Intelligent Canaries at SKILup Days 2024

Hello, my name is Mark Callahan. I'm the founder of Cloud Canaries. And, uh, thank you for joining my session at, uh, skills Day.

We're gonna spend some time and go through what intelligent canaries are. I'm hoping that the presentation will be enlightening and a little bit entertaining too. So the first thing that we're gonna do is actually look at what an scur is.

And I learned about s-curves a long, long time ago at the MIT Sloan School talking about innovation. Um, and it's followed me, you know, through my entire career. An SCUR is this kind of path of technology.

And on this cover slide, just remember where the inflection point line is. It, it will come back. Um, again, Mark Callahan, founder of Cloud Canaries.

You can find all about me on LinkedIn today. We're in this intelligent era, uh, present and future cloud canaries. Cloud canaries are work, uh, use workload data.

AI and compute cloud canaries are microservices. They perform many different tasks such as monitoring, observability and repair. Intelligent compare, uh, intelligent canaries are powered by AI with workload data and compute intelligent things along with intelligent canaries.

Are you prepared to make the shift risk benefit analysis trade off? Here's a, uh, cartoon with a intelligent thing in a cage, and the, the quote is, if we let it out, there's an 85% chance it will cure cancer. 0, 1% chance that it'll take over the world.

It's the intelligence era, and you have to decide, are we at an inflection point? Is it time to open the cage? Think about that as we go through the, go through the slides.

But before diving into intelligent conne canaries, we're gonna go back in time and we're gonna go to the, uh, 18 hundreds, uh, and look at schooner, beautiful schooner and steamships and the s-curve. A schooner, you know, was used for, uh, freight for well over a hundred years, if not more. Very efficient.

Use tradewinds as fuel only limited by weather. Still, the sales technology was highly advanced. Steamships new technology, the steam engine and coal, as the fuel allowed new freighter designs, shipping was not limited by weather.

Technology advancement was accelerating. The scur technology, a I'm gonna say is a schooner and technology BA little bit further in time is the steam ship. What the scur does is describes the diffusion of technology and how it accelerates, and then over a period of time kind of matures, stabilizes as the market saturates.

Which technology was the, at the end of the, of the s curve of ITSs curve. And the answer is the cells tech. Within a decade of steamships, uh, cells, uh, were replaced with steam engines powered by coal.

While the schooner cell technology had advanced greatly, it was unable to stay with the new scur s-curve, the technology B it was, it was a new S-curve on the block, steam engines and coal. We're gonna take two different quick looks at phases of the S curve. I call one side the emotional phase, and, and the other side is the economic phase.

The emotional phase is like, is this ever gonna work? It's beginning to work, it's taking off, we're gonna be rich. Oh my gosh, what is happening?

I noticed the inflection point down at the, as the S trip begins to take off on the economic side, it's very similar, uh, but it's based on economics, pilots, early adopters, acceleration, new integrations, and finally, uh, market saturation. Remember this, the S curve drives everything. So now let's begin to look at the intelligence era data and compute our AI's rocket fuel Data.

Properly cleaned and organized is the rocket fuel of tomorrow's powerful AI solutions. And I would say today's smart services, new customer experiences and AI powered tools and intelligent agents sounds very familiar with intelligent canaries, by the way, that will augment the employee employees, give them superhero level capabilities and simulate and simultaneously boost their job satisfaction. Famous guy, Steve, um, brown said that, um, I, I didn't say it very well, but you can go look at his quotes.

So, data and compute are the AI rocket fuels. Let's take a quick dive in this. I'm not gonna spend too much time.

Just one slide each, but it for data, data quality is obviously important. Accuracy, consistency, completeness, timeliness, and relevancy data organization, which is kind of interesting because the way data gets organized for, uh, let's say using an AI model, a neural network is a little bit different than, you know, you or me actually using or organizing data for ourselves. Quantity of data, big, big, big thing.

The more data, wider, deeper, uh, will improve the, uh, ultimately improve the, the models, uh, capabilities. Now, the key with all these things here is that when you do this change data or modify it to improve its quality, organize it and just deal with the volume of it. Um, it's a fairly expensive process.

So the platform that you use to do this should be, um, as inexpensive as possible. And it, and, and in other words, you should be getting this data, you know, very good quality already organized and in massive amounts. Few seconds with the, um, the other side of the, the rocket fuel and, and that's compute.

Um, there's really three things here as far as compute, the logic, the CPU, the GPU, um, the, uh, graphical processing used for neur networks and, and AI models in general. Lots of them run them in parallel memory, lots of it. And, and the interconnection of, of all that, that surprisingly looks like a neural network.

And we'll talk about that at, at, uh, you know, shortly. This is all driven. The intelligence era is driven by Moore's Law, and basically it says everything doubles every two years.

So the density of of chips, you know, increases, which means the prices, uh, go down, but it also just means the power of compute increases dramatically over a very short period of time. And that's really along with data, the big drivers for, for ai, the new tech kit on the block, existing observability solutions are obsolete. So now we've come back to the present And we're gonna take the s-curve information and what's the rocket fuel?

And we're gonna begin to talk about AI and some of existing cloud solutions that all of you are using. The key here is, um, we, yes, actual workload data is used for alarms and notification, but the paradigm shift is that the, the data and compute that we're generating becomes the fuel for ai, the AI steam engine, I might say. So log and trace data, it's fun to look at.

Um, but observations generated using a neural network does not use logs or trace data. We remove the psychological bias, very important. It's like the captain of the steam engine.

Um, looking and looking at the, what is the apparent wind, uh, which means the wind that's used to power sails for today. The captain of that steam engine really doesn't have to do that because he's looking at old technology. He has new technology.

Workloads do not lie, they do not keep secrets. Workloads are the beacons. The more beacons, the better, the wider, the deeper as possible.

We're gonna talk about workload data and how it can impact observability and a lot of other solutions. Neural networks, you know, they mimic the, the brain, um, uh, neurons and they're all connected and they all have weights and, um, they can identify, uh, things that we just don't normally easily identify with logs or, or, or trace data. Um, and it can provide answers that sometimes are like, where did it get that answer?

Uh, and we'll talk about that in a little bit too. Compute. The point here is, is that as, as compute increases, you can take simpler models, general models and, and not really use any fancy algorithms, but you can just use general models to actually generate, um, your recommendations, your predictions, your forecasts.

And so the big driver is not how fancy your algorithm is or model per se is, but how much compute memory you have. And that's, that will dramatically drive ai. Intelligent canaries come all the way back to intelligent canaries, back to the present and future.

They use workload data, AI and compute. As I said before, intelligent canaries are micro surfaces, can call them agents. Uh, they perform many different things.

They can perform monitoring, observability, repair, but it's, uh, the blue sky is the limit. They use billions of workload, billions of workloads, each with data. Um, again, workloads are the mariners beacon, the north star.

And when you have billions of workloads, um, you can collect masses amount of data that you can use for your, your model generation. Artificial intelligence and the, I call it the AI shrug, uh, neural networks can be head spinning even for the best neural scientists. I know a few, and sometimes I just say, um, I don't know.

How did it make that prediction again? Uh, when you're talking about millions of, of, of nodes each with weights and this all interconnected, um, you can actually have situations where people don't really have a clear idea how those answers came out. How were they generated?

It's not as clear as log data traces or profiling, but it's more accurate. Uh, compute Moores law, again, this is all driven by the fact that every two years in general, the number of I, uh, integrated circuits are a, aren't as CPU or, or other logic doubles the amount of memory, the density of memory, you know, doubles and therefore the cost to, to use those new technologies decreases. And that really drives, is really gonna drive this, uh, intelligence era intelligent canaries in observability.

So we're gonna use observability as a, as an example, a solution Observability with instrumentation. Instrumentation is a collection of log files, traces possibly profiling used, uh, using, uh, the collected data for forecasting, um, observability without instrumentation. And that's, I call it technology B observability with instrumentation is technology.

A workload data and AI compute insights from workload data are added to the model that that limits the psychological biases that you might see otherwise. And, uh, you get these cleaner predictions, forecasts, insights from that, from using this massive amount of workload data. Again, the S curve comes back.

So observability with instrumentation technology, a observability without instrumentation technology being, and as we've seen in the past, um, the newer technology that's rapidly developing going up that scur usually wins, pick the next cloud solution. And there's, you know, there's so many of them. Um, you know, and I'll actually go through that.

And most of these are obsolete. They are the s-curve technology A and the new technology B, which is AI driven, is gonna take over. So it's not whether they're obsolete today, but they will become obsolete tomorrow.

So pick your rabbit from the hat. Uh, here are just a few, uh, of the many different applications that intelligent canaries can, uh, uh, can be used in. Again, it's workload data, AI and compute, Digital experience, SLAs, contract negotiations, crisis response, market analysis, sales forecasting.

I can just go on and on. Observability risk management. So, um, you know, Greenfield, so we've only, you know, we're, we're basically, uh, you know, at the pilot or the early adopter phase right now, and that s-curve is gonna take off.

So the last thing we're gonna do is, is take a look at a digital delivery lifecycle. AI powered and intelligent. This is the new one versus the old one, which really was an AI po wasn't AI powered.

And, and in each of the phases, you can apply intelligent canaries, uh, insight and analysis. Very obvious how you can use the, uh, intelligent canaries observability monitoring, uh, portfolio and backlog. Basically predictive analysis, you know, uh, AI augments that are even replaces it.

Continuous integration, uh, continuous testing or continuous mo, continuous monitoring, uh, using canaries and continuous delivery. All these can be, can leverage intelligent canaries. Thank you very much.

And let us know if you wanna open the cage.