.
Aubrey King
like, unless it's all done in containers. Then obviously that's
Joel Moses
Right.
00:15:27:20 - 00:16:05:01
Lori MacVittie
So, what I'm hearing is these things are great for some workflows, but maybe not production level IT enterprise type flows, right? They already have scripts, right? I mean, we've been tracking the status of automation in IT for years in our research every year and it's pretty high. Most people are either using a script that is automatically executed or that they type in and kick off, but it's a script that already has that flow.
Joel Moses
Yeah.
00:16:05:03 - 00:16:23:21
Lori MacVittie
So the move to AI is more, "hey AI, here are the possible scripts you can use, do this thing." And let it go do that work. Right? I mean, that seems like the natural evolution rather than, hey, here's this tool, build this visual thing, and maybe it'll work. I mean...
00:16:23:23 - 00:16:42:16
Joel Moses
Well, my attitude about these systems is you can actually put these things in production, but if you have some
Aubrey King
Okay.
Joel Moses
and it's, but it's good to have a non-technical user who can design a workflow to get it to work. But if that is the end of the process, you haven't done what's necessary to make it work properly. Right?
00:16:42:16 - 00:17:09:10
Joel Moses
Just getting it to work is one bar. The other bar that you have to leap over is to get it to work scalably and reliably in a production environment. And so what I would propose is that if people are intent on using these low code tools in production, start with a non-technical user who defines and lays out how the workflow should work, get it prototyped, and then hand it over to an actual developer who can then look at the workflow,
00:17:09:10 - 00:17:23:15
Joel Moses
decorate it with things that help you with troubleshooting, with the log statements in the right place, get it to production quality. And that's where I don't see a lot of people spending a lot of time on these systems.
00:17:23:18 - 00:17:41:28
Lori MacVittie
I would love to push back on the idea of non-technical people writing workflows, which are basically codifying processes in environments that they don't understand.
Joel Moses
Well,
Lori MacVittie
I would love to have this discussion, but we're
00:17:42:00 - 00:17:42:18
Joel Moses
Yeah, look
00:17:42:18 - 00:17:43:24
Lori MacVittie
starting to run out of time, man.
00:17:43:24 - 00:18:01:13
Joel Moses
I know, I know. It's that deficit between people who understand the use case and the people who understand the systems. That's the problem here. The non-technical user knows how they use things. They know the flow. They just don't know why the flow has to work certain ways. And that's where the developer comes in.
00:18:01:15 - 00:18:24:17
Lori MacVittie
So prototype. Using these kind of tools very
Aubrey King
Oh yeah.
Lori MacVittie
Remember VB back in the day right? It was great. Visual Basic was great for letting non-technical people build out user interfaces and describe flows for prototyping. Excellent use case. And then someone would actually take that and turn that into a production grade thing.
Joel Moses
Yeah.
Lori MacVittie
Yeah. Aubrey?
00:18:24:19 - 00:18:26:15
Lori MacVittie
Good idea?
00:18:26:17 - 00:18:31:09
Aubrey King
You know
00:18:31:11 - 00:18:31:14
Lori MacVittie
Uh oh. Can of worms.
00:18:31:14 - 00:18:34:12
Joel Moses
You're talking to him after he's been troubleshooting one of the systems.
Lori MacVittie
Can of worms.
00:18:34:12 - 00:18:56:18
Aubrey King
Yeah, that's the problem. I mean, coming from a troubleshooting session direct into asking me about that is tough. I like the idea of these really more to get an idea of what code you should make for production. So like, let me give you an example. There are, as I go, we talk about low code right?
00:18:56:18 - 00:19:18:03
Aubrey King
So as I have gone through creating this flow that I've had for a while, I've had to go through it and do it a couple of different ways. I ended up having to do it non deterministically, where I actually just broke out the JavaScript and started writing. Just to make sure that that the non-deterministic flow worked and when it did then I could go back to, oh, I'm sorry that the deterministic flow worked,
00:19:18:06 - 00:19:40:26
Aubrey King
then I could go back to a non-deterministic pathway and continue that. But those little bits of JavaScript, those little bits of JSON that you end up dealing with while you're working these flows are, I think, the gems that you pull out of an N8N and put into, you know, a production grade Linux system or something like that. I really do.
00:19:40:26 - 00:19:49:27
Aubrey King
It's great to get started, but I don't, at this point, I don't think this is ready for primetime. I wouldn't put it in my production environment.
00:19:50:00 - 00:19:53:27
Lori MacVittie
Okay. That sounds like your takeaway.
00:19:54:00 - 00:19:55:01
Aubrey King
It's not there yet.
00:19:55:04 - 00:19:56:18
Lori MacVittie
Okay. All right. Joel? Anything you wanna...
00:19:56:20 - 00:20:20:07
Joel Moses
Well, you know, my takeaway is actually the inverse of that. I think they are ready for production, provided you give them enough care and feeding from a developer who knows how to decorate it with things that are production grade. But and the reason I think that is, first of all, it's great to have a non-technical user who can define a workflow because they know where the data elements are that they're interested in. They know the systems that they work with.
00:20:20:07 - 00:20:41:08
Joel Moses
They know the flow that they use. I'll give you an example. An accountant knows how to navigate the ERP system and go to various pages and craft a manual workflow. And what they're trying to do is visually represent a manual workflow. That's incredibly valuable to a developer. A developer would have to figure all that stuff out on their own by studying the user's behavior.
00:20:41:15 - 00:20:55:22
Joel Moses
If you put these in the hands of a non-technical user, the user can model it for them, and that gets you to production faster. But it doesn't get you to production right after the non-technical user produces it. That's my point.
00:20:55:24 - 00:21:18:12
Aubrey King
Little bit of time, yeah.
Lori MacVittie
Wow. All right. Well, you know, as we close it, I guess my takeaway is probably not ready yet. You still have paths to go. If your scripts are working, maybe don't rip them out just yet. Yeah, that's what I'm going with there.
Aubrey King
Fair.
Lori MacVittie
So, that's a wrap for Pop Goes the Stack.
00:21:18:14 - 00:21:29:27
Lori MacVittie
If you enjoyed today's interrogation, smash subscribe. Then try emailing an AI some sarcasm. You'll get why it's needed. We'll be back to grill the next shiny thing soon.