Latest / The MonkCast / Platform Engineering 2.0: Evolution, Not a Reset with Pankaj Gupta
Transcript
- speaker-0: This episode is sponsored by VMware by Broadcom. Good morning, good afternoon, good evening. I am Stephen El Grady. ⁓ I am the co-founder of Red Monk, ⁓ the developer-focused ⁓ industry analyst firm, and I am here today with Pankaj. Pankaj, would you care to introduce yourself? speaker-1: Thank you, Steve, for having me on your show. I'm Pankaj Gupta. I'm part of the ⁓ Broadcom VCF division. speaker-0: Well, that is a perfect introduction since we are here today to talk about ⁓ platform engineering and sort of what the what the future is for the platform. So as we know, like there there's a we'll get into this obviously over the course of the conversation, but there's a lot going on with ⁓ platform engineering, ⁓ you know, any number of trends, ⁓ you know, sort of our impact again. So what is your just to sort of open, you know, with a broad strokes question, what's your what's your sort of take on the current state of platform engineering today? speaker-1: Platform engineering is no longer optional. More and more organizations are deploying or have built a platform and have invested into platform engineering teams for that. If you look at the reports ⁓ from ⁓ Google Dora report, which came out last year and ⁓ December last year, we sponsored as Broadcom a state of platform engineering report with platformengineering.org, all the data suggests that more than ninety percent of organizations have adopted platform engineering. And adoption is continued to increase significantly of now the AI workloads are coming into that. However, few challenges still remain ahead of us. Adoption is good. But the maturity of platform engineering and use of platform by the developers varies from organization to organization. It's still a very significant road to go. Another thing what platform engineering teams have done, they have sold the value or delivered or articulated the value to their ⁓ IT and business leadership also for that. One of the major To what one of the ways you see the success of platform engineering is the when you start measuring the success of platform engineering. And the data suggests from 70 to 80 percent of organizations have mechanisms to measure the success of platform, whether it's a Dora metric or developer productivity or how fast they can take the code to the production variety of that. So in a nutshell, Platform engineering is really living to the promise of delivering the reducing the cognitive load on the application developer, making things much more efficient, bringing the golden pathization across multiple engineering organizations for that. Also improving the developer experience, improving the developer productivity. moving from the ⁓ ticket ops to the click ops for that, offering self-service portal for developers for that. And also platform engineering has also become a substrate on which ⁓ you can implement shift left because now you have the standard pipeline CI C D. So you integrate CI C D. So it's a very multifaceted ⁓ function. It is complex. Adoption is getting more universal, but challenges still remain ahead for that. Many advanced organizations have started managing platform as a product. So when you start managing in the platform, you start teething developer as your customers. What is their need? What is their requirement for that? So we are pretty excited about the platform engineering evolution and I think it is it is at the fore roads today in the context of AI. speaker-0: Okay. So, you know, when when we have conversations with platform teams, and as you say, you know, i i if they're not ubiquitous, they are sort of becoming close to that, right? It's just a sort of necessary ⁓ component of of sort of any technical organization. you know, we see them sort of hitting you know, e either hitting or approaching a ceiling. ⁓ you know, do you see that and sort of in your experience, you know, sort of why is that happening? speaker-1: I think last two or three years has been a big changes in the industry as you know for that. Okay. The originally when platform engineering started, it is it was a AI blind architecture. And platform is becoming a substrate to ⁓ build, deploy, and govern the AI-based applications for that. So that's the one ceiling the some of the platforms have started hitting, not being the AI-aware architecture or not able to enable. But the good news in the last two years, many platforms like VCF and others have crossed the red boundary of not being a AI, being an AI-blind architecture. So that's the one big ceiling is ⁓ hitting. Second big ceiling, which ⁓ Current platform has to evolve or go beyond its ceiling, is the developer-only focus. When the platform engineering started, the primary focus or primary persona was the developer. Now there are a variety of new ⁓ personas who will use and have a benefit from the platform. First and foremost is the data scientist. Next one will be the FinOps. The security teams. So platform engineering has to evolve to support multiple personas for that. And not just the human personas, now the agents also come into that. And I think platform engineering has been mostly reactive instead of proactive, especially in a few areas. One of them is, of course, the FinOps. You know what is spent has happened in the rear-view mirror. We should look at How spent is going to be in the future or with the new environment for that. Security has been a ⁓ bolt-on security. Security has been, or the compliance has been more rear view mirror rather than the forward view mirror into that. And there is another fundamental shift which is happening for certain use cases, especially for AI. The golden paths are excellent and will remain a foundation for platform engineering. But in last two years, where a lot of experimentation and learning is happening for AI, and every six months or every two weeks, there is something new that comes out. So you need a lot of flexibility, experimentation, fast fail, fast learning. So golden paths are good for a standard application, but when you're learning significantly, sometimes golden paths start feeling like the golden cages. So you need a lot of flexibility into that. The current platform engineering or the original platform engineering were more focused on rigid in a static platform. If you talked to me last year, we would have not talked about MCP servers. We would have not talked about the AI gateway. And then just all the new building blocks coming up and there are choices of that. So those are some of the ceilings for that. But many of platforms have already started moving to the platform engineering 2.0. And that's the reason Dotcom worked. So shaping the narrative or shaping the where platform engineering has to go next as a platform engineering 2.0. But I will reinforce that the next iteration of platform engineering is not a reset, it's an evolution. It builds on the fun fundamental principles of platform engineering, improving the productivity, standardization, reducing the cognitive. load for that, managing platform as a product. So all those fundamental principles of platform engineering is still remain extremely valid and will be there for a long time. speaker-0: And so for the next question, you've already touched on this. ⁓ you know, obviously everything in this industry at this point in time is being impacted by AI. and platform engineering is no exception. ⁓ so in in your view, where does you know, where do AI and and platform engineering intersect? speaker-1: I think it's a very strong intersection and timing is absolutely right for that. And let me go a little bit more deeper. If you talk to any developer today, most of the developers are using the AI for writing the code. So there is a significant AI-driven coding acceleration for that. In independent of how big organization is the developer or how many developers are there for that. More and more developers are using AI tools, and AI tools are extremely effective, highly productive for that. But it has a two major implications for that. Previous bottleneck was that writing the code. Now writing code is becoming like a commodity, or you can create much more code. So the Bottleneck is shifting from writing the code to taking code to the production faster. The organizations who did four releases in a year, now they can do 40 releases in a year for that. Second big point, really, which really comes with the role of developer, is changing. Previously, developer was more focused on writing the code. Now they are more focused on validating the code and taking code to the production faster. And a much scalable way into the ⁓ deployment. Second big impact which AI is really building bringing is the agentic future. Today, maybe 30 to 35 million developers who are going to use the platform or will use the platform, but the agents have already outnumbered the developers or going to outnumber. It's pretty likely next year there will be 100 million. Agents will be using the platform. They will require govern they will require the same infrastructure, they will require API, they will governance for that. So that's a huge part of that. Next big AI is really driving is the last six months you saw the conversation changing from token maxing to the token economics for that. It is top of the mind for that. The last one which AI actually ⁓ drove is driving platform engineering a lot is the sovereignty and the compliance for that, putting the guardrails over here. So AI has an implication, changing the developer persona and behavior, changing the frequency of the code development, bringing the new developers, it has an impact on FinOps, it has impacts on the security for that. And the bigger impact is that. AI gives a brand new persona, which platform has to cater with is the agents for that. So this is the perfect storm, and this perfect storm cannot be at more appropriate time where it's a lot of other things are happening. The sovereignty requirements are growing for that. You have seen the hardware pricing is increasing so much. So you see the how AI is creating the ⁓ DRAM pricing challenges in the server cost for that. So now the AI is also driving FinOps. Previously FinOps was bolt-on. Now it has to be integrated. So wherever you see that, AI has impact which is just avalanche across the whole platform engineering. ⁓ speaker-0: So, you know, you know, Ponakash, you you mentioned that the, you know, the sort of ballpark hundred million number could be less, could be more. ⁓ you know, who knows? It's a ballpark. The net of it is we're gonna have a lot of these agents running around ⁓ interacting with these systems. So how do you think about this as a persona, right? We're you we're used to personas like de ⁓ developer, operator, admin, et cetera. This is a new persona, you know, from an agent standpoint. So, like, where does the platform fit in here? speaker-1: So agents will be using the platform for writing the code, executing the code independently. So that's the biggest, biggest impact. And their frequency of writing the code and executing the code may be exponentially higher than what the human can do that. That's the first part. Second part, they will have the unique needs for that. They will require the APIs, they will need the models, they will require More importantly, what they will require is the traceability. They will require very strong guardrails for that. You heard a few months back it brought in one of the company's infrastructure down for that. You saw an example that one of the agents deleted their customers customer data for that. So the the requirement for guardrails is going to be so significant over here for that. observability is going to be very huge. The audit is going to be pretty huge for that. Because there is no human interaction. The agents can do whatever at 3 a.m. for that and then you have to figure out what happened for that. So that's the reason there is so much brand new conversation in the industry for last eight to twelve months about the MCP server, the MCP ⁓ definitions and their requirements and the ⁓ Execution is just changing or evolving to bring a lot of guardrails into that. Same is conversation associated with the AI gateway. So now platform engineering's biggest job is to understand that what is the roadmap for their agents look like? What are their their unique needs for that? What is the guardrails they have to ⁓ put it? How they are going to troubleshoot it, how they are going to put the traceability or audit. Audit is going to be huge, huge for that. And all these things have to be done at a scale. They're not just 30 developers or 100 developers who are using this, is the agents are doing which is autonomous for that. So I think this is a very big ⁓ persona. And this is the first time platform gets a non-human persona. Okay. speaker-0: Now, you know, when we talk about that, how do we engineer for that? And and you and I have have sort of talked about this, you know, sort of as as five pillars. Yes. So, you know, we don't have time to go through them at any level of of great detail, but can you give us sort of a quick sort of recap of of the you know what you consider the five pillars to be? speaker-1: The first one is the AI native platform. So the platform has to support AI workloads and agents natively into the platform. It could be the GPU provisioning, self-revice for GPU, it could be the MCP server, it could be the AI server, the virtualization list goes on and on for that. Second one is the platform has to evolve from a single developer focused persona to multi-persona experience for that. Which includes the agents, of course, as non-human. It will include the business leaders who are going to look at ROI, the FinOps cost predictability, ⁓ the security and the compliance team who are going to move from a static compliance one is in a year to every day or every hour compliance or every action compliance. The third one is the platform engineering has to evolve from a bolt-on FinOps to more embedded FinOps. So putting the gates when you provision the environment for that and constant monitoring, not just the regular workloads, but also the AI workload. And a stronger security to complement shift left. Security shifts down into the platform. Security comes into the runtime for that. And how AI is expanding the security gap that has to be addressed for that. That's a brand new attack surface for that. And the last one is that composable by design. The platform has to evolve to support new components. And more and more choices is going to come into here. So you should be able to replace one CI-CD tool with another CI-CD tool, one model with another model, one gateway with another gateway without effects going through the rest of your. stacked for that. It is a tall order, but those are the five things together frame the platform engineering 2.0. speaker-0: Okay. So that all makes sense. ⁓ as we wrap up though, you know, I would like to I'd like you to leave the audience with sort of your ideas. Like if if I'm a platform engineer, sort of what are the next steps? Like what should I be looking at? What should I be thinking about? speaker-1: I think there are two things. ⁓ f first and foremost is ⁓ familiarize yourself where the platform engineering evolution is happening. So ⁓ Broadcom has published a ⁓ platform engineering two dot zero, a white paper which is a vendor neutral, talks about the industry trend, what the platform engineering has to evolve. That's one, it is published on our website. Just search platform engineering two dot zero. Second ⁓ action. Is that go back to your organization and look at these five pillars, which is has the more urgency for that? Most probably every organization is looking at how they enable the AI workloads on this platform. Look at is it the AI has to be supported natively, if they have already done it, how they are addressing multi-persona, or if they are behind on FinOps, figure out where your starting point for that. But start looking at platform in a much broader perspective into your organization for that. Also bring infrastructure teams with you. The people who manage the servers, the compute, the networking, the storage with that, because they remain the foundation for every ⁓ platform. And keep learning because this field is going to evolve for the next two or three years very significantly. Platform engineers have the right foundation and the already know many of the things with platform engineering will continue to deliver and continue required to deliver. speaker-0: Can't close on a better note than that. ⁓ because particularly in this day and age, as quickly as things are moving, you have to keep learning. ⁓ so with that, we'll we'll close out. I really appreciate everybody's time and ⁓ Ponkash, thank you so much for joining us. speaker-1: Thank you, Steve, for having me.