AWS Financial Services Symposium | Dan Holmes & Nimish Doshi

Clips
News
More from AWS Financial Services Symposium

Nimish Doshi

Director Technical Advisory

Splunk

Daniel Holmes

Director Observability Advisory

Splunk

Observability at the data core: Examining its interplay with gen AI and cloud-native technologies

Observability has become a cornerstone for enterprises as companies harness the full potential of generative artificial intelligence capabilities driven by cloud-native tools. Several emerging trends show that more than a buzzword, gen AI in observability is a critical capability driving the future of IT infrastructure and application resilience.Splunk’s Nimish Doshi and Dan Holmes talk to theCUBE about gen AI in observability

play_circle_outline Importance of observability in the cloud-native and GenAI movement

play_circle_outline The role of the cloud and AI in driving change and resilience in applications

play_circle_outline Maximizing Observability: Unifying User Experience, Synthetics, APMs, and Infrastructure Monitoring for Enhanced Insights

play_circle_outline The challenges of scale and cardinality in managing observability data

Info

Dan Holmes & Nimish Doshi | Splunk

Nimish Doshi

Director Technical Advisory Splunk

Daniel Holmes

Director Observability Advisory Splunk

>> The AWS Financial Services Symposium, presented by theCUBE.

>> Welcome back, everyone. We're here in New York City with theCUBE. I'm John Furrier, your host. In this session, we're going to talk about observability. We're going to go deep, we're going to do a deep dive. We're here at the AWS Financial Services Symposium. Dan, Nimish, great coming on theCUBE with Splunk, Director of Technical Advisory and Director of Observability. Guys, we got the players here for the observability conversation.

>> Absolutely.

>> Thanks for coming on.

>> Thank you for having us.

>> Observability is the hottest topic right now besides governance, some of the data AI stuff. But at the foundation level right now, cloud-native is propelling the GenAI movement, we all see that. That's the hype cycle. We're right out of the meat and potatoes and the steak on the grill, so to speak, is where's the data? What's happening in the network, the observability? And it's changing and you guys have been on the forefront of this with Splunk from log files in the early days, to being part of every enterprise's fabric. Where are we right now with observability? What's the key trend that's, I won't say being retool, but being focused on in terms of managing observability so that the enterprises could be positioned to take advantage of the generative AI?

>> Yeah, absolutely, I can share a couple of comments. First is the first disruptor that really opened people's eyes around observability or introducing change was cloud, move to cloud. So as the architectures and infrastructures supporting applications move to cloud, it's introducing a change in how organizations build applications, how organizations run and operate those applications, and then more importantly, how organizations ensure resilience in that application experience. I think that's the first. The second one which I think AI is starting to introduce is we're all coming back to the killer app again. And every time there's a new killer app or a new generation of killer apps, it starts inspiring change in everything from how you architect those apps, but again, to how you drive resilience and a quality customer experience.

>> Dan, that's a great point. I'd also say too, that data is a big part of it, right? When you talk about observability, it's touching all the hot buttons. Security, user experience, new experiences that become expectations for this next generation. There's a lot going on and I know that there's been discussions around the network observability, there's different layers. Take us through, take a minute to explain what is the current state of the art of observability and how do we start framing this?

>> There's a lot of different players out there and a lot of different technologies, but one of the things they start off with, just the users themselves. Real user experience or user monitoring. People forget that part of observability. Then there's things such as synthetics, the ability to monitor websites, APIs and do it in an automatic fashion. So the APMs and the metrics of the world get all the highlights, but those little things out there are just as important. Then you get into, of course, the service analyzers, the ability to do application performance monitoring, and you get in infrastructure monitoring. All those things are nice, but putting that all together as one platform is actually what we call state-of-the-art. As you mentioned network, that is something that's just as important and Splunk is part of Cisco. There's a lot of different network capabilities we have in there.

>> We could do a whole rabbit hole discussion about ThousandEye to Splunk, which I think I'm very bullish on, just get it out there. Again, I keep saying that, I keep pounding the table. Cisco, but let's a whole other conversation. But I want to get your thoughts on that concept of putting it all together. There's more to observe now, so the surface area and volume and diversity of signals. Can you guys share your thoughts on what that means and how that translates to how customers are thinking about architecting and deploying new solutions?

>> Yeah, I think actually, one of the more exciting, when you talk state-of-the-art or what are the trends that customers are really starting to lean into is as Splunk, I can't not talk about the data. And actually, if you look at what's happening at the data level of analytics is the initiative around OpenTelemetry. Because you're absolutely right, the diversity of infrastructures, platforms, applications have exploded. And actually, anticipating that explosion is what inspired the cloud native compute foundation to really mature OpenTelemetry, which gives a high-quality data source regardless of its source.

>> And OpenTelemetry also vectors into the whole open-source movement around where GenAI developers with data is going on. All the open-source models are coming in, there's a developer frenzy right now in open source. Huge, the demand and the appetite to code with data is incredible and the cloud scale is also causing some opportunities and challenges around how do you look at observability if you have now large-scale connections?

>> Distributed tracing, distributed hybrid deployments, that's the new thing. We were talking about it in the morning. Most applications that are in the financial services, they're not going to be a 100% cloud. They're going to be sometimes on-prem, a lot of it on cloud, depending on what the services are. And the ability to actually trace on different platforms, different clouds for that matter, hybrid clouds, that is going to be a new trend as time goes on.

>> Guys, I want to get your thoughts. Again, we're bringing up, we're getting deep here and this is deep tech. It's important because traditionally we're seeing, I mean, you mentioned Unified, which is a good concept we want to get into. But traditionally, most enterprises have all these different IT tools, right? And you bring up OpenTelemetry, the reason why it's popular is that you can have these that could spin and generate data without having all these different tool sets. Can you guys share what the pain is and suffering involved with having managing multiple tool sets versus say, using OpenTelemetry to generate agents and data across?

>> Absolutely, yeah. I mean common, the pains of when you have multiple tools is one, it's more expensive, there's a high cost. Two, when you're troubleshooting, you get a sore neck because your swivel chair, you kind of look in between different solutions. And three, you're not able to allow software to do the heavy lifting on the analytics to say not only this is a problem, where is the problem occurring and what's the cause of it? And so when you can separate the quality data stream from the backend processing and actually make the telemetry open source, it opens the door now for customers to have greater choice in how they want to process it.

>> It also saves on cost because if you had to maintain multiple agents and multiple vendors at the same time, the cost is going to be tremendous.

>> Oh, yeah. And also the managing it. And I think ultimately what I hear people say is that, "Look, I want visibility. I got to have visibility, what's going on, number one." And then this other stuff, get rid of all that. That brings up a good point. You guys had the observability cloud, Amazon's got CloudWatch.

>> Absolutely.

>> How does that all work together? Can you share your relationship with AWS? If I'm a customer, okay, I buy into OpenTelemetry because CNCS is a really big part of our audience and ecosystem, been covering them since it's founded and we see the benefits. But now I'm an enterprise, I have some people that go to these shows, I get it. How do I get it going? What's this, what's the playbook?

>> Yeah, I think the way to think of it is it's actually our step one with customers when they want to improve their observability practices, take the data from AWS CloudWatch as kind of your tier one first data source. There's a lot of tremendous information that you can get from there. Then as you start looking at how you want to look at those custom apps or how you want to be able to trace activity that may not only be running on AWS, but out to other third parties or out to application workloads running on other infrastructures, then you can deploy OpenTelemetry and then just bring all that data together in Splunk and see end-to-end, what's happening.

>> Nimish, scope the impact of what he's saying, because if OpenTelemetry becomes the industry standard and it looks like it's going to be on that track-

>> Yeah, there's even a mainframe committee now.

>> Talk about the scope of the impact in the ecosystem for the technology ecosystem. Because Gen AI is essentially now a platform shift and everything's moving up the stack in terms of capabilities. More performance is accelerating from the bottom, the middle layer, data layer's going and software's going to intertwine. More observability's going to get smarter, faster, more visibilities and probably have agents in there too, doing work. But the impact of the ecosystem, everyone has to play together because it's the platform.

>> It makes interoperability more of a successful thing than it was in the past. Because if you have different agents and different vendors who, they didn't even like each other, aren't going to talk. Now they have to because they have the same standards. It also helps the customer because here's the thing, I could go from one vendor to another vendor and don't have to worry about the actual collecting the data. I buy the software for one reason. It gives you the best performance, the best use, the best cost, all those things all together, which in the past, it was like, okay, maybe I'll hedge my bets on two or three vendors because that's the way we do it. Now no, just choose one. If you don't like it, take it out. Bring the next one on because it's still the same agents.

>> It's a changing world. Talk about scale, how important and how big does the scale factor in? Because some of these pipelines are petabytes.

>> Absolutely.

>> Talk about the scale, the impact of scale. Why does it matter, when does it matter?

>> No, it's a great question and it's another trend that we've observed, is understanding what does it mean to scale when you are applying your focused on an observability use case where A, yeah, I've got to scale my ability to get data in and process it. That's actually the easy one. Secondly, I've got a scale aligned to the economics of doing that. So a lot of the feature sets that we focused on in our observability solution has been to help customers actually focus on optimizing, is that data necessary to accomplish what you want? Can you aggregate it, can you filter it? Can you put it to a lower cost data store so that you can balance the challenges of technically gathering the data and processing it with the constraints people have on doing that at the right economics?

>> Can I add to that?

>> Sure, absolutely.

>> One of the things that people don't really realize, it's something called cardinality. I'll give you an example of that. When I grab a metric, I get a timestamp, I get a key and I get a value and maybe a name part, a dimension, we call it. Each one of those is what we call part of a cardinality. The thing is, when you get into billions of these metrics, the cardinality bloats, so our actually When that happens, you have to have the capability of handling that. A lot of vendors, they struggle. A lot of open source definitely struggles. So when you talk about observability, one of the things that people don't look at is not the number of petabytes that are coming in. What's the cardinality also? Because if you can't store it in a way that matters that you can retrieve it, then you don't scale the retrieval.

>> And AI, we know, eats storage for breakfast and lunch and dinner. You've got to feed the AI with the right data at the right time. This is where, again, OpenTelemetry opened up a lot of doors for me in my mind. Questioning here because Savannah Peterson, who's here with me today, was with Paul Gillan at Reinvent, I think last year. They talked about observability and one of the headlines of I had to pull it up and look at it, is, "OpenTelemetry promises the demise of commodified data work." I want to get your reaction to this. "The adoption of OpenTelemetry is going to allow developers to do more building of interesting analysis features rather than repeating commodified work." What's your reaction to that? Do you agree with that statement?

>> I think so.

>> Does it help developers? What's the developer impact?

>> If you think about it, the other movement has been around the maturity of DevOps, meaning what does a day in the life of the developer look like and how do you provide a high quality developer experience? And so if developers are now more, you build it, you run it, they're going to first want to leverage the same type of software they use in building their own software, so they like open source. And two, if that open source is of a high quality type and can be processed effectively to solve the problems they have, then they are focused on actually delivering the value that their company wants them to deliver and not focused on-

>> So they're more productive?

>> So they're more productive, exactly.

>> So the headline essentially was referring to managing probably multiple tool sets or getting in the weeds on integration or-

>> Even simplest possible thing. Ever since the beginning of computing, getting data into a system has been a real big problem, so that's commoditized work. Every developer in the world has figured out ways of doing that, and it changes from application to application. If you don't have to spend your time doing that, then you can spend more time on analytics and what the business is supposed to do, that goes away. So that's where-

>> I love the concept. Let's go through the scenario of AWS, Splunk together, okay, working with the customers. Cloud check, public cloud looking good. You guys have been there from the day one with Amazon, great relationship. Cloud scale, you guys done that, Splunk is A+ there, good. On-premise, cloud operations, hybrid looking good. You've been on-premise, check, A+ there. Edge is going to bring a lot of latency challenges to data movement. You mentioned data movement. How do you guys look at observability when you start bringing in edge devices in, whether it's handheld, wearables, IOT devices? We're starting to get into that distributed computing full end-to-end paradigm. Certainly with GenAI, we're hearing a lot more end-to-end. How does that change the observability if it does? Or is the game still the same on edge?

>> I mean, from my view, again, if you take a look at it from a data perspective, the game's still the same. Now, obviously you have more things that are sending you data that actually could be vulnerable for a number of reasons, so obviously you have to develop a strategy around protecting those. And then yes, if the customer experience is going to be impacted by either that edge processing or by a dependency on that edge processing, then again, you're going to want to have visibility into that so that you can proactively protect that customer experience.

>> What about security on the edge, too? It's obviously a big factor, too.

>> Yeah, and that is something we're working on. We actually created our edge processor last year and that would be deployed in, let's say traditional places to measure things like temperature, humidity, usage, just things that are just part of IoT. Now, security has to be built in, that's the main thing. You can't just have a vulnerability there and then have it propagate all the way to the servers. So Splunk, we as a security company, we do take care of that. Now, the problem comes in, what if you buy a third-party version of this because anybody can sell these things? If they don't have that built in, the only way to protect it is to protect it at a central place, because they can send you anything they want and it could be malware. But some place in the central that has to be monitored and dealt with.

>> Guys, great conversation. Last question is obviously we're here in New York for AWS's Financial Services Symposium. Impact, the finance, obviously they're on the cutting edge, one part of finance. They're spending as much money to get an edge as possible, go fast. Remember the old high-frequency trading models. To the slow banks who are regulated kind of, they are. There's a lot of innovation, it's a cutthroat, competitive market. Where's the advantage for the customers in working with AWS and Splunk as you guys roll out and try to prepare, get that foundation set for this big gen AI wave, which we all see coming? And it's a little hyped up right now, but you're starting to see practical solutions dropping in, but it's all going to be grounded in large-scale data management. Large-scale data intelligence, data will be at the center of it, which is what Splunk does.

>> I gave a talk a half hour ago on exact same question was asked. And I said, "Well, I can walk to downtown right now and get there in about an hour, or I could take a car and get there in five minutes." That's the same thing here. You don't need AWS and Splunk to solve these problems. But if you combine them together, this time to value increases, that's the main thing. So if you want to take days, months, years, yeah, anyone can do it. But if you want to do it in a fast-paced way where vendors take care of some of the infrastructure and some of the other plumbing, we'd call it, that's what AWS and Splunk can do.

>> And everyone has their own unique environment, too. They could walk a certain way or drive a certain way. Take the subway.

>> Yeah, absolutely.

>> So this is the key. Customization is coming, it's here.

>> Yeah, absolutely. And I like to think too, what excites me most about the AWS-Splunk partnership and really just seeing how it's evolved with our joint customers is that we're focusing now more and more on protecting innovation velocity and driving a higher quality developer experience. And I think as we continue together to innovate more ways of doing that, it's going to help every financial service customer in New York.

>> Well, you guys are the center of the conversation. Our research team, we have conversations all the time around, and we've been promoting and reporting on research around this. We're in a systems revolution right now. It's a systems reboot. Some call it the AI mainframe, I've heard that. We had a post on SiliconANGLE on all that today. But at the end of the day, it's still the same game. Compute, run some software. You got data in the equation, it's just going to look a little bit different, but it's still the same game. Data is at the center of it. You got software and you got hardware, it's just distributed.

>> It is.

>> Yeah, it is. And behind all of it remains the creativity. I mean, the innovation, the creativity, that's why we still love... I mean, I love where the company name Splunk came from, is to go explore those caves. And I think this third wave of AI is just opening up some really exciting new caves for customers to go explore.

>> Dan, Nimish, thank you so much for coming on theCUBE. Thanks for sharing the insight on the deep dive on observability. I'm John Furrier with theCUBE. Stay with us, more action from fabulous program here today at AWS in New York City. We'll be right back.

Dan Holmes & Nimish Doshi | Splunk

>> The AWS Financial Services Symposium, presented by theCUBE.

>> Absolutely.

>> Thanks for coming on.

>> Thank you for having us.

>> It also saves on cost because if you had to maintain multiple agents and multiple vendors at the same time, the cost is going to be tremendous.

>> Absolutely.

>> Nimish, scope the impact of what he's saying, because if OpenTelemetry becomes the industry standard and it looks like it's going to be on that track-

>> Yeah, there's even a mainframe committee now.

>> It's a changing world. Talk about scale, how important and how big does the scale factor in? Because some of these pipelines are petabytes.

>> Absolutely.

>> Talk about the scale, the impact of scale. Why does it matter, when does it matter?

>> Can I add to that?

>> Sure, absolutely.

>> I think so.

>> Does it help developers? What's the developer impact?

>> So they're more productive?

>> So they're more productive, exactly.

>> So the headline essentially was referring to managing probably multiple tool sets or getting in the weeds on integration or-

>> What about security on the edge, too? It's obviously a big factor, too.

>> And everyone has their own unique environment, too. They could walk a certain way or drive a certain way. Take the subway.

>> Yeah, absolutely.

>> So this is the key. Customization is coming, it's here.

>> It is.