IT Environment Complexity Will Drive Observability – Christine Yen, Honeycomb

This is texturing TV. Hey guys. Thanks for the throw.

We're here with Christine Yen. Who is CEO for honeycomb? And they just landed 50 million dollars in additional funding kind of hard to come by these days.

But question immediately Christine is hey do people really get observability. We've been talking about it for a while and a lot of people not their head. I think they still think it's continuous monitoring.

Maybe they get confused about what the difference is. What's your sense of? What is our real maturity level these days for observability.

I think it is increasing but a lot of people out there are still hungry for answers. I think one of the most exciting things to watch last year was the rise and just dominance of open Telemetry as the the standard way to get data out of your applications and into something that can help me make sense of them. Open Telemetry is driving some really great things.

Tracing Chief among them and honestly, I think maybe correlated with this Movement. We are hearing many fewer people associate observability with oh just logs metrics metrics and traces and be much more curious about what they can do. What new things they can do with an observability tool something really modern and meant to understand, you know, today is applications.

Doesn't hurt that are the O'Reilly observability engineering book came out authored by three honeycombers. We've seen it ton of interest and hunger in that as well. And again, I think it just reflects.

A maturing interest in observability isn't being more than just a buzzword, but still hunger to really understand. What does this mean? How do I get started?

And how can this benefit my team? I think you touched on the issue a lot of folks are used to predefined metrics, but when they're confronted with observability, I think sometimes they're just overwhelmed and they don't even know what questions to ask so and where to get started. So how do we get people down the path?

Because they say that the sign of intelligence is knowing the right questions to ask in the first place. So, how do we make everybody smarter? Your question and frankly.

This is something that we have been thinking about and you know trying to build pieces to help with at honeycomb for forever. There's a couple things here right the first being It's hard to break a habit and many of us who are used to metrics and monitoring dashboards have just adapted to have that be our habit how we think about systems. You you look at the dashboards that have already been built and if the data is not there then you're like, okay.

Well the date is not there. And a lot of what I think of as being the movement around observability is challenging folks to say hey. Don't just be satisfied with what's already there take an active hand in exploring what your applications are trying to tell you.

So that's sort of a cognitive precursor piece. But this question of what questions do I ask of my data. Absolutely.

It's the hard part any data tool. Like how do you get started using it? I could I could but I won't get into his room for the ux and exciting technical developments in the industry right now on that front.

I I'll say that for another for another chat for here. Really I will just draw it back to I think what you and I talked about in the past which is helping Engineers. All Engineers see their worlds in a given tool as a developer.

It was really frankly off putting to me to be given a monitoring dashboard and told to figure out what my application was doing with these graphs of mango throughput and CPU utilization. That's not the language that I used doesn't map to my tests doesn't matter to my users. Now that we are seeing more and more tools support.

It's Rich custom metadata. I cardinality high dimensionality all those phrases. An engineer can say well I care about this customer.

I'm looking at this endpoint where I care about this logical path for this type of user. And the more that users can see their world reflected and this data tool that's supposed to tell them what all their software is doing the more they can start to tap into that Natural Curiosity to figure out what questions to ask about their software. What is your sense of?

The competitive landscape right now. I mean everybody and his brother ever sold an APM platform says they're in observability and then there's all kinds of other folks that have jumped in. So what differentiates honeycomb and what's your sense of how crowded is this space?

And how crowded will it remain? I think that there's there are many. Let's see.

Well what differentiates honeycomb is that? We look at. How first we're very much aligned with the application itself.

We think that that's where all the interesting stuff that's happening infrastructure metrics infrastructure. Telemetry is good context But ultimately what you're trying to do is understand why your software isn't behaving. that's what's impacting your customers the way that we approach that differently than the committed other folks on in the competitive landscape and say is that Many other tools are very focused on this problem of how do we correlate across these different data sources.

How do we correlate a log and a trace and a metric over here and present that as a single pane of glass? Honeycomb has always approached this space for most perspective of that's the wrong. That's the wrong question to be asking.

It is a data. It becomes a data correlation problem. If you accept the premise that you should capture logs and metrics and traces separately that they should live separately and and be stored separately and be queried separately.

but in the reality logs metrics and traces are just visualizations of something that happened in your system. And if we can accept that premise, then you can build a tool and you can use a tool that allows you to work with a single source of Truth in our world. This is traces.

We think traces are the richest and most flexible form of data that we now have the computation power to process quickly starting from there and capturing and visualizing it in a graph or as individual log lines if you need this and again, I know it sounds a little pedantic to say it goes down to the comes out of the data store. But in the end all every tool in this space is in the business of capturing a bunch of telemetry reflecting it back to users. And so what ends up being possible with that tool comes down to how is the data stored?

What can I do with it? And honeycomb has always been focused on building this. Seamless integrated experience not gluing together different product experiences under a single pane of glass which again, who knows what kinds of seams there are behind that glass.

Do you think the current level of complexity that we're seeing in? It is going to force all these observability issues one way or another. So now it's just a question of how long is it going to take for me to get there but one way or another we're all going.

Seven years ago when honeycomb started no one was using the phrase High cardinality. And all it means is for a given field. There's many many possible values this increasing complexity at a technical level.

Now that we're instead of running five app servers. We have 500 microservices across each running on a hundred different containers. This technical complexity is absolutely itself driving a change in how we think about data and working with that data.

So 100% to layer on top of that technical technical complexity. I'll point at two sort of social trends SRE and platform engineering as new ways that engineering teams are adopting our new practices that they're adopting to look at their software systems differently. both of these disciplines through Elevate the expectation that we be thinking about our software from the customer's perspective.

And doing so again Force engineering teams to not just look at containers and processes but customer IDs Merchant IDs shopping cart IDs all of these these ways of Licensing through the noise to find a signal that actually matters to you to solve the answer whatever question you're trying to ask. We of course hear a lot about AI you cannot walk down the street today without somebody showing you their great new AI large language model this or machine learning algorithm that what is the role of AI and observability this are they complementary? It is one drive.

The other does one need the other. Now it's gonna be complementary AI is a way of trying to navigate a path but as we've seen through, you know, everything that we've seen in the news. Hey, it's not always right.

I think that in the past I have been misused in this space in trying to trying to come up with the answers for the humans. The problem with with over Reliance and AI in that way any sort of magical we'll figure out what's going wrong for you. We'll we'll tell you when things are going wrong.

You don't you know, you humans don't have to worry about it problem with that is that in? This world false positives and false negatives are equally damaging when the when the action is do I wake an engineer up in the middle of the night. Either way, you're either burning folks out or you're allowing damage the system to continue where I think machine learning is really interesting and note I see machine learning not artificial intelligence because ultimately, it's we're not trying to replace the humans here where I think ml can be really well applied is in helping folks understand the questions to ask helping them translate what is in their head to?

how to navigate this data set that the human may not be familiar with There's some really interesting interesting Explorations that can happen there and I can't wait to see what the next few months in quarters bring. So once your best advice to folks about observability how to get started what have you seen you're more successful customers do and others have not. I think I've learned is.

What makes an engineer great if being able to look at a set of tools understand those trade-offs take those rules put them into their heads and then build build their software or lives on top of those rules. but in a world where the rules are changing so quickly. From oh, well, I have to work with pre aggregated metrics or you know, I have to understand that data set myself.

I think the folks who I've seen must be most successful in. Embracing an observability practice and allowing to improve their cultures is stepping back. and through always Trying to figure out what is the problem?

We're trying to solve. What is the question? I'm actually trying to ask.

and then figuring out how to map that question to the tool or data set they're working with So much more often, you know to the flip side. The anti-pattern is when folks go from this is the data I have to work with how do I you know, what do I do with it? We're in a world today where you know the rules of how we work with data are changing.

It seems daily. Why constrain ourselves to what we've always done in the past? Right in my patterns may have patterns.

Hey Christine. Thanks for being on the show. Thanks so much.

Back to you guys in the studio.