KubeCon + CloudNativeCon NA 2025 | Graham Siener, Honeycomb.io

Clips
More from KubeCon + CloudNativeCon NA 2025

Graham Siener

VP Product

Honeycomb.io

play_circle_outline Meet Graham Siener: Honeycomb's VP of Product and the Pioneers of Observability at KubeCon

play_circle_outline Trend of consolidating tools to manage observability data in one platform.

play_circle_outline Enhancing Observability: Honeycomb's Role in the OpenTelemetry Community and the Shift Towards Collaborative Industry Standards

play_circle_outline Honeycomb expanding support to larger companies while maintaining scalability.

Info
Transcript

Graham Siener, Honeycomb.io

Graham Siener

VP Product Honeycomb.io

In this conversation at KubeCon + CloudNativeCon North America 2025 in Atlanta, Graham Siener, vice president of product at Honeycomb.io, joins theCUBE’s Rob Strechay to unpack how observability is evolving for cloud-native and AI-driven applications. Siener traces Honeycomb’s roots to co-founders Christine and Charity coining the term “observability” from their experience with Facebook’s Scuba tool, and explains why customers are consolidating metrics, logs and traces into a single, connected telemetry platform. He shares how the rise of code-generating agen... Read more

explore Keep Exploring

What is Honeycomb's role and perspective in the field of observability? add

What is the reason behind the creation of Honeycomb, and what trend is occurring in the field of observability? add

What is OpenTelemetry, and how is it influencing the approach to telemetry data in companies? add

What has changed regarding the focus of Honeycomb's target audience over time? add

bolt Powered by CUBE AI

Graham Siener, Honeycomb.io

search

Rob Strechay

>> Hello and welcome back to day three of KubeCon CloudNativeCon North America from moderately warm Atlanta today. Again, the heat is rising on the set here. We're getting and diving in and really looking for visibility into what's next. So what better way to do that than I'm joined by Graham Siener, who's the VP of Product for Honeycomb. Welcome on board.

Graham Siener

>> Thanks. Great to be here.

Rob Strechay

>> So observability has been a big topic. It's a big topic for many different reasons in the community here. Kind of help people understand where Honeycomb fits from an observability perspective and what your platform does.

Graham Siener

>> Sure. So our co-founders, Christine and Charity actually coined the term observability about 10 years ago. I mean, their origin story is they were working at Facebook, they had a tool there called Scuba, and it really was the difference to understanding, monitoring, we have a challenge, to we actually understand what's causing this, here's the solution. And that's really the genesis for them starting Honeycomb. What we've seen is people are consolidating, right? They're using a lot of different tools and they're realizing it's actually more valuable if they have all their data in one spot. And so I would say the macro trend we're seeing in observability is people are looking at all these different applications. They're looking at all as telemetry, and they're saying this needs to connect together. And that's how we've been built from day one. It's great to see everyone, I think, getting on board with that. And there's certainly a lot of tools that are helping people do that.

Rob Strechay

>> Yeah, I mean I think it probably has never been more important than it is today, especially with the complexities of things like AI and cloud native architectures and people building AI on. But oh by the way, it's still connected back to my legacy and I still have VMs and in some cases mainframes and all the way back. So how do you see things evolving and what have some of the conversations been like this week?

Graham Siener

>> Yeah, you're absolutely right and there are so many people that are generating code faster than ever with these agents now, and what we see is they don't have the time or energy to maintain some of those old applications. They're actually just rewriting. It's really fascinating to hear all these companies who are now looking at that calculus and saying, it didn't make sense for us to rebuild or port this application over, and actually we can just rebuild it now. We've got the right interfaces, we understand how it should work, and now we can vibe code something to get in production. The problem is they have no context on how it actually works. They didn't actually write it. And so we're hearing from a lot of companies that they're really nervous. They love that first hit of getting stuff into production faster, having some of these agents do a lot of that heavy lifting, the first production incident, they have no idea what to do. They're basically sifting through code for the first time. So I think they're starting to understand the value of having some better observability around those apps.

Rob Strechay

>> And sitting there and instrumenting it from the beginning so that they can understand and when things go bump in the night, where do I go and what do I do, from that perspective. So also talk to... This week, the community here has been very vibrant this week. There was also a lot of day zero stuff going on. You guys were very involved with what's going on in OpenTelemetry, kind of help people understand what was going on with you guys on day zero as well.

Graham Siener

>> Yeah, so OpenTelemetry is something that has really found its footing. It was a very nascent project, and we made a really big bet that that was what we should be adopting. Those should be the open rails that we're building on top of. You look at a lot of companies that have these proprietary systems. They've got all this proprietary telemetry data coming off, and they get increasingly nervous that that's harder for them to switch. It's harder to find library support if they're adopting new standards. And so yeah, the OpenTelemetry just first day here was amazing because we have so many maintainers now at Honeycomb, but just everywhere here at KubeCon. And it's really cool to see all these different vendors come together and say, this should be a standard, we should make this good for everyone. We don't need to compete on the telemetry that's going over the wire. We should be competing on the experience on top of that.

Rob Strechay

>> Yeah, I think that to me makes a lot of sense, and that's what I love about this event in general and some of the other CNCF events where they're bringing together a lot of the maintainers and people who are running the different groups in there. In fact, you guys are running one of the groups as well, or heading up one of the SIGs here as well.

Graham Siener

>> Yes, that's right.

Rob Strechay

>> So I mean, I think you're giving a lot back to the community, which I think is key. I mean, that's how, again, the community grows and you compete on other things. That's right. It's coopetition in the middle and compete around it. Help people understand where best to kind of join up with Honeycomb and what some of the use cases and problems you're solving for them, beyond just the vibe coding one that we talked about.

Graham Siener

>> What we do really well is help people get to identification of problem faster. And one of the reasons is because again, all of this data's connected. So in a lot of systems you might have metrics coming off, they're giving you insight into CPU, but then if you have an error, basically now you're going through and scanning what happened in this time period. Let me look at the logs, let me look to see which applications are doing what. When you connect all these things, when you have metrics that can get correlated with your traces, correlated with the logs, it gets a lot easier to say, okay, I know there's a problem and I actually know it's this that's causing it, right? It's this API service here that's slow because of this misconfiguration that we just pushed into production. And so here is the issue and here's how we fix it. So we have a lot of customers who really have time sensitivity. They're an e-commerce company, they're a streaming company, and every second that they're down or they're degraded in performance, they're losing money. And so that's where we find a really good fit.

Rob Strechay

>> So I mean, I think you hit it with the metrics logs and traces and bringing it together. I could see why Honeycomb, connecting it all together in a mesh there. Do you see that people are trying to figure out... Again, I look at it and go, AI is definitely changing the game, but it's still an application. And depending on how you're building your agentic or if you're building agents or even chatbots, there's a lot of complexity and a lot of different pieces that can go to... Is that where again, you're bringing it all together can help people understand these are the different moving parts underneath there?

Graham Siener

>> That's right. We have a lot of companies that are not just using LLMs to make code, but now they're building experiences on top of LLMs. Duolingo is a big customer of ours, and they talk a lot about what it looks like to debug those chat experiences. It's not, this is working or it's broken, it's degrading in unusual ways. And so having the visibility to understand, here's the prompt, here's the sequence of events that happen, here are the tokens, really helps them troubleshoot faster. And I mean, this is a new world. I mean, these things are behaving in unusual ways that we can't really predict. And so if it's not well-instrumented, if you don't have a sense of what's going on, you're just going to spend a lot of time just kicking tires, troubleshooting, understanding, "Did this fix it? No. Did this fix it? No."

Rob Strechay

>> Yeah, I think that the mean time to resolution, it means something, especially with everybody trying to figure out, because the switching cost, especially for things like e-commerce is nascent to put it... Or small, very small, not nascent. And I think that people want to have good experiences. Are you finding that it's not even just the e-commerce companies, but it's traditional companies even with their internal apps, if they have a customer service app or things like that as well?

Graham Siener

>> Yeah, we support SLOs, and SLOs I think are really good best practice as you start to adopt SRE mindset and being able to say, here's what our customer experience should be, and here's how we measure if we're meeting that standard, is really helpful. Even for those internal apps. We talk to a lot of teams that they treat their developers just as important as their customers because they want to make sure they're productive, they want to make sure they've got environments up and running. And so every environment, doesn't matter. If it's down, someone's not productive.

Rob Strechay

>> Right. And I think when we look at all of what's going on and we've gone around the floor here, everybody's trying to get a handle on it was with a number of platform engineers on Monday night at the Infrastructure as Code Conference connect, and we were having this discussion about, to your point about from the prompt back and how do you understand that? Has that been an ask for companies that is increasing? Because they're trying to figure out, to your point, how they track all the way down through these applications that sometimes misbehave in weird ways and things like that.

Graham Siener

>> Yeah, absolutely. A lot of teams are... They're reaching for off the shelf or whatever tool is coming with the agents that they're building. The problem is that lives in a silo. And so if you can start to connect these things, again, just like everything else, it's not just what did the agent do, it's what was the user doing on their mobile app that connected to the server that was talking to the agent? Being able to trace that end to end is where you get some of the visibility that helps you understand, "Oh, this is where the problem is."

Rob Strechay

>> And I would expect that when you start to look at the ecosystem of everything that's going on, things like O-Tel really help because it's there to begin with. And we're talking about some of the new updates that they've had to it and things like that, that have really helped in that injector and be able to grasp more data. Are you seeing that the community is really leaning in, especially with these new brands of applications, which may or may not act the same as the previous ones as well?

Graham Siener

>> Yeah. Yeah, that's right. I think what's cool is, again, because it's open standards, people are seeing those gaps and they're filling them. Gen AI is an area where we started to create some semantic conventions, started to think about how do we treat the prompt? How do we think about token usage? All of these things are getting baked into O-Tel, and so you just plug it in. If you're a Claude code user that actually emits open telemetry, so you can start to look at all your developers in their sessions and understand how are they using Claude code.

Rob Strechay

>> Wow. Yeah. I mean, I didn't even know that. So that's very interesting, and we have to check that out because I use Claude. So when you look at all of these companies that are here, I mean the companies are wide-ranging. What is an ideal size of company, or does it not matter, and verticals that Honeycomb really excels at?

Graham Siener

>> Yeah, it used to matter a little bit more. I would say where we really started out was focused on that really elite cloud-native company. If you're building cloud-native, you're on the cloud, great, go ahead, Honeycomb's a good fit. And we took a hard look and said, hey, we've got some bigger companies that are using Honeycomb now, what does it look like to really support them? And to your point, that looks like first-class log support, metrics, being able to understand what that data pipeline looks like for all of these different data sources that they're managing, all the sampling and just to handle data volume. So yeah, at this point, we're supporting some of the biggest companies in the world, and it's the same principles. It's just at their scale.

Rob Strechay

>> Yeah, I mean, in some of our research, we do a lot of surveys out and bring back the research. It always comes out that companies are using four or five different observability products, but they all want to consolidate and move towards a platform. That's kind of, I think what you're talking about there is, people are looking to bring their data somewhere and really not have swivel chair for the SREs about, hey, this is doing my logs, this is doing my metrics, and this is doing my traces. Has that been a great advantage that has provided to your customers?

Graham Siener

>> Absolutely. And we started building out an AI assistant, and we have this workspace called Canvas, and you can ask it questions. And what's actually been great, and it's kind of coincidental, is because the data's so connected, because we encourage customers to send these really wide events with a lot of schema data, without a lot of training, these models can now do a really good job of understanding your data. And so if it's all connected, it can make better decisions, it has more context. And so that's actually been a really nice loop to get customers to really just consolidate, send more of your data to one place. And so now you have a model that can actually look across that context and make better decisions.

Rob Strechay

>> It would seem that it's a way that organizations can get their head around it. And I think also part of it is SRE is sometimes a first step into a dev role, or hey, they were in a junior dev role, but they came into that. So definitely that persona I know, because I had some that worked for me at AWS. It's like those are the people who are on the front lines every day and really doing a great job and things like that. When you look at this, and we're going to be hopefully sitting down together a year from now in Salt Lake City, what do you hope that you can say then that you can't say today?

Graham Siener

>> Well, first of all, I hope that people learn a little bit more about what AI can and can't do. Certainly we're talking to a lot of companies that are really asking, can AI do my SRE's job for me? And I don't think we're at the point where that's going to happen. I will say, a year from now, we're going to get more comfortable pushing that slider of autonomy over to more autonomous and just seeing some of the work that even really skilled engineers and SRE's do. It's repetitive, there are things that they don't need to do as a first step. There's a lot more that we can do to not just say, here is the incident, but also here's a remediation step, would you like to take it? And that's where I think everyone's going. And I'm pretty excited that people are going to be able to spend more of their time actually solving higher order problems instead of doing some of the lower level work.

Rob Strechay

>> Totally agree with you. Hey Graham, I really appreciate this. This has been fantastic and have a great rest of KubeCon CloudNativeCon.

Graham Siener

>> Yeah, thanks for having me.

Rob Strechay

>> Thank you. And thank you for watching this episode from KubeCon CloudNativeCon North America 2025 from Atlanta. We'll be right back.