James Higginbotham: Hi, I am James Higginbotham. I'm the author of "The Principles of Web API Design" and founder of LaunchAny, an API consulting firm. I'm here with Vaughn Vernon today. Vaughn is a software developer and architect with over 35 years of experience across a broad range of business domains. He's also a leading expert in domain-driven design, and a champion of simplicity, which I love. In addition, he's the founder of an open-source product or platform called VLINGO XOOM, which is a set of DDD-friendly distributed computing tools. The tools simplify concurrent, reactive, event-driven, and microservice architectures, and it's all built on the Java virtual machine. Now, if that's not enough, he's also the author of the book "Strategic Monoliths and Microservices," as well as an editor of a new Signature Series from Pearson and Addison-Wesley. Vaughn's here today to share some insights from his new book. Great to speak with you, Vaughn. Let's start by having you tell us a bit about your Signature Series with Pearson Addison-Wesley.
The Pearson Addison-Wesley Signature Series by Vaughn Vernon
Vaughn Vernon: Sure. Thank you to GOTO for inviting me to be interviewed, and my suggestion to have you, James Higginbotham, interview me. It's good to be with all of you. Well, I would say that my Signature Series has the first goal of trying to help developers, software developers of all kinds, to mature. And by maturity, I don't mean simply knowing more about some programming language. Of course, that's always important. But my thinking on maturity is more around having a business-centric focus on software development, where software developers are less drawn to tinkering with new toys and hacking code, but more about learning what the benefits are of drawing close to the business, having a good partnership with business stakeholders, and developing software that has strategic value.
In doing that, the series also emphasizes organic refinement, meaning that things emerge as we learn more, and we look at software projects more as learning experiences and through experimentation and discovery.
As we learn more, the software gets better. As we go through that, we're specifically looking into topics such as reactive object and functional-based architectures and programming, as well as domain modeling and, of course, including domain-driven design, creating services that might fit what would be called microservices by some, but again, demonstrating purposeful architecture around that. Also patterns and like your book, James Higginbotham, APIs on the patterns that you talk about in your book.
We're trying to promote the best uses of such tools and any underlying technologies that match up with those. So really, an architecture and design focus where technology and programming underpin that, but those things are not the primary focus.
James Higginbotham: Yes, that's fascinating. A lot of the things that you mention, you even call that out, it's not just about programming languages. I mean, there's an aspect to that, but there's more to it. You're helping the development community move beyond just the technology. And it sounds like you're helping them engage in the business side of things, in the product thinking side of things, other elements of that. Is that right?
Vaughn Vernon: Yes. And, you know, James, as we've spoken about before in my interview with you on another podcast, we're quite interested in learning what the business needs are, whether we're consulting first and then training, or training-consulting, whatever it happens to be, we're looking for ways to improve the business value. And not just showing developers new techy type things that, of course, it is technology. But yes, trying to improve the practices first.
James Higginbotham: That's great. I was excited about this Signature Series when you mentioned that you had this opportunity. I think those that are probably familiar with the Signature Series from Pearson in the past with Martin Fowler's series and others, I think there's a lot of people that have taken advantage of those and matured their career and their thinking and changed the way maybe they've approached software design, systems design, solution architecture as a result of those. So this is going to be a great series. I've seen some of the titles already and I'm looking forward to some that are still being written as well.
Vaughn Vernon: Yes. And it gives a great opportunity to introduce new authors sort of under maybe the wing of the series, where maybe they aren't as well known to the community, but they should be. Being in the series helps too, I guess, both validate, but more promote the ideas of other people and show that there's a lot of wisdom out there to be shared. And thank you again for being an author in the series.
Who should read Strategic Monoliths and Microservices
James Higginbotham: Yeah, thanks for the opportunity. It's an honor to be part of that series. Let's dive in to the book now. Your book is "Strategic Monoliths and Microservices." Who did you write that for? And why did you write it? You know, what drove you? I mean, writing a book is a passion project that takes a lot of energy, a lot of time. So tell us a little bit about why you wrote that book.
Vaughn Vernon: I've noticed and probably, mostly since my "Red Book," the "Implementing Domain-Driven Design" book that there continues to be a big chasm maybe or gap between business people, you know, those who are primarily interested in the innovation around their business and developers. Developers latch on to domain-driven design. They want to learn it. But what I found is there's always just sort of this still very technical bent to those who want to learn DDD. And even when given a clear opportunity to interact with business people, they tend to either shy away from it or decline the opportunity. They'll say things like, "No, we've got this." So, it's still this kind of, like, closed circle of developers who are trying to think about the ways that the software should work without really engaging with the business.
While making those observations and writing books that are primarily for architects and developers, I started to think, "Well, wouldn't it be nice if executives had the opportunity to know, at a slightly less technical level than, say, architects or programmers of various levels, where they could understand the kinds of decisions that could be made, and the possibilities without actually stepping on the toes of developers at the same time?" And I just want to make clear that the book is for C-suite, senior VPs, VPs, directors, managers of software projects. But it can range through chief architect, senior architect, enterprise architect, senior developer, mid-level developer, even junior developers, because these are the kinds of things the mindset that needs to occur to be, developed in software teams if they truly want to innovate.
The book is meant for a range of roles, but it's written in such a way that certainly is compatible with C-level executives. And I just have to say, too, that C-level executives these days, especially in technology, are way different than they were 30 or 40 years ago, you know, sort of when I was getting my start in software. And I talk about the sort of Wall Street or, you know, midtown Manhattan look with pinstripe suits, and vests, and all this stuff. There's not anything wrong with that, but, you know, when you really look at the founders today of startups, and even executives and, you know, large enterprise or fortune companies and global companies, they have a completely different mindset than that. And they're more casually dressed, but also, they're very technically savvy. Tomasz, my co-author, and I didn't in any way dumb down the book. But we made sure that we didn't dive too deeply into bits and bytes and things like that because that isn't mostly about an innovation most of the time.
James Higginbotham: Wow, that's a great point. And I've seen that as well in the organizations I work with. The C-suite, they're deeply technical, we're having deep technical conversations with them. They're sitting down and, at least in my experience, drawing diagrams and things and trying to work through the problems at a much deeper level. So having a resource like this that helps to unify the thinking and to speak to that broad audience allows the C-suite down to the junior developer to all be kind of speaking the same language, thinking about the same things in the same way. And that seems to be a really powerful way to approach a book like this.
At the frontier between business and technical strategy
James Higginbotham: So, "Strategic Monoliths and Microservices," I think that's a really interesting title. It feels like a blend between the business or technical strategy worlds and some of the architectural decisions we have to make. Why did you choose that title? Did it come to you instantly? Was it something that took a little while to emerge? Tell me a little bit about the origins of the title of the book and kind of the general, you know, vision or mission of this book and what you're trying to set out to do and how all that ties together?
Vaughn Vernon: Yeah, great question. First of all, titles of books, that's a really interesting subject in itself. Actually, "Implementing Domain-Driven Design", my first book, the "Red Book," I had that title from the beginning, and it stuck. I think that this book's title changed about five times before it was ready and we chose one to use. It took a lot of ruminating over time to consider all the angles and so forth. We knew that we wanted to write about monoliths and microservices, and sort of the overarching ideas of strategy and innovation.
I think that what we finally landed on was both the main title, "Strategic Monoliths and Microservices," and a subtitle that kind of harmonize. And the way that they do that is "strategic" and the subtitle "driving innovation" go together. And then "monoliths" and "microservices" in the main title and using "purposeful architecture." So those two titles are I guess a bit parallel and woven together. So that was the idea is put more emphasis on strategic innovation. And, of course, software architecture is extremely important, but you can't have an architecture unless you know what software you're developing. And when someone rushes into architecture because this is what they love and they're just sort of, like, trying to fit things together that they assume will be necessary, I think that often there's a lack of real thought as to purpose. So trying to emphasize that first comes to the strategy, the understanding of how we will be innovative, and then what do we need to make that happen in terms of software architecture?
What does digital/business transformation stand for?
James Higginbotham: In the first chapter, you seem to focus on kind of three areas at least that stood out to me as business transformation, software architecture, and the inevitable element of change. And you seem to have kind of set up the whole book from at least those three elements. That's what stood out to me.
We talked about kind of how the title was inspired and what you're looking at. But let's talk a little bit more about what people are experiencing in their day-to-day. Every organization's going through a digital transformation of some sort and we're all having to deal with changes and evolution of software architecture and the change throughout. So how are these elements connected? How are you tying those together in the book?
Vaughn Vernon: Yes. So definitely every company that I engage with is going through a transformation, digital transformation. And what I find, though, is that sometimes digital transformation is simply like a lift and shift to the cloud. And, to me, that's actually really not digital transformation. And I think where innovation is really a sought-after goal, most people wouldn't really consider that innovation or digital transformation. Like I say Amazon is sort of making their enterprise computing stack available to the outside world, that's innovation, right? I mean, they had it, it was working, it scales tremendously. And they say, "Okay, then let's sell timeshares for this, basically, right, usage, metering for the use of this." Okay, that's transformation, definitely innovation.
But, subscribing to that, spending money for it, and just moving your same software over there is not digital transformation. It's just sort of, like new hardware in a different form or something like that. So, I like to emphasize that digital transformation is all about innovation, that you're trying to transform in areas where either the digital aspects of your business were simple, like too simplistic, not carrying enough business weight, or there are still manual processes within the business, and we're trying to take every situation there and turn it into a digital experience that will drive big revenues for the companies that engage in that process and take the risk in investing in it.
The four main pillars of the book
James Higginbotham: Given that background about digital transformation and the view you're seeing of how software is having to change and organizations are changing now, can you walk us through the four main topics or parts of your book and talk about how they give everybody a chance to understand? What does your book cover? What are you approaching in this book?
Vaughn Vernon: Sure. And actually, I realized I didn't fully answer your question. So how do architecture and change fit into that? Well, if you're going through a digital transformation, transformation doesn't happen overnight. And certainly doesn't happen even extremely quickly when you're moving from sort of this big ball of mud, you know, tangled legacy system. So you need to have an architecture that will facilitate change. And even if you're starting kind of, you know, from scratch in a lot of digital transformation areas, like really innovating, you still need an architecture that facilitates change. Right? So hopefully, that sort of answers the first part of that.
The four sections then are: how do we look at the transformation in general? That's kind of the first part of the book, and how do we develop the right mindset for innovation for actually engaging meaningfully with business stakeholders, and software developers, and, you know, users or customers of the software that will be eventually available for use so that we can make the best software and create breakthroughs, realize breakthroughs in what we offer? Then some tools that can be used to do this in a very lightweight, very flexible, and tools that support agile as in agility. And we kind of, you know, in the book kind of steer away from the heavyweight, what is considered agile, but where it tends to have a lot of ceremonies, you know, associated with it. So we kind of formulate that in the first part.
The second part is about domain-driven success. So domain being identified as a sphere of knowledge, but also the business domain this sphere of knowledge about the business, but then in smaller parts of the business where we're focused on innovation, what do those spheres of knowledge look like? What's driving us to try to innovate within those? But then also balancing that kind of investment with a recognition that all software is not strategic. It's necessary, but it might be supporting or generic. And so, how do we balance the kind of investment that we need to make without putting too much money and time into areas? Or at least not the human capital time that you know, the business needs to use for these innovations, but maybe spend some money on off-the-shelf or download open source, you know, components and so forth to support that. So there's a good balance there to be found.
This is where we do talk about domain-driven design, but we delay the introduction of the term "domain-driven design" because we didn't want higher-level executive management and so forth to hear yet another one of these, what might appear to them to be buzzword. While you, and I, and others understand that it's not a buzzword, it's a way of thinking and an approach to software development. But we kind of sneak up on them a little bit in that sense and describe what we're doing and why we do it, and then say, "Oh, and by the way you know, Eric Evans wrote about this, whatever, nearly, I guess, 18 years ago or 19 years ago, now I don't remember for sure. And this is what it's called, and here are the parts of it." And then we just don't, like, keep saying stuff like that. We just show them how to use it.
Then the third part of the book is event-driven architectural or event-first architectures where we're talking about how events are extremely important to the development of any software. And I'll get back to that in a moment. Then what are the software patterns, both whether it's web-based, or messaging-based, or whatever it happens to be, what are the technical tools that we have, and the patterns that can be used to highlight and accommodate events as the first-class idea in software development, and how that helps with innovation?
The fourth part of the book, pardon me, is about monoliths themselves and microservices, but taking kind of two approaches to this. One is "Monoliths Like You Mean It," right, is the name of one of the chapters. What that means is if you're going to develop monoliths and you find this to be a purposeful architecture, this is how you want to go about doing it so that it doesn't turn into a tangled, you know, big ball of mud mess. And what if you already have a tangled big ball of mud mess but you need to transform that into a well-modularized monolith before you do anything else, how do you deal with that kind of situation?
And then the next chapter, it's in fact, chapter 11 is introducing microservices like a boss, right? So this is where you have a good reason to use microservices. Perhaps this includes using both monoliths and microservices in the same system solution. And then how do you go from, you know, the tangled big ball of mud and reach microservices? Well, you might first want to go back and re-modularize or modularize that big ball of mud using the advice in the previous chapter. But if you're brave enough to sort of just bulldoze forward with, we're just going to go right from the big ball of mud to microservices while we discuss that as well, and how you can be successful with that. And we've, Tomasz and I have both had experience with this. So we feel that we provide very good advice and techniques for doing this.
Finally, we kind of wrap up with this, what I think is a very sort of poignant last chapter, chapter 12, which requires balance, demand innovation, right, or demand strategy. So it's saying again, let's just sort of review this in a few words. Remember that architecture is providing a purposeful means to develop software innovation on a kind of foundation, but the innovation itself needs to happen for there to be any reason for software architecture to even be a concern at all. So those are the four parts.
Why are events important in software architecture?
Vaughn Vernon: Now, I said I would come back to that whole idea of events. And a really important thing to remember is that we introduce, you know, Mel Conway's, you know, the amazing brainchild of Conway's Law, what's become known as Conway's Law, from the year 1967 didn't get any daylight until 1968 and then Fred Brooks recognized it as, "Wow, this is really important work, you know," and included in the "Mythical Man-Month" book. And since then we've really learned, pardon me, that Conway's Law is undeniable, it's there, it's like the law of gravity. You don't get better at the law of gravity, you simply have to work within the bounds of the law of gravity, realizing that if you fall even a few meters, or yards, or whatever measurement system you're using, you can get very hurt or even die from such an impact.
So, given that, what do we do with the law of gravity? Well, we know that we can only jump and stay in the air so long. So we learn to coordinate certain things. Like if we're playing basketball, or football, or, whatever it happens to be, we learn how to use our knowledge of the law of gravity and the feeling that we have in our bodies for, you know, how long we can sustain some move, or, you know, airtime, or whatever it happens to be. And in the same way when we're aware that communication within an organization will have a direct impact on the software or the system, actually any kind of system, but a software system, in this case, and the outcome of that, then we better make some very good decisions about the kind of organizational communication structures that we put together.
And what's amazing to me is that Mel Conway not only had the knowledge to observe that organizations will create systems according to their organizational communication structure, but that he said, "Okay then, this means that you should basically incentivize leaders in your organization who are flexible and create flexibility within the organization communication in order to do this." So, okay, what does this have to do with events? Well, events are all about human communication. They are.
What happens when I finish talking to you, James Higginbotham? When I finally run out of breath and you can talk again, what happens? Well, it's sort of like an event, right? You recognize that there's this Vaughn Vernon -stop-talking event, and now you're gonna react to that. Well, it's the same way with software. When various things happen in software, we want to record as a communicative record that this occurred. And because it's an important occurrence, then even, the monolith or microservice that we're building, or others throughout the system that we're working with can consume that as a means of telling them that this happened, it's a form of communication, and our communication itself is being relayed around the entire system and the system can react to it. So this is why we focus on events through the book, and the tools that we provide in the book are very much event-centric. So this is, I guess maybe more than you wanted me to say about the parts of the book and Conway's Law and events. But there you go, that's the whole kind of, you know, thing there, the whole idea.
James Higginbotham: Yes. No, I have witnessed quite a few organizations that think in a synchronous fashion, they think in request-response, they think in calls. And then it tends to get them in trouble, particularly when they're going down a microservices path. Because, as you said, they maybe have a big ball of mud, or maybe they're just trying to transform the way that they do things or trying to scale up. They're trying to reorganize the way that they work. And so they break things, decompose big problems into smaller problems.
But they create this level of fragility, as a result of making a call to a service, which makes another synchronous call to a service and they create these call chains. The result of that, then, is instead of a single process with code that's chaining calls together, we're chaining calls over the network. And anywhere along the way, in that call chain, we could have a network problem. We could have something that's not available.
How can events make organizations more resilient?
James Higginbotham: So can you talk a little bit about how events really help to transform the organization into being more resilient, less fragile by shifting toward an event model, thinking more in terms of events rather than synchronous call chains and service calls,s and encountering these kinds of problems?
Vaughn Vernon: Sure. And actually, what you said is a really good way to highlight that communication is, again, a very fundamental part of this. Because life life is not a podcast, right? I mean, right now, you're waiting for me to stop talking. So it's sort of like that service call service, right? Where you're waiting for me to stop talking so that you can do the next thing, or I'm waiting for you to stop talking, ask me a question, and then I do the next thing. But human life doesn't work that way. Rarely do I say to you talk to someone about some topic, and then I literally wait for minutes, hours, days, weeks, months for you too, like, fulfill that request that I made, because it might take a lot of time for that whole, you know, your part of the conversation to conclude how productive would that be?
So when you just think about just normal human interaction, events make perfect sense because events help us to, in essence, embrace latency. Latency happens on the network. Even you might think that you know, well, we're deploying to Google and Google has the fastest network on Earth. Nobody can\, do better than Google. But talk to Google about how much latency they have. I mean, I'm sure they're super-fast, but I will guarantee you that their engineers, they're trying to squeeze more out of it, trying to find ways to make it even faster and more reliable.
So given that, when you capture an event and you persist it, and that is successful, then you have this long-term, even forever kind of record that can be distributed once to other parts of the system, or it can be distributed time and time again to newer systems that didn't exist before. When that distribution occurs, as in a broadcast or publishing such an event, you simply have to know that the event was published. If you know that when the event was published, you kind of can now relax and forget about it. Right? So if you ever think of software relaxing, it's probably more about our brains are relaxing. Okay, I know that it's persisted and I know that the messaging mechanism has this and it's going to guarantee me that at some point it's going to be delivered, and now, given that at some future time, some other component will react to that event and do something about it, and I might be interested in the fact that some conclusion was reached.
I would want to know about the event that happened in the other system, but I'm not worried about setting timeouts for REST requests. I'm not saying, "Well, if this REST request doesn't succeed within five seconds, we're done here, right? Failure." And how many cascading impacts will that have, potentially across a large part of the enterprise? Instead, we're just saying, "Well, we know that this worked, and we know that we're expecting something back, but we'll just react to it when we see it." Then if it happens in 6 seconds, or 16 seconds, or 60 seconds, as long as that's within an acceptable business timeframe, we're good, right? We're not worried about, you know, user requests stacking up on our web tier waiting to be handled, because this thread is tied up on waiting for something else and we can't let it wait for more than five seconds, otherwise, we start experiencing other problems upstream. So it just really changes your whole mindset.
And the resiliency part comes in simply because if there is a failure in delivery then, for example, that event will be eventually delivered. Given that the cloud, many companies are very much dependent on the cloud now, if there is a failure in some machine or part of the network, we can bring up nodes, whatever it is, or make a function available or a lambda available in another area to handle that. Well, you're going to get some very kind of welcoming and naturally-feeling resiliency. Of course, this is a bigger topic than that. I don't want to just oversimplify it. But that is the basic idea of the benefits.
Shifting from asynchronous call request-response to event-driven reactive
James Higginbotham: I loved how you said, "software relaxing." I had this picture, this microservice on a recliner kickback, and just kind of waiting for an event to come its way. But no, I fully agree with you. I've worked with event-driven architectures in the past. I first got introduced to them with message-oriented middleware and accusing topics on message brokers and things in connecting different enterprise systems. But having worked with a lot of organizations that have these, either long-running processes, as you said, where it could take a while before we can fulfill the task at hand, the work that needs to be done, or maybe just, as you said, the increased latency of dealing with cohabiting software with other people on shared hardware in the cloud, or dealing with crossing boundaries across regions, or just simply some system gets overwhelmed or overloaded, and therefore would have a negative impact on something that's working synchronously, to shift ourselves to thinking asynchronously and to start to embrace that as an architecture to make that shift will take time but has a lot of benefits to it.
Do you have any examples or something you can share, even if the parties remain nameless, of how you've helped people to transform that mindset from asynchronous call request-response, maybe a more REST-oriented way of building software internally, to a more event-driven reactive approach to doing things and how that, you know, kind of evolved and kind of the results and the fruits of that labor?
Vaughn Vernon: Yes. Well, I'm working with multiple teams right now in areas like that. Yes, it's a very different way of thinking. Again, the whole service call makes service call, it seems natural to developers because they're used to making function calls or method invocations and then those methods that are called make other calls. And it's like, well, that always works. Very rarely will it fail. And so it just seems natural, and it is a big mind shift for people to go through to accept that.
And what I try to do is just make the whole idea sound much simpler than they tend to consider it upfront. And when they hear about this asynchrony, "Wow, that's difficult stuff. Never done that before." Well, of course, they have. If you've ever written a web UI or an old Windows API app, even if you're using Visual Basic you have worked with asynchronous messaging, right? You do not know when a user clicks the mouse, or when they've selected some button, or click some button, whatever it happens to be until you receive a notification that that occurred. And you're not constantly saying, you know, "Hey, UI, did they click that button yet? Did they click, you know, an item in the list box? Did they scroll the list box?" You just receive notifications of those things.
That is just what we're talking about. So anybody who is in fear of this they don't need to be. It is far simpler than most people have the idea, but you do kind of, like, need to shift your mindset. As I said, if you send, you know, an event to another party and you're expecting something back within some timeframe, well, you simply remain inactive or doing other things until that occurs. And so, you have to have in mind that you won't know immediately the result of that. So you have some state somewhere that has been saved in.. persisted in some way, whether it's in memory or whether it's on disk, and so you can then recall that when you're notified about the result that you were interested in seeing.
I would love to talk about some of the concrete experiences, but it's been everywhere from just trying to get people to drop the idea of everything as a REST call, helping teams see that if they don't make those eventually consistent service-to-service-to-service calls that will fail and fail daily, not necessarily in large massive dose, but then you won't need to hire a team or your team will be much smaller and dealing with the inconsistent data that exists when you wake up in the morning or throughout the day because one of those services failed. It is really difficult to get people to understand that and buy into it. But again, I think it's mostly just the kind of fear factor of doing something that seems unfamiliar until maybe you can help them to see that it is not completely foreign to them.
What is purposeful architecture?
James Higginbotham: So let me hit on purposeful architecture, the stepwise refinement, then we'll go ahead and wrap things up by talking about the next book that you're working on. Does that work?
Vaughn Vernon: Sure.
James Higginbotham: Ok. All right. So you had previously mentioned the idea of purposeful architecture. What is purposeful architecture? Tell us a little bit about that.
Vaughn Vernon: Well, it means that the architecture has a design that facilitates the software that the solution to the business problem that you're trying to solve. This means that it's reflecting what the business needs. So, if a developer goes in, or an architect goes in and starts making architecture decisions early before they've even engaged with the business, it's like don't confuse me with the facts, I've got my mind made up, right? You're just concluding that we are going to use these facilities because that's what I want to use, or that's what we as a technical team want to use on this.
And so it's got to be Kafka, and it's got to be Kubernetes, and it's got to be all this stuff. For example, go to the Kubernetes leaders and they'll be the first to tell you, you may not want to be using Kubernetes just because it's out there. There is a price to pay for going there. And it does a lot of great things. I don't want to be misunderstood that I'm knocking on Kafka or Kubernetes. But hey, make those decisions when you need to. That's purpose-driven. It's a responsibility, right? It's showing that you're responsible for your decisions and that you're making decisions based on what you are working on.
James Higginbotham: So to kind of tie that back, we talked about events and event-driven architecture and thinking about things in that way to introduce resiliency, overcome latency issues, and handle long-running transactions or workflows, those kinds of things, as well as many other scenarios. It sounds like then what you're saying is purposeful architecture is really about what does the business need or what would it benefit from? And so, therefore, would it benefit the business for me to perhaps select a different approach than maybe something that I would normally use because there's a lot more benefit to the business as a result? So maybe event-driven would be beneficial in some cases, in many cases, perhaps, as a result of that. Is that what you're saying here with purposeful architecture?
Vaughn Vernon: Yes, absolutely. But also, maybe event-driven is the very wrong thing to use in some cases, right? If it's a simple web app that is just collecting data and somehow you can access that data in batch form or, you know, simple queries, maybe that's all that you really need to do. So, yeah, it's just really being responsible, making decisions based on the real needs not because your mind is already made up on how this is going to work.
James Higginbotham: Yes, those automatic decisions that we go to just make because we've done it before, let's do it again. We achieved success before, so that must automatically mean we're gonna achieve success again. When what you're saying is slow down and be very thoughtful about an intentional, rather than just defaulting to either something new that's emerging or something that we've done in the past that has been successful. Let's reason through this and determine the right path to the benefit.
Vaughn Vernon: In Chapter 8 of the book, we start just kind of, like, taking some previous definitions of software architecture, what is a software architecture, right? It seems like every book on architecture has to try to define what software architecture is. But we show what others have said about software architecture, they're generally quite brief. We wrote two small paragraphs about software architecture. The thought that we're trying to convey is you cannot decide what your architecture should be until you've had communication. The communication is going to tell you the kinds of software, you know, modeling that you're going to do, the kinds of components that will need to be created. And that might lead to, you know, one architecture decision or another.
There are sort of some go-to architecture patterns or styles that I think are generally always useful, but these don't impose a certain sort of way of using distributed computing. I would say the hexagonal or ports and adapters architecture is just sort of like a smart way to go. And that's far from setting things in stone. It's actually making the software much more flexible and capable of introducing new mechanisms, new kinds of databases, messaging, REST, rich GUI, or whatever it happens to be, right, you can make those decisions as you realize you need them.
If you look at architecture decision records, that is what they actually are. You're making an architecture decision at the last responsible moment, right? You're delaying those things purposely because you don't know yet if you need them. Or even you might know that you need some kind of thing, but should it be GraphQL? Or should it be an API gateway with different queries happening, you know, for a single endpoint? Or do you allow GraphQL to handle that for you? I don't know, I can't tell you right now because we haven't spoken enough about it. That's the real answer.
James Higginbotham: And that's fantastic. I know, when I first started, when I was a senior developer and I was being challenged by someone who was, you know, encouraging me to grow in my career. One of the things they recommended is they gave me a list of books they gave me a stack of books. He said, "Here's a stack of books, I want you to read these." And some of these are books you'd recognize and some less so. But he said, "You know, I think you need to start getting into software architecture, more really start thinking at a higher level about how all the components communicate, how the systems communicate between one another. This is going to stretch you and help you to grow as a developer and go in the direction that you feel like your career's taking you." And so I sat down, and he must have given me six or seven books, and not one of the defined architecture or gave me kind of a working model of how do I think about this? I sort of had to derive my own perspective from it until I had time to figure it out.
So it's exciting to me that you've devoted some time in Chapter 8 to that. That other coming along, whether they just need a refresher, or they want to see a different way than it's defined or described, or that junior developer that's starting to try to figure out "What is this architecture thing? I've been coding for a while now." Maybe they've progressed up in there and they're starting to grow in their career and they're trying to understand it. So that's exciting to me to have that as part of the book as well.
Vaughn Vernon: Yes. And just to make a point. I mean, I think that software architecture has been defined by some very smart and experienced people as, like, the parts of the software that are hard to change. And you know what, that shouldn't be true. If you're using the ports and adapters architecture, you might just find that it might not be really easy to change, but it's going to be simpler to introduce or reintroduce different mechanisms that you found were insufficient for whatever reason, you know.
Say that you were using REST as a way to query logs of events, right, that other outside subsystems could do that, and you decide, "Well, we're now going to use Kafka. Instead, we're going to use, you know, Azure message bus or whatever it happens to be, SNS, SQS from AWS. And so we're now going to make that available." Well, the API hopefully doesn't have to change, or it doesn't have to change much. And you can completely change that out, that mechanism, the approach to doing that without, you know, a lot of overhead. I wouldn't want to automatically conclude that software architecture is about the things that are hard to change. Although I respect that at one time, that may have been the case.
Recommendations for refining your monoliths or microservices
James Higginbotham: So, as we wind down, I have two things to discuss to wrap this discussion up. The first thing is for those that are faced today with some monoliths that are less than ideal, whether they're a big ball of mud or they're just, you know, really difficult to work with, and they're thinking about microservices, or maybe they're using microservices and they've encountered some challenges or something else. What are some recommendations that your book provides for how to step through refining your monoliths and microservices to make them a bit more resilient and adaptable, how to make it easier to work with, how to perhaps either transition to microservices or to make it a bit less of a distributed monolith and more, you know, of an independent series of services that can communicate together? How do you or how does your book approach guiding teams that are likely in that kind of situation?
Vaughn Vernon: Yes. Probably the quickest way to explain it is to say, if you're using domain-driven design, think of either a monolithic module as being a bounded context, or think of a microservice as being a bounded context, right? As in this is the way that we deploy things. So in a monolith, if we have seven bounded contexts, we have seven major modules within it. And we create the same kind of boundaries where we don't want another module within the monolith reaching inside either into our database directly forbid, right? But also just using our domain model directly, we create the same kinds of interfaces that we would if we were on a network, but it's just that we don't have to pay the price of the network and the potential failures that the network is going to sometimes require you to deal with.
So that's sort of the first guidance, and then being purposeful. And let me just say one more thing about that. The reason is that you can develop much more rapidly if you are using monoliths initially, because you're not dealing with the failures of the network, and you're not dealing with all the complexities of distributed computing. So use a monolith very early, you might learn that the monolith is just what you need long-term. But if you need to, for whatever reason, rate of change, or some scalability issue, or something, if you need to break a module out of that monolith and make it a microservice, well, again, that's a pretty simple, I would hope, step to achieve. But as soon as you start tangling things up and making one module deeply connected to another one, forget about that. You've already lost that opportunity unless you do a lot of rework.
So that's what we're saying is start with a monolith, introduce microservices as you realize that you need them. But even if you know right away that you are going to use microservices you're better off running everything in a sort of single VM with, you know, multiple threads across bounded contexts or microservices. Because, again, you're just not going to deal with the kinds of complexity that you can spend days or weeks just trying to figure out a silly networking kind of thing. And that just derails your business value production, right? So start with the sort of uncool way of doing things but recognize that in the sandbox that you've created for yourself, you're going to have a lot of traction, you're going to move quickly. And then when the distributed computing part of it becomes a responsible thing to do with a purpose behind it, then you make that decision to break them out.
James Higginbotham: That's great advice. So let's wind this down. I hear that you're working on a second book in the series. Can you tell us a little bit about it, give us a little bit of a preview or hints about it?
Upcoming book: hands-on coding for purposeful architectures
Vaughn Vernon: Yes. And I have to say, I thought that we would be able to write the second book within a very short time after the first book was done. Because I got the series right after the lockdown from COVID and everything just stopped. So here's this great opportunity to write a couple of books. Well, things started getting busy again sooner than I expected. So we're about we're quite a bit behind in the second book. So I'll just preface that. We hope to have it out earlier.
But the second book is read the first book and everything that you're wondering about, how would I actually do that in code? How would I do that with an object-oriented language? How would I do that with a functional programming language? How would I do that with, you know, various architectural mechanisms, architecture patterns, and so forth? How would I go about doing that? Well, that's what we are diving into with the second book, implementing strategic monoliths and microservices. And so, if you know "Implementing Domain-Driven Design," my "Red Book," well, you're going to see a similar reflection, different code, of course, different examples, but you're going to see that kind of balance between content prose and, you know, source code to a moderate extent, with full examples available on GitHub afterward in polyglot, right, multiple programming languages.
James Higginbotham: Oh, that's great. I can see why that's gonna take a bit of time. I mean, that's a lot of work ahead of you, but it's gonna be really valuable and a great complement to your book. Well, thanks, Vaughn, for coming on and speaking about your book. Again, the book title is "Strategic Monoliths and Microservices: Driving Innovation Using Purposeful Architecture." Vaughn, it's been great chatting with you, and I appreciate you taking the time.
Vaughn Vernon: Sure, absolutely. Thank you, James Higginbotham. Thank you GOTO.