 Hi, this is your Sapin Bhartiya and welcome to D3M, our topic of this month and topic of this month is observability. And today we have with us, once again, Justin Coppett, work marketing manager at Akamai. Justin, it's good to have you back on the show. Hey, it's good to be back. Good to see you. If I asked you, how would you define observability? I think one of the things we need to clarify is that it's not so much a feature. It's not so much saying you can purchase observability. You can go out and buy it. You can buy a platform that is observability. It's a condition or a state of your environment. You know, the term for observability goes all the way back to the 60s like a lot of things do. It's an offshoot of control theory and it's ways to see what's happening in complex systems that originates in industry and manufacturing or aircraft, et cetera. So the difference between that and something like say just monitoring is monitoring is a component of observability. So just like a logging in telemetry and actually the CNCF, they have their own category for observability and they put monitoring, logging, tracing and chaos engineering, which was a fun one that I was reading about a few days ago. And it's one of the coolest sounding things that I've heard in a long time. But yeah, so the biggest difference there is that monitoring can tell if your server has a problem whereas you can look at something like, hey, the CPU is spiking or even something as simple as this is up or down whereas observability you're looking to see not just that it's working, that it's how it's working. You need things to be able to get data, you need to be able to process and correlate data, root cause analysis and identify incident response. All that together in your platform is what makes observability. I will talk about observability's evolution as you talk about chaos engineering because tools, getting all the data is good but what do you do with the data? How do you build a culture? That's where we talk about chaos engineering or we can talk about site reliability and when we look at cloud, cloud native these things are not separate silos, they all work together, obviously play a very big role in security, it can play a big role in performance. So talk about the evolution that you have seen and then we can also talk about the scope, how you have seen that has expanded. Yeah, I think a lot of it has to do with what I talked about last time was on the show is just the sheer amounts of data that organizations are bringing in with observability you need to make sure you need to make sure your teams are able to work more with the data that they have and really the only way to do that quickly and concisely to actually keep up with the amount of vulnerabilities, the amount of problems and the environments are just exploding in size. So the velocity and the variety and the sheer amount of the data coming in is expanding the size that the environments need to be. So to get observability running across your entire platform which is often gonna expand multiple clouds, even hybrid cloud, you have to have a solution that is out of the box, you don't need to be a data scientist to use but you have to be able to have machine learning capabilities but it also needs to have the flexibility in case that organization does have data scientists involved in their organization because increasingly we are seeing that happen more often particularly the larger ones but starting to trickle down to the smaller ones as well and the flexibility they can bring their own machine learning models and they will be able to fit into that platform. SaaS companies themselves are racing to gain share in the market which is, I think it's $22 billion market right now which is growing pretty much as I speak. I can see the dollar amounts going up and they're starting to shift to different pricing models as well. I think that's another big thing that we're seeing starting to happen where they're doing consumption based pricing models instead of usage based. So subscription plans were almost discouraging from IT leaders from actually monitoring all their data. They were often only monitoring it at an application level. So in addition to that, we're also seeing way more Kubernetes and serverless monitoring tools that are starting to be used quite frankly a lot more and some of them aren't necessarily for Kubernetes or serverless but Prometheus is one that we're all very familiar with but that's just starting to expand starting to get more attention and it's starting to evolve just like Kubernetes itself came out in 2015, the 1.0 but it took a while for it to start to become more mature and a little bit more accessible. So we're starting to see that with observability platform tools as well. If you look at observability and look at some existing practices like XIME or security information and event management what is the difference there? Because once again, there are a lot of things a lot of new technologies coming or there are new labels actually with the existing practices. So can you talk about the difference there? Yeah, absolutely. So it comes down to yeah, there's a ton of overlap. Of course there is and that's one of the reasons it gets floored a little bit. So if observability is essentially the state of your environment you are able to achieve observability something like a SEM is really a practice even though we most often refer to it as like a platform you can buy as like a SaaS solution. I'm gonna buy a SEM platform. It's really a practice, you know the term goes back to I think 2005-ish give or take like a lot of things in our industry that really started to bloom up. It was a combination of security information and security events management. So a SEM tool that enables security information and event management is we're talking about collecting and analyzing log data from really a wide range of sources. So making sure you're not just getting stuff from all of your servers getting things from your networking getting things from any device that's plugged in that you have an ability to capture and really wanna be able to manage these data and events. Again, it's a practice and it's part of observability but not exclusively, they're not a one-to-one because you also have SOAR platforms, SOAR stands for I'm gonna quiz myself security orchestration automation response. There's and that's pretty much the name is in the title they're able to get that sounding really cool without stretching it too much. That's when we're automating the incident response. So we're figuring out what's going on we're able to automate who's supposed to do what what systems are supposed to be involved. Then we have XDR tools where we have the classic X is used as the abbreviation here extended detection and response. I looked at my notes for that one because I forgot what it meant but that's where we're looking at correlating data from all of the sources you get. Now none of these individual platforms or practices are observability themselves they're all a component of it. So that's probably the biggest difference and I understand why there's overlap because you can't have one without the other with a lot of these technologies or practices of course there are different users in different stages of the journey. A lot of folks they have already embraced these at a very early age. There are a lot of them which are actually champions of this technology but then there are a lot of folks which are at a very early stage and when we look at Akamai or Linode I mean you cater to almost the whole spectrum. How much adoption of observability you are seeing which is already there or when you interact with customers where you see hey these are some of the either challenges or some of the struggles where they are kind of hesitant in embracing these practices or you're like no everybody understands this very well and everybody has embraced it. We are actually on the second phase of observability adoption. Oh it's it varies so much by I don't even want to say industry but by company to company almost department to department to depend on who the CEO is and the philosophy even if they don't use observability by name. There's really no one else no one out there at this point who isn't at least looking at some type of a SEM tool or isn't looking at some type of a SOAR platform because you know every time ransomware comes in the news these increase substantially. I mean that obviously increases the need for disaster recovery but observability comes into play as well and the biggest one is trying to tune your application and figure out why and how downtime happens. So there was actually an article this year published by Morgan Stanley. This quote came from Melissa Knox she's the head of a global head of software investment banking said digital businesses can lose upward of $5 million per hour when their applications or infrastructure are down. And that is enough to get the ball rolling on projects that departments otherwise would have had difficult to justifying security disaster recovery security even basic antivirus software they're like insurance you don't they don't generate revenue they prevent you from losing revenue. So when you're able to put a dollar amount behind it that helps get things moving. So the focus and now is on predicting what will happen to be able to prevent outages and downtime and poor experiences and we're also seeing that come a lot into AI. So AI and observability starting to roll into one of course I mean AI is just an octopus it's just touching pretty much everything in our society right now and it's understandable in a lot of ways. But yeah it really depends the back to your original question it really depends on the industry and how much they're affected and what they do. If they're relatively closed maybe they're a construction company and their services are only used by their own staff you're gonna see that a lot less even though there is more of a shock to their systems whenever they are similar industries have vulnerabilities and they're down for a while they still lose money. Whereas someone who's generating a massive streaming service data analytics platform oh they are all over it and they've been all over it for a very long time because they know that their lifeblood is cut the moment they have bad performances and they start losing dollars in the in the minis of zeros right away. And since you brought up the point of AI and Genetic AI I also want to talk about when we look at Genetic AI I want to look at it from two different perspectives one is as organizations are running Genetic AI workloads at the same time Genetic AI may help observability how do you see those two expect of Genetic AI and observability? AI is pretty much the only way to really especially when we're dealing with large amount of data it's the only way to really rapidly and reliably identify the patterns of the problems here. I mean that's probably the biggest thing we're seeing and we're starting to see that come in terms of something called AI ops which they're starting to use the term hyper automation this industry loves the term hyper hypervisor hyper scale or we love that word and you kind of used to it but so yeah I mean that is becoming its own branch which is essentially combining what Gartner actually defines it as combining big data machine learning to automate IT operation processes. So we're seeing that a lot so traditionally when I refer to traditional IT I refer to traditional development models and basic infrastructure standing things up making sure things are working and growing them at a relatively lighter scale whereas DevOps models were just rapid change very quickly, very quick so not only are you getting a lot of data from big data and data analytics but you're starting to develop more versions more software you're starting to put things out way faster than they did before so with new solutions brings new problems and then we look for new solutions again so there we are back to AI. So with AI you're looking at granular data feeds across multiple parts of an organization so you're able to even pull your own information to increase the amount of data fed into your observability systems be able to pull better conclusions from your own environments even locked in so it's not always AI isn't always just grabbing scraping everything from the internet but you have to keep in mind that especially larger enterprises even of a thousand people, two thousand people and especially up these are very disparate systems and there's a lot of legacy stuff so observability is being sort of a unifying umbrella to capture the information from this and AI is playing a big role in making sure that that information is actually useful and we capture sort of remove the humanity from that because quite frankly we're just too slow to be able to do that anymore it's unreasonable for us to be able to do that and it's one of those as we talked about last time AI in this part is enabling teams to do more with the data that they have and not replacing them so you need to sort of power them with AI ops and quickly discern those hidden patterns and figure out what's going on and that's where we're starting to see the era of hyper automation just in time infrastructure where data, data, colo, edge computing all that kind of stuff can really be put together very quickly depending on a business needs that was a huge thing in the economy when shipping containers, giant container ships came out in the 1960s where people had to order things instead of storing them in their warehouse we're gonna start to see that in the data center model and AI is able to provide insight in the complex systems and give course of action onto how to get the best performance and bang for your buck out of that for both providers and consumers. Since we are talking about some of the emerging technology like GenITAI and we initially talked about the whole evolution of observity where are we heading in terms of observity what do you kind of see in the future for or in observability? Oh yeah it comes back to more out of the box solutions we talked a little bit about that earlier but not everyone has data scientists everyone would like to have them and more people do but they don't so we really need teams and security teams in particular to be able to use SEM platforms or platforms and set up observability packages without having a degree in data science and we also need them to be flexible for people to do have data scientists in there so that is one thing they're shifting towards to capture a piece of the market usage space and consumption model pricing is a big one we're starting to see platforms go to and more talking about Kubernetes and serverless so it's something that's been around for a while but it's the uptick is steady it's steady for people using Kubernetes and some might even skip over that and go right to using a functions as a service platform that is often powered by Kubernetes on the back end and being able to get an observable ability platform that accommodates that and more familiarity with all the tools for people who are rolling your own in the cloud native compute foundation and chaos engineering which says we assume something was gonna go wrong let's plan it, let's see how it goes and see what the effects are we're starting to see a lot more of that so I think it's gonna we're gonna see more out of the box solutions that are also flexible that sounds like having your cake and eating it too but that's exactly what the market needs right now and we're seeing the pricing model shift so it's becoming more important it's certainly not going to go away but to have a more accessible solution to be able to achieve observability within your system that is growing increasingly more complicated but to also have more friendly models on pricing so that people aren't afraid to have every part of the environment watched because I have an anecdotal story where there was a company that got hit by ransomware they were using a provider that I won't name but it was so expensive they only had maybe one third of the servers actually being monitored and one of the ones that didn't got hit spread to the other ones of course East West inside the data center and that has significant outage so these consumption based models are going to be key to wider adoption and small medium business and of course enterprises don't have as much of a problem there but that's gonna be friendly to them as well. As you're talking about different companies are in different stages of their observability journey what advice do you have how they should build tools out there tools exist a lot of open source tools out there as you mentioned CNCF they have though and then there are vendors who help them but it's more about culture practices so talk about how they should build how they should approach observability and build the right culture for it. Yeah, first thing you have to do is you have to look at your team look at the team you have and not the one you wish you had and a more wish you had to accomplish the goal you have to see what their skill set is what people's cycles and availability are and that's where you're gonna make your decision is how much you wanna actually buy the off the rack solutions which work for a variety of people and how much you wanna start rolling your own to watch your own environments. You know here at Akamai we have to we build our own solutions you know every observability isn't something we're directly selling but we have observability into the platforms that we create for our own sake to keep them up and available but that's a different model depending on what you do if you're a manufacturer, if you're healthcare, et cetera so you look at your team you see what you can do and you see what usually the case is gonna be a SaaS provider quite frankly unless you're large enough to be able to have the team that's gonna be able to go in pull the open source tools and bring everything down so it's gonna be a mix it's just gonna be evaluating your team and the culture is similar to Zero Trust where it's something where you have to say keep it in the back of your mind and everything that you do what, how can you pull logs telemetry from every step of the application process and funnel it to a location where even if you don't have the platform yet you will have one that you will be able to use so plan everything you're developing for observability keep that as a philosophy as you build your entire environment now just your application has to be your environment to your infrastructure and that's where you go from there and build it around your team Justin, thank you so much for taking time out today and share your insights in the context of observability I love all those insights I would love to have you back on the show thank you Absolutely, it's great to be here and as always, thanks for having me